10248

Download the full version of the ebook at
https://ebookfinal.com
Quadratic mappings and Clifford algebras 1st

Edition Jacques Helmstetter
https://ebookfinal.com/download/quadratic-
mappings-and-clifford-algebras-1st-edition-
jacques-helmstetter/
Explore and download more ebook at https://ebookfinal.com

Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.
Clifford Algebras Applications to Mathematics Physics and

Engineering 1st Edition Carlos A. Berenstein
https://ebookfinal.com/download/clifford-algebras-applications-to-
mathematics-physics-and-engineering-1st-edition-carlos-a-berenstein/
ebookfinal.com
Quasiconformal Mappings Riemann Surfaces and Teichmuller

Spaces Ams Special Session in Honor of Clifford J Earle
October 2 3 2010 Syracuse Syracuse New York Yunping Jiang
https://ebookfinal.com/download/quasiconformal-mappings-riemann-
surfaces-and-teichmuller-spaces-ams-special-session-in-honor-of-
clifford-j-earle-october-2-3-2010-syracuse-syracuse-new-york-yunping-
jiang/
ebookfinal.com
Handbook of Conformal Mappings and Applications 1st

Edition Prem K. Kythe
https://ebookfinal.com/download/handbook-of-conformal-mappings-and-
applications-1st-edition-prem-k-kythe/
ebookfinal.com
Hyperbolic Manifolds and Holomorphic Mappings An

Introduction Second Edition Shoshichi Kobayashi
https://ebookfinal.com/download/hyperbolic-manifolds-and-holomorphic-
mappings-an-introduction-second-edition-shoshichi-kobayashi/
ebookfinal.com
Reflections on America by Jacques Maritain Jacques
Maritain
https://ebookfinal.com/download/reflections-on-america-by-jacques-
maritain-jacques-maritain/
ebookfinal.com
Elements of the Representation Theory of Associative

Algebras Volume 2 Tubes and Concealed Algebras of
Euclidean Type 1st Edition Daniel Simson
https://ebookfinal.com/download/elements-of-the-representation-theory-
of-associative-algebras-volume-2-tubes-and-concealed-algebras-of-
euclidean-type-1st-edition-daniel-simson/
ebookfinal.com
Re Mapping Archaeology Critical Perspectives Alternative

Mappings Mark Gillings
https://ebookfinal.com/download/re-mapping-archaeology-critical-
perspectives-alternative-mappings-mark-gillings/
ebookfinal.com
Monomial Algebras Second Edition Rafael Villarreal
https://ebookfinal.com/download/monomial-algebras-second-edition-
rafael-villarreal/
ebookfinal.com
Neuroradiology cases 1st Edition Clifford J. Eskey
https://ebookfinal.com/download/neuroradiology-cases-1st-edition-
clifford-j-eskey/
ebookfinal.com
Quadratic mappings and Clifford algebras 1st Edition
Jacques Helmstetter Digital Instant Download
Author(s): Jacques Helmstetter, Artibano Micali
ISBN(s): 9783764386061, 3764386061
Edition: 1
File Details: PDF, 3.18 MB
Year: 2008
Language: english
Jacques Helmstetter
Artibano Micali
Quadratic Mappings
and Clifford Algebras
Birkhäuser
Basel · Boston · Berlin
Authors:
Jacques Helmstetter Artibano Micali

Institut Fourier (Mathématiques) Département des Sciences mathématiques
Université Grenoble I Université Montpellier II
B.P. 74 Place Eugène Bataillon, CC 051
38402 Saint-Martin d’Hères Cedex 34095 Montpellier Cedex 5
France France
e-mail: [email protected] e-mail: [email protected]
2000 Mathematical Subject Classification: 15A66, 15A69, 15A63, 15A75, 16H05
Library of Congress Control Number: 2007942636
Bibliographic information published by Die Deutsche Bibliothek. Die Deutsche Bibliothek lists
this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in
the Internet at http://dnb.ddb.de
ISBN 978-3-7643-8605-4 Birkhäuser Verlag AG, Basel - Boston - Berlin
This work is subject to copyright. All rights are reserved, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, re-use of illustrations,
recitation, broadcasting, reproduction on microfilms or in other ways, and storage in data
banks. For any kind of use permission of the copyright owner must be obtained.
© 2008 Birkhäuser Verlag AG

Basel · Boston · Berlin
P.O. Box 133, CH-4010 Basel, Switzerland
Part of Springer Science+Business Media
Printed on acid-free paper produced from chlorine-free pulp. TCF∞
Printed in Germany
ISBN 978-3-7643-8605-4 e-ISBN 978-3-7643-8606-1
987654321 www.birkhauser.ch
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1 Algebraic Preliminaries
1.1 Some general notation and definitions . . . . . . . . . . . . . . . . 1
1.2 Universal objects in a category . . . . . . . . . . . . . . . . . . . . 2
1.3 Examples of universal objects . . . . . . . . . . . . . . . . . . . . . 4
1.4 Tensor algebras and symmetric algebras . . . . . . . . . . . . . . . 8
1.5 Functors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 Exact sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Projective modules and flat modules . . . . . . . . . . . . . . . . . 14
1.8 Finitely presented modules . . . . . . . . . . . . . . . . . . . . . . 15
1.9 Changes of basic rings . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.10 Rings and modules of fractions . . . . . . . . . . . . . . . . . . . . 21
1.11 Localization and globalization . . . . . . . . . . . . . . . . . . . . . 25
1.12 Finitely generated modules . . . . . . . . . . . . . . . . . . . . . . 30
1.13 Some applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2 Quadratic Mappings
2.1 Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.2 Changes of basic ring and localizations . . . . . . . . . . . . . . . . 57
2.3 Nondegenerate quadratic mappings . . . . . . . . . . . . . . . . . . 59
2.4 Operations on quadratic mappings and symmetric
bilinear mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.5 Hyperbolic and metabolic spaces . . . . . . . . . . . . . . . . . . . 67
2.6 Orthogonal decompositions of quadratic spaces . . . . . . . . . . . 72
2.7 Witt rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.8 Examples of Witt rings . . . . . . . . . . . . . . . . . . . . . . . . 83
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
vi Contents
3 Clifford Algebras
3.1 Definitions and elementary properties . . . . . . . . . . . . . . . . 105
3.2 The parity grading of Clifford algebras . . . . . . . . . . . . . . . . 112
3.3 Clifford algebras of free modules of rank 2 . . . . . . . . . . . . . . 117
3.4 Graded quadratic extensions . . . . . . . . . . . . . . . . . . . . . . 122
3.5 Graded Azumaya algebras . . . . . . . . . . . . . . . . . . . . . . . 131
3.6 Traces and determinants . . . . . . . . . . . . . . . . . . . . . . . . 144
3.7 Clifford algebras of quadratic spaces . . . . . . . . . . . . . . . . . 149
3.8 Discriminant modules, quadratic extensions and
quaternion algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
4 Comultiplications. Exponentials. Deformations
4.1 Coalgebras and comodules . . . . . . . . . . . . . . . . . . . . . . . 175
4.2 Algebras and coalgebras graded by parities . . . . . . . . . . . . . 181
4.3 Exterior algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
4.4 Interior products in Clifford algebras . . . . . . . . . . . . . . . . . 190
4.5 Exponentials in even exterior subalgebras . . . . . . . . . . . . . . 195
4.6 Systems of divided powers . . . . . . . . . . . . . . . . . . . . . . . 199
4.7 Deformations of Clifford algebras . . . . . . . . . . . . . . . . . . . 201
4.8 Applications of deformations . . . . . . . . . . . . . . . . . . . . . 210
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
5 Orthogonal Groups and Lipschitz Groups
5.1 Twisted inner automorphisms and orthogonal groups . . . . . . . . 232
5.2 Filtrations of Clifford algebras . . . . . . . . . . . . . . . . . . . . . 243
5.3 Lipschitz monoids and derived groups . . . . . . . . . . . . . . . . 249
5.4 The invariance property . . . . . . . . . . . . . . . . . . . . . . . . 258
5.5 Associated Lie algebras . . . . . . . . . . . . . . . . . . . . . . . . 261
5.6 First results about orthogonal transformations . . . . . . . . . . . 267
5.7 Products of reflections when K is a local ring . . . . . . . . . . . . 272
5.8 Further results about orthogonal transformations . . . . . . . . . . 281
5.9 More information about exterior algebras . . . . . . . . . . . . . . 287
5.10 The Lipschitz monoid Lip(M ) when K is a field . . . . . . . . . . . 296
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
6 Further Algebraic Developments
6.1 Modules over a noncommutative algebra . . . . . . . . . . . . . . . 323
6.2 Graded modules over a graded algebra . . . . . . . . . . . . . . . . 327
6.3 Graded semi-simple modules . . . . . . . . . . . . . . . . . . . . . 334
6.4 Graded Morita theory . . . . . . . . . . . . . . . . . . . . . . . . . 337
6.5 Graded separable algebras . . . . . . . . . . . . . . . . . . . . . . . 344
Contents vii
6.6 Graded central simple algebras over a field . . . . . . . . . . . . . . 353

6.7 More information about graded Azumaya algebras . . . . . . . . . 358
6.8 Involutions on graded central simple algebras . . . . . . . . . . . . 365
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
7 Hyperbolic Spaces
7.1 Some representations of Clifford algebras . . . . . . . . . . . . . . . 391
7.2 The Cartan–Chevalley mapping . . . . . . . . . . . . . . . . . . . . 395
max
7.3 The bijection C(V ) ⊗ (U ) ⊗ C(V ) → C(M ) . . . . . . . . . 403
7.4 The Cartan–Chevalley criterion . . . . . . . . . . . . . . . . . . . . 409
7.5 Applications to Lipschitz monoids . . . . . . . . . . . . . . . . . . 414
7.6 Applications to totally isotropic direct summands
of maximal rank . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
8 Complements about Witt Rings and Other Topics
8.1 Witt rings over local rings when 2 is invertible . . . . . . . . . . . 439
8.2 Continuation when 2 is not invertible . . . . . . . . . . . . . . . . . 445
8.3 Witt rings of Prüfer rings . . . . . . . . . . . . . . . . . . . . . . . 453
8.4 Quadratic forms in characteristic 2 . . . . . . . . . . . . . . . . . . 461
8.5 Clifford algebras in characteristic 2 . . . . . . . . . . . . . . . . . . 464
8.6 The group of classes of Clifford algebras . . . . . . . . . . . . . . . 468
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
Bibliography
Books and booklets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
Index of Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Index of Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
Introduction
The first purpose of an introduction is to explain what distinguishes the newly
written book from other books that might as well have the same title. This book
deals with quadratic mappings between modules over an arbitrary ring K (com-
mutative, associative, with unit element); therefore it requires an effective mastery
of some little part of commutative algebra. It is especially interested in quadratic
forms and in their Clifford algebras. The most common object under consideration
is a quadratic module (M, q), that is any module M provided with a quadratic form
q : M → K, and the deepest results are obtained when (M, q) is a quadratic space,
in other words, when M is a finitely generated projective module and q induces a
bijection from M onto the dual module M ∗ . In particular the study of Clifford al-
gebras of quadratic spaces shall (very progressively) lead to sophisticated theories
involving noncommutative algebras over the ring K (Azumaya algebras, Morita
theory, separability).
This book is almost never interested in results that would follow from some
special properties of the basic ring K; therefore much more emphasis has been put
on a serious study of Clifford algebras than on sophisticated properties of quadratic
forms which always depend on subtle hypotheses on the ring K. Here, when K is
not an arbitrary ring, it is a local ring, or even a field; the consideration of such
particular rings is justified by the importance of localization and globalization in
many chapters, and the important role of residue fields at some critical moments.
Besides, many useful applications of Clifford algebras outside mathematics involve
quadratic spaces over fields.
Another essential feature of this book is the narrowness of the set of pre-
requisites, and its constancy from the beginning to the end. These prerequisites
are made precise below, and although they are not elementary, they are much less
difficult and fewer than would be required for a pioneering or scholarly work. All
essential properties of Clifford algebras have been reached by elementary means
in the first five chapters before more difficult theories are presented in Chapter 6.
The concern of the authors about teaching has led them to limit the amount of
prerequisites, and to prove all results in the core of the book (almost the whole
book) on the basis of these prerequisites; for all these results the complete path
leading to their proof (sometimes by new simpler means) is explained.
Of course it has not been possible to impose the above-mentioned features on
the whole book. We thought it sensible to present interesting examples involving
theories outside the scope of the book, and to give information about related topics
which do not appear in the core of the book. Thus for the proof of several state-
ments it has been necessary to refer to other publications. For instance, quadratic
forms over the ring of integers often afford illuminating applications of general
theories; but since this book does not deal with arithmetic, it just mentions which
arithmetical knowledge is indispensable.
x Introduction
Readers are assumed already to know elementary algebra (rings, fields,

groups, quotients,. . . ), and also linear and multilinear algebra over fields, espe-
cially tensor products and exterior algebras. They are assumed to know the usual
properties of quadratic forms over the usual fields R and C, which should enable a
rapid understanding of the properties of more general quadratic mappings. Even
some knowledge of linear algebra over rings (over commutative, associative rings
with unit) is required: exact sequences, projective and flat modules,. . . . Most of
these prerequisites are briefly recalled, especially in Chapter 1. A self-contained
yet concise exposition of commutative algebra is provided; it only covers the small
part that is needed, essentially localization and globalization, and finitely gener-
ated modules. Homological algebra is never involved, except in isolated allusions.
Many pages are devoted to “exercises”; their purposes are varied. Some of
them are training exercises, in other words, direct applications. Others present
still more results, which have seemed less important to the authors, but which
nevertheless deserve to be stated with indications about how to prove them. Others
present examples enlightening the reader on some particular features or some
unexpected difficulties. There are also developments showing applications in other
domains, and some few extracts from the existing literature. The levels of difficulty
are varied; when an exercise has seemed to be very difficult, or to require some
knowledge that is not treated in the book, an asterisk has been put on its number,
and often a hint has been supplied.
In the opinion of the authors, many applications of Clifford algebras outside
algebra, and even outside mathematics, raise problems that are universally inter-
esting, even for algebraists. It is the duty of algebraists to find clear concepts and
effective treatments, especially in places that are usually obscured by a lot of cum-
bersome calculations. In many applications of Clifford algebras there are interior
multiplications; here (in Chapter 4) it is explained that they can be derived from
the comultiplication that makes every Clifford algebra become a comodule over the
exterior algebra (treated as a coalgebra). In many applications of Clifford algebras
the calculations need two multiplications, a Clifford multiplication and an exterior
one; here (in Chapter 4) this practice is related to the concept of “deformation of
Clifford algebra”, which allows an elaborate presentation of a well-known result
stated for instance in [Chevalley 1954], §2.1, and with more generality in [Bourbaki
1959, Algèbre, Chap. 9] (see Proposition 3 in §9, no 3). But the true meaning of
this essential result only appears when it is stated that it gives isomorphisms of
comodules over exterior algebras, and not merely isomorphisms of K-modules.
Spinor spaces in quantum mechanics raise problems for which insightful alge-
braic interpretation and smart proof eschewing tedious calculations are still objects
of discussion. Spinor spaces are often said to be Clifford modules although they
are actually graded Clifford modules (see Example (6.2.2) in this book); the word
“graded” refers to a parity grading which distinguishes even and odd elements.
Whereas the theory of Clifford modules is a long sequence of particular cases,
graded Clifford modules come under a unified and effective theory. The last ex-
Introduction xi
ercises of Chapter 6 propose a smart and effective path to the essential algebraic
properties that are needed in quantum mechanics.
This comment about spinor spaces is just one example of the constant em-
phasis put on parity gradings (from Chapter 3 to the end), in full agreement with
C.T.C. Wall and H. Bass. In many cases the reversion of two odd factors must
be compensated by a multiplication by −1, and here this rule is systematically
enforced in all contexts in which it is relevant; indeed only a systematic treatment
of parity gradings can avoid repeated hesitations about such multiplications by
−1. For instance if f and g are
2 linear forms on M , their exterior product can be
defined as the linear form on (M ) that takes this value on the exterior product
of two elements x and y of M :
(f ∧ g)(x ∧ y) = −f (x) g(y) + f (y) g(x) ;
the sign − before f (x)g(y) comes from the reversion of the odd factors g and x;
but in f (y)g(x) the odd factor y has jumped over two odd factors x and g, whence
the sign +.
Lipschitz, the forgotten pioneer

Rudolf O.S. Lipschitz (1832–1903) discovered Clifford algebras in 1880, two years
after William K. Clifford (1845–1879) and independently of him, and he was the
first to use them in the study of orthogonal transformations. Up to 1950 peo-
ple mentioned “Clifford-Lipschitz numbers” when they referred to this discovery
of Lipschitz. Yet Lipschitz’s name suddenly disappeared from the publications in-
volving Clifford algebras; for instance Claude Chevalley (1909–1984) gave the name
“Clifford group” to an object that is never mentioned in Clifford’s works, but stems
from Lipschitz’s. The oblivion of Lipschitz’s role is corroborated by [Weil], a letter
that A. Weil first published anonymously, probably to protest against authors who
discovered again some of Lipschitz’s results in complete ignorance of his priority.
Pertti Lounesto (1945–2002) contributed greatly to recalling the importance of
Lipschitz’s role: see his historical comment in [Riesz, 1993].
This extraordinary oblivion has generated two different controversies, a his-
torical one and a mathematical one. On one side, some people claimed that the
name “Clifford group” was historically incorrect and should be replaced with “Lip-
schitz group”; their action at least convinced other mathematicians to make correct
references to Lipschitz when they had to invent new terms for objects that still
had no name, even when they were reluctant to forsake the name “Clifford group”.
On the other side, some people were not satisfied with Chevalley’s presentation
of the so-called Clifford group, and completed it with additional developments
that meant a return to Lipschitz’s ideas; this is especially flagrant in [Sato, Miwa,
Jimbo 1978], where the authors discovered again some of Lipschitz’s results and
gave them much more generality and effectiveness; the same might be said about
[Helmstetter 1977, 1982]; but since the Japanese team showed applications of his
xii Introduction
cliffordian ideas to difficult problems involving differential operators (the “holo-

nomic quantum fields”), the necessity of going beyond Chevalley’s ideas became
obvious for external reasons too. The fact that all these authors at that time com-
pletely ignored Lipschitz’s contribution proves that the mathematical controversy
is independent of the historical one.
The part of this book devoted to orthogonal transformations can be under-
stood as a modernization of Lipschitz’s theory. Whereas Lipschitz only considered
real positive definite quadratic forms for which Clifford–Lipschitz groups may look
quite satisfying, with more general quadratic forms it becomes necessary to attach
importance to “Lipschitz monoids” from which “Lipschitz groups” are derived.
Here the historically incorrect “Clifford groups” are still accepted (with the usual
improved definition that pays due attention to the parity grading), but they only
play an incidental role. They coincide with the Lipschitz groups in the classical
case of quadratic spaces; but when beyond this classical case Clifford groups and
Lipschitz groups no longer coincide, the latter prove to be much more interesting.
Thus the mathematical controversy happens to prevail over the historical one.
Contraction and expansion are opposite and equally indispensable stages in
all scientific research. At Chevalley’s time it was opportune to contract the argu-
ments and to exclude developments that no longer looked useful; but in Chevalley’s
works there is at least one part (in [Chevalley 1954], Chapter 3) that should have
led him to reinstate Lipschitz if he had continued developing it. Our Chapter 7 is an
expansion of this part of Chevalley’s work, which for a long time has remained as
he left it. This expansion involves the contributions of both Lipschitz and Cheval-
ley, and should give evidence that it is much better to accept the whole heritage
from all pioneers without prolonging inopportune exclusions. Besides, Lipschitz’s
ideas also proved to be very helpful in the cliffordian treatment of Weyl algebras.
Weyl algebras
Weyl algebras represent for alternate bilinear forms the same structure as Clifford
algebras for quadratic forms, and in some publications they are even called “sym-
plectic Clifford algebras”. In [Dixmier 1968] you can find a concise exposition of
what was known about them before cliffordian mathematicians became interested
in them. Revoy was probably the first to propose a cliffordian treatment of Weyl
algebras; see [Nouazé, Revoy 1972] and [Revoy 1978]. Later and independently,
Crumeyrolle in France and the Japanese team Sato–Miwa–Jimbo produced some
publications developing the cliffordian treatment of Weyl algebras, although they
ignored (at least in their first publications) that these algebras had been already
studied, and had received H. Weyl’s name. Revoy’s isolated work was hardly no-
ticed, the cliffordian ideas of the Japanese team (which the renewal of Lipschitz’s
ideas mentioned above) were inserted in a very long and difficult work devoted to
differential operators, which discouraged many people, and Crumeyrolle’s state-
ments bumped up against severe and serious objections. That is why the cliffordian
treatment of Weyl algebras has not yet won complete acknowledgement.
Introduction xiii
There is no systematic presentation of Weyl algebras in this book, which

already deals with a large number of other subjects. But at the end of Chapters 4,
5 and 7, many exercises about them have been proposed; Weyl algebras are defined
in (4.ex.18). These exercises explain the cliffordian treatment of Weyl algebras as
long as it is an imitation, or at least an adaptation, of the analogous treatment of
Clifford algebras. For the most difficult results that require Fourier analysis and
related theories, a short summary has been supplied; it should help readers to
understand the purposes and the achievements of this new theory.
Acknowledgements
A. Micali is pleased to recall that he has worked for a long time with Orlando E.
Villamayor (1923–1998), who efficiently contributed with some other authors to
much progress that is now common knowledge and presented as such in this book.
J. Helmstetter declares that he is indebted to Chevalley for his interest in Clifford
algebras. Both authors have also learned much from H. Bass’s publications. For
some difficult topics we have also consulted [Knus 1991].
During the writing of our text, we took advantage of the services of the Math-
ematical Department (Fourier Institute) of the University of Grenoble; our text
was prepared in this Institute and its Library was very often visited. Therefore we
are grateful to the Director and to the Librarian for their help. Several colleagues
in this Institute suggested some good ideas, or tried to answer embarrassing ques-
tions; they also deserve our gratitude.
The authors also thank the Director and the Librarian of the Mathemati-
cal Department of the University of Montpellier, who always gave us a helpful
welcome. Our colleague Philippe Revoy followed our work and gave much good
advice, for which we are indebted to him.
The authors are also grateful to Max-Albert Knus (Zürich, ETH-Zentrum)
for his critical reading of a large part of the book, and for his judicious suggestions.
Finally we thank all the persons that have worked on our text to improve its
literary quality and its presentation; despite our efforts, we yet needed their help.
Without the benevolence of the editorial board of Birkhäuser Verlag, our project
would not have been completed.
Grenoble, September 2007.
Jacques Helmstetter Artibano Micali

Université Grenoble I Université Montpellier II
Institut Fourier (Mathématiques) Département des Sciences
B.P. 74 Mathématiques
F-38402 Saint-Martin d’Hères Place Eugène Bataillon
France F-34095 Montpellier Cedex 5
France
Chapter 1
Algebraic Preliminaries
This preliminary chapter is devoted to the following three subjects, which to-
gether allows us to review a great part of the prerequisites, and to add some more
specialized knowledge:
(a) a very simple presentation of the notion of universal property, with many
examples; Sections 1.2 to 1.4 are devoted to this subject.
(b) additional information about categories of modules; Sections 1.5 to 1.9 con-
tain reminders about exact sequences, usual functors, projective modules and
changes of basic rings; but in 1.8 finitely presented modules are treated with
more detail.
(c) a self-contained presentation of rings and modules of fractions, localization
and globalization; although this material is already well treated in the existing
literature, a concise exposition of the exact part that is here actually useful
should prevent beginners from wandering in too-specialized topics.
Complete knowledge of all this chapter is not indispensable, because precise refer-
ences will always be given when the most difficult or specialized results are needed
in the following chapters.
1.1 Some general notation and definitions

The following notation and definitions will be used in all chapters.
As usual, N, Z, Q, R, C, H are respectively the set of all integers ≥ 0, the ring
of all integers, the fields of rational, real and complex numbers, and the division
ring of real quaternions.
A ring A is always an associative ring with a unit element 1A unless otherwise
stated; the group of all invertible elements of A is denoted by A× . If a mapping
f : A → B between two rings is called a ring morphism, it must be understood
2 Chapter 1. Algebraic Preliminaries
that the equality f (1A ) = 1B is also required. By definition a subring of A must

contain 1A .
Every module over the ring A is a left module, unless it is stated that it is
a right module; the category of all left A-modules is denoted by Mod(A). Right
A-modules are often treated as left modules over the opposite algebra Ao (defined
in 3.1) or over the twisted opposite algebra Ato (defined in 3.2) and therefore their
category is denoted by Mod(Ao ) or Mod(Ato ). If M and N are two modules over
A, the set of all A-linear mappings from M into N is denoted by HomA (M, N ); the
ring of all A-linear endomorphisms of M is denoted by EndA (M ). When A = Z,
Mod(Z) is also the category of all additive groups.
The letter K always refers to a commutative ring, the unit element of which
is denoted by 1; it is silently assumed that K is not reduced to 0 (in other words,
1 = 0). When gradings (or filtrations) get involved, K is always trivially graded (or
filtered); in other words, all its nonzero elements have degree 0. At some places it
will be assumed that K is a local ring (in which there is only one maximal ideal m)
or even a field. When there is no mention of another ring, all algebraic notions such
as linearity, tensor products,.
. . , refer to this ring K; consequently notation like
Hom(M,
N ), M ⊗ N , (M ),. . . , must be understood as HomK (M, N ), M ⊗K N ,
K (M ),. . . .
Unless otherwise stated, an algebra A over K is a ring provided with a ring
morphism K → A that maps K into the center of A. If this ring morphism is
injective, K can be identified with a subring of A, and thus 1A is identified with
1. Every ring is an algebra over the ring Z of integers. The category of all K-
algebras is denoted by Alg(K); thus the notation HomAlg(K) (A, B) means the set
of all K-linear ring morphisms from A into B. The subcategory of all commutative
algebras in Alg(K) is denoted by Com(K).
Internal references must be understood in this way: the notation 4.2 means
the second section of the fourth chapter; inside each section, all emphasized state-
ments or formulas are numbered in a single file; for instance (4.2.3) means the third
statement (theorem, or definition, or remark, or example, or formula, or anything
else) in 4.2. References to other works (listed in the bibliography at the end of the
book) are indicated by brackets: for instance [Lounesto 1981].
1.2 Universal objects in a category

There are different ways to introduce universal properties, but here the simplest
way, based on the notion of universal object in a category, is already sufficient.
A category C consists of objects and morphisms (also called homomorphisms or
arrows); each morphism relates two objects, called the source and the target; the
set of morphisms relating the objects M and N in the category C is denoted by
HomC (M, N ). A morphism f from M to N and a morphism g from N to P can be
linked together to give a morphism from M to P denoted by g ◦ f or gf , and it is
1.2. Universal objects in a category 3
required that the equality h(gf ) = (hg)f is true whenever it is meaningful (in other
words, whenever the targets of f and g are respectively the sources of g and h). In
the definition of a category is also mentioned the existence of an identity morphism
idN for each object N , with the requirement that the equalities idN f = f and
g idN = g hold whenever they are meaningful. A morphism f : M → N is called
an isomorphism if there exists g : N → M such that gf = idM and f g = idN ;
this morphism g is unique and is called the reciprocal isomorphism.
An object U in a category C is called an initial universal object (resp. a
final universal object) if for every object M in C the set HomC (U, M ) (resp.
HomC (M, U )) contains exactly one element. This definition implies that idU is
the only element in HomC (U, U ). Here almost all universal objects under consid-
eration will be initial ones. We must keep in mind the following evident theorem.
(1.2.1) Theorem. If a category contains two initial universal objects (resp. two final
universal objects) U and V , the only morphism from U to V is an isomorphism.
Proof. There is one morphism f from U to V and also one morphism g from V to
U ; since gf (resp. f g) is a morphism from U (resp. V ) to itself, it must be equal
to idU (resp. idV ); therefore f and g are reciprocal isomorphisms.
The most evident category is the category of all sets in which any mapping
is a morphism; the empty set is the only initial universal object, and any set
containing exactly one element is a final universal object.
In the category Mod(K) of all modules over K, the morphisms are the K-
linear mappings; the notation Hom(M, N ) (or HomK (M, N ) when more precision
is necessary) is the usual abbreviation for HomMod(K) (M, N ). In this category any
module reduced to zero is both an initial universal object and a final one.
Now let us consider the category Alg(K) of all algebras over the ring K, in
which the morphisms are the K-linear ring morphisms. The algebras reduced to
zero are the only ones in which the unit element is equal to 0, and they are final
universal objects. Since the morphisms must respect the unit elements (see 1.1),
the ring K itself is an initial universal object. The category Alg(Z) is the category
of all rings; for any ring A, the only morphism Z → A determines the characteristic
of A : it is the integer n ≥ 0 such that nZ is the kernel of this morphism.
Before more interesting examples are given in the next two sections, it must
be recalled that sometimes the objects of a category are morphisms in another
category. In particular, with each category C is associated a category Chom in
which the objects are all the morphisms of C; if u : M1 → M2 and v : N1 → N2
are two objects of Chom , a morphism from u to v is a couple (f1 , f2 ) of morphisms
f1 : M1 → N1 and f2 : M2 → N2 such that vf1 = f2 u. The composition of two
morphisms (f1 , f2 ) and (g1 , g2 ) in Chom is given by the formula (g1 , g2 ) ◦ (f1 , f2 ) =
(g1 f1 , g2 f2 ) whenever it is meaningful. If U is an initial universal object in C, then
idU is an initial universal object in Chom .
Sometimes it is convenient to associate with a given category C a dual category
C ∗ containing the same objects and the same morphisms; but the target (resp. the
source) of a morphism in C ∗ is by definition its source (resp. its target) in C.

If gf : M → N → P is a product of morphisms in C, it must be written as
f g : P → N → M in C ∗ . Every final universal object in C is an initial one in C ∗ ,
and this explains why both kinds of universal objects are considered as dual to
each other.
1.3 Examples of universal objects

Later in 3.1 Clifford algebras are introduced by means of a universal property,
meaning that they afford an initial universal object in some category; because of
(1.2.1), this property characterizes a Clifford algebra up to isomorphy. To prove
that a universal property can often distinguish an interesting object, we show
several already known objects, the interest of which actually depends on some
universal property.
Quotient modules
Let M be a module over K, and N a submodule of M ; when the quotient module
M/N and the quotient morphism ϕ : M → M/N are involved in an argument,
the following proposition is often referred to.
(1.3.1) Proposition. For every linear mapping f : M → P such that f (N ) = 0,
there exists a unique linear mapping f : M/N → P such that f = f ϕ. Moreover
if M is a K-algebra and N an ideal of M , then M/N is also a K-algebra and ϕ
is an algebra morphism.
This property of M/N and ϕ is said to be a universal property because it
means that ϕ is an initial universal object in the category of all linear mappings f
from M into any module P such that f (N ) = 0. If f : M → P and g : M → Q are
objects in this category, a morphism from f to g is a linear mapping u : P → Q
such that g = uf . When M is a K-algebra and N an ideal, we may require f and u
to be algebra morphisms. The proof of (1.3.1) is based on the fact that the image
and kernel of ϕ are exactly M/N and N ; therefore these properties characterize
M/N and ϕ up to isomorphy.
The null morphism M → 0 with a trivial target is obviously a final uni-
versal object, but quite uninteresting. In the following examples such trivial final
universal objects will not even be mentioned.
Freely generated modules

Let S be a set. The K-module freely generated by S is the set K (S) of all mappings
ε : S → K such that ε(s) vanishes for all s ∈ S except a finite number; it is a
K-module in an evident way; it is reduced to 0 if S is empty. Let ϕ be the mapping
from S to K (S) which maps each s ∈ S to the mapping es : S → K such that
1.3. Examples of universal objects 5
es (s) = 1 and es (t) = 0 whenever s = t. It is clear that K (S) is a free K-module

in which the family of all es is a basis; this property explains the name given to
K (S) . It also implies that ϕ is an initial universal object in the category of all
mappings f from S into any K-module P ; if f : S → P and g : S → Q are objects
in this category, a morphism from f to g is a linear mapping u : P → Q such
that g = uf . Indeed every mapping f from S into a K-module P determines a
unique linear mapping f from K (S) into P such that f = f ϕ, or equivalently,
f (es ) = f (s) for all s ∈ S.
Let P be any K-module, and S any subset of P ; the universal property of
K (S) implies the existence of a canonical linear mapping K (S) → P which maps
every es to s. It is surjective if and only if P is generated by S; therefore every
generating subset S makes P isomorphic to a quotient of the free module K (S) .
If P is finitely generated, it is isomorphic to the quotient of a free module with a
finite basis.
Tensor products of modules

Let M and N be K-modules. The canonical bilinear mapping ϕ from M × N into
the tensor product M ⊗ N (that is (x, y) −→ x ⊗ y) is also an initial universal
object because of the following universal property.
(1.3.2) Proposition. For any bilinear mapping f : M × N → P there exists a

unique linear mapping f : M ⊗ N → P such that f = f ϕ , or equivalently,
f (x ⊗ y) = f (x, y) for all (x, y) ∈ M × N .
The objects of the category under consideration are the bilinear mappings f
defined on M × N ; if f : M × N → P and g : M × N → Q are two objects, a
morphism from f to g is a linear mapping u : P → Q such that g = uf .
The precise definition of M ⊗ N is very seldom needed, since it is character-
ized by (1.3.2) up to isomorphy. For instance (1.3.2) is sufficient to explain that
two linear mappings v : M → M and w : N → N determine a linear mapping
v ⊗ w from M ⊗ M into N ⊗ N ; indeed v ⊗ w is the only linear mapping such
that (v ⊗ w)(x ⊗ y) = v(x) ⊗ w(y) for all (x, y) ∈ M × N .
Of course it is possible to state an analogous universal property for a ten-
sor product of several modules. The so-called commutativity property of tensor
products refers to the evident isomorphisms M ⊗ N ←→ N ⊗ M . There is also an
associativity property:
(M ⊗K N ) ⊗L P ∼
= M ⊗K (N ⊗L P );
it is valid when M is a K-module, P an L-module, and N a module over K and

L such that κ(λy) = λ(κy) for all κ ∈ K, λ ∈ L and y ∈ N .
Tensor products of algebras

When A and B are K-algebras, the tensor product A⊗B is also an algebra; indeed
the quadrilinear mapping (x, y, x , y ) −→ xx ⊗ yy from A × B × A × B into A ⊗ B
determines a linear mapping from A⊗B ⊗A⊗B into A⊗B; thus A⊗B is provided
with a multiplication such that
(x ⊗ y) (x ⊗ y ) = xx ⊗ yy ;
it is easy to prove that it is associative and that 1A ⊗ 1B is a unit element. Observe
that the mappings x −→ x⊗1B and y −→ 1A ⊗y are algebra morphisms ϕ1 and ϕ2
from respectively A and B into A ⊗ B, and that ϕ1 (x) and ϕ2 (y) always commute
in A ⊗ B :
(x ⊗ 1B )(1A ⊗ y) = x ⊗ y = (1A ⊗ y)(x ⊗ 1B ).
As an algebra, A ⊗ B has the following universal property.
(1.3.3) Proposition. If f1 : A → P and f2 : B → P are algebra morphisms, and
if f1 (x) and f2 (y) commute in the algebra P for all (x, y) ∈ A × B, there exists
a unique algebra morphism f : A ⊗ B → P such that f1 = f ϕ1 and f2 = f ϕ2 ,
or in other words, f1 (x) = f (x ⊗ 1B ) for all x ∈ A and f2 (y) = f (1A ⊗ y) for
all y ∈ B.
From (1.3.3) it should be easy to deduce the category in which (ϕ1 , ϕ2 ) is an
initial universal object. Nevertheless it has become usual to say that (1.3.3) is a
universal property of the target A ⊗ B.
Direct products and direct sums

With
every family (Mj )j∈J
of modules over K are associated a direct product
j Mj and a direct sum j Mj which coincide whenever the set J of indices is
finite; the direct product is the ordinary cartesian product consisting of all families
(xj ) such that xj ∈ Mj for all j ∈ J, whereas the direct sum is the submodule
of all families (xj ) in which all xj vanish except a finite number. Let us realize
that the direct product is the source of a final universal object in some category
C , whereas the direct sum is the target of an initial universal object in another
category C .
The objects of C are the families (fj ) of linear mappings P → Mj with
the same source P , and a morphism from (fj ) to (gj : Q → Mj ) is a morphism
u : Q → P such that gj = fj u forall j ∈ J. The following proposition means that
the family of all projections ψj : i∈J Mi → Mj is a final universal object.
of linear mappings fj : P → Mj determines a

(1.3.4) Proposition. Every family
unique linear mapping f : P → j Mj such that fj = ψj f for all j ∈ J; in other
words there is a canonical (and linear) bijection

Hom(P, Mj ) −→ Hom(P, Mj ), f −→ (ψj f )j∈J .
j j
1.3. Examples of universal objects 7
The objects of C are the families (fj ) of linear mappings Mj → P with

the same target P , and a morphism from (fj ) to (gj : Mj → Q) is a morphism
u : P → Q such that gj = ufj for all j ∈ J.The next proposition means that the
family of all natural injections ϕj : Mj → i Mi is an initial universal object.
(1.3.5) Proposition. Every family of linear mappings fj : Mj → P determines a
unique linear mapping f : j Mj → P such that fj = f ϕj for all j ∈ J; in other
words there is a canonical (and linear) bijection

Hom Mj , P −→ Hom(Mj , P ), f −→ (f ϕj )j∈J .
j j
For a family of two modules M and N the direct product and the direct sum
coincide; but the notation M ⊕ N is generally preferred when it is the source of
a linear mapping, whereas M × N is preferred when it is the source of a bilinear
mapping or the target of a mapping of any kind.
Obviously the categories C and C have been derived from the category C =
Mod(K) and the family (Mj ) in such a way that no special properties of Mod(K)
have been needed; consequently it is possible to repeat the same construction of
C and C by starting from any category C and any family (Mj ) of objects of C.
If C contains a final universal object, its source is called the direct product of the
family (Mj ) in C; and if C contains an initial universal object, its target is called
its direct sum in C.
When the objects of C are sets provided with some common structure, and
when its morphisms are the mappings respecting this structure, it often happens
that every cartesian product of objects is still an object, that the canonical pro-
jections from this cartesian product onto its components are morphisms in C, and
that finally this cartesian product is the direct product in C. This is true for the
category of all sets, the category Alg(K) and its subcategory Com(K).
The concepts of direct product and direct sum may be understood as dual
to each other if we remember the dual category C ∗ defined in 1.2. Indeed if we
replace C with C ∗ in the construction of C (without changing the family (Mj )),
we get the dual category of C ; and conversely. Therefore a direct sum in C is a
direct product in C ∗ , and conversely.
Despite this duality, the existence of direct sums is often a more difficult prob-
lem than the existence of direct products when the objects of C are sets provided
with some structure as above. Besides the previous example with C = Mod(K),
we will also look for direct sums of two objects in two other categories. First, in the
category of all sets, the direct sum of two sets M and N is the so-called disjoint
union. When M and N are actually disjoint, it is merely M ∪ N ; if not, it may be
the union of two disjoints sets M and N respectively isomorphic to M and N .
Secondly the direct sum of two commutative algebras A and B in Com(K)
is their tensor product A ⊗ B ; indeed (1.3.3) implies that there is a canonical
bijection
HomCom(K) (A ⊗ B, P ) −→ HomCom(K) (A, P ) × HomCom(K) (B, P ).
1.4 Tensor algebras and symmetric algebras

Let M be a K-module. We consider the category of all linear mappings from M
into a K-algebra P ; a morphism from f : M → P to g : M → Q is an algebra
morphism u : P → Q such that g = uf . If there is an initial universal object
ϕ : M → U , its target U (which is thus defined up to isomorphy because of
(1.2.1)) should be called the algebra freely generated by M . Two properties are
understood in this name. First U coincides with the subalgebra U generated by
ϕ(M ) in U ; indeed there must be an algebra morphism u : U → U such that
ϕ = uϕ, and then U → U → U must be equal to idU , whence U = U (since
U → U is the natural injection). Secondly, if the algebra P is generated by the
image of f : M → P , the unique algebra morphism f : U → P satisfying f = f ϕ
is surjective and makes P become a quotient of U ; in other words, any algebra
P generated by a linear image f (M ) of M is determined (up to isomorphy) by
the knowledge of Ker(f ) (or that of any subset generating it as an ideal). This
is what you must understand when one says that P can be deduced from U by
adding more relations between the generators; and these relations correspond to
the elements of Ker(f ). This point of view leads to an effective construction of
U , starting with a K-algebra V of noncommutative polynomials in noncommuting
indeterminates ex (indexed by x ∈ M ), and then taking the quotient of V by the
ideal W generated by all ex + ey − ex+y and λex − eλx with x, y ∈ M and λ ∈ K,
so that a linear mapping ϕ : M → V /W can be defined.
Nonetheless the above name of U is very seldom used, because there is an-
other construction of U that has become more popular, and that has led to it
being called the tensor algebra of M . Let us set
T0 (M ) = K , T1 (M ) = M , T2 (M ) = M ⊗M , T3 (M ) = M ⊗M ⊗M, . . . ;
because of the so-called associativity of tensor products, for all (i, j) ∈ N2 there
is a canonical isomorphism from Ti (M ) ⊗ Tj (M ) onto Ti+j (M ), whence a mul-
tiplication mapping Ti (M ) × Tj (M ) −→ Ti+j (M ); this multiplication can be
extended by linearity to the direct sum T(M ) of all the tensor powers Tk (M ), and
thus T(M ) is an associative algebra (even a graded algebra). The next proposition
means that the natural injection ϕ : M → T(M ) is the desired universal object.
(1.4.1) Proposition. Any linear mapping f from M into an algebra P extends
in a unique way to an algebra morphism f : T(M ) → P . For every sequence
(x1 , x2 , . . . , xk ) of elements of M (of any length k ≥ 1),
f (x1 ⊗ x2 ⊗ · · · ⊗ xk ) = f (x1 )f (x2 ) · · · f (xk ).
Since there is an algebra freely generated by the module M in the category

Alg(K), it is sensible to look for a commutative algebra freely generated by M
in the subcategory Com(K). Since the former can be constructed as a quotient of
an algebra of noncommutative polynomials, the latter can be constructed in the
1.4. Tensor algebras and symmetric algebras 9
same way by means of commutative polynomials. As explained above, we can also

obtain the latter by imposing more relations (relations of commutation) between
the generators of the former; in other words, it suffices to take the quotient of the
tensor algebra T(M ) by the ideal R generated by all x ⊗ y − y ⊗ x with x, y ∈ M .
The resulting algebra is called the symmetric algebra of M , and denoted by S(M );
the symbol ∨ is often used to mean the product of two elements of S(M ).
This algebra S(M ) inherits the grading of T(M ), because the ideal R is the
direct sum of the intersections Rk = R ∩ Tk (M ); thus S(M ) can be identified with
the direct sum of the quotients Sk (M ) = Tk (M )/Rk . Obviously R0 = R1 = 0,
whence S0 (M ) = K and S1 (M ) = M . The natural injection M → S(M ) is an
initial universal object in the category of all linear mappings M → P with target
a commutative algebra, as stated in the next proposition.
(1.4.2) Proposition. Any linear mapping from M into a commutative algebra P
extends in a unique way to an algebra morphism S(M ) → P .
In the algebra T(M ) each component Tk (M ) has a universal property, be-
cause it is a tensor power of M : for every k-linear mapping f from M k = M ×
M × · · ·× M into any module P , there is a unique linear mapping f : Tk (M ) → P
such that
f (x1 ⊗ x2 ⊗ · · · ⊗ xk ) = f (x1 , x2 , . . . , xk )
for all x1 , x2 , . . . , xk ∈ M . The components of S(M ) also have their particular
universal property.
(1.4.3) Proposition. For every symmetric k-linear mapping f : M k → P there is
a unique linear mapping f : Sk (M ) → P such that
f (x1 ∨ x2 ∨ · · · ∨ xk ) = f (x1 , x2 , . . . , xk )
for all x1 , x2 , . . . , xk ∈ M .
Proof. We suppose k ≥ 2, otherwise the symmetry hypothesis is empty. First f
induces a linear mapping f : Tk (M ) → P . As above, we treat Sk (M ) as the
quotient Tk (M )/Rk , where Rk is spanned by the products
x1 ⊗ x2 ⊗ · · · ⊗ xj−1 ⊗ (xj ⊗ xj+1 − xj+1 ⊗ xj ) ⊗ xj+2 ⊗ · · · ⊗ xk
with j ∈ {1, 2, . . . , k − 1} ; if the k-linear mapping f is symmetric, the mapping

f vanishes on all these products, whence a linear mapping f from Tk (M )/Rk =
Sk (M ) into P with the desired property. Since Sk (M ) is spanned by the products
x1 ∨ x2 ∨ · · · ∨ xk , the unicity of f is obvious.
1.5 Functors
Let C and D be two categories. A covariant functor F from C to D associates
with each object M of C an object F (M ) of D, and with each morphism f of C a
morphism F (f ) of D in such a way that the following three conditions are always
fulfilled:
if f ∈ HomC (M, N ) , then F (f ) ∈ HomD (F (M ), F (N )) ;
F (gf ) = F (g)F (f ) whenever gf exists;
if f is an identity morphism, then so is F (f ).
A contravariant functor F from C to D is a covariant functor from the dual category
of C (see 1.2) to the category D; this means that:
if f ∈ HomC (N, M ) , then F (f ) ∈ HomD (F (M ), F (N )) ;
F (f g) = F (g)F (f ) whenever f g exists;
if f is an identity morphism, then so is F (f ).
Many universal properties give rise to a functor. For instance there is a func-
tor T associated with tensor algebras; it is a covariant functor from Mod(K) to
Alg(K); the tensor algebra T(M ) of a module M has been defined in 1.4; if f
is a linear mapping from M into N , then T(f ) is the unique algebra morphism
T(M ) → T(N ) extending the linear mapping M → T(N ) defined by x −→ f (x).
It is easy to verify that a functor has been defined in this way. In the same way we
can define a functor S from Mod(K) to Com(K) by means of symmetric algebras.
The covariant functors between two categories are themselves the objects of
a category. When F and G are covariant functors from C to D, a morphism Φ from
F to G associates with each object M of C a morphism Φ(M ) from F (M ) to G(M )
in D, in such a way that G(f ) ◦ Φ(M ) = Φ(N ) ◦ F (f ) for every f ∈ HomC (M, N ).
This morphism Φ : F −→ G is an isomorphism if and only if all morphisms Φ(M )
are isomorphisms.
Let F be a covariant functor from C to D, and M and N two objects of C;
we assume that C contains morphisms ϕ1 : M → P and ϕ2 : N → P for which P
is the direct sum of M and N in this category, and that D contains morphisms
ψ1 : F (M ) → Q and ψ2 : F (N ) → Q for which Q is the direct sum of F (M ) and
F (N ) in this category. Now F (ϕ1 ) and F (ϕ2 ) have the same sources as ψ1 and
ψ2 , but their common target is F (P ); since (ψ1 , ψ2 ) is universal, all this results in
a morphism Q → F (P ). In other words, the existence of direct sums in C and in
D implies the existence of canonical morphisms
F (M ) ⊕D F (N ) −→ F (M ⊕C N ).
Some functors F have the nice property that these canonical morphisms are iso-
morphisms. For instance the functor S defined by means of symmetric algebras
has this property, as stated in the next theorem; remember that the direct sum of
two objects in Com(K) is their tensor product (see 1.3).
1.5. Functors 11
(1.5.1) Theorem. The symmetric algebra of the module M ⊕ N is canonically

isomorphic to the tensor product S(M ) ⊗ S(N ).
Proof. For every x ∈ M and y ∈ N the canonical algebra morphism S(M ) ⊗
S(N ) → S(M ⊕ N ) resulting from the previous argument maps x ⊗ 1 and 1 ⊗ y
respectively to (x, 0) and (0, y) in M ⊕ N . Conversely, because of the universal
property of S(M ⊕ N ), the linear mapping (x, y) −→ x ⊗ 1 + 1 ⊗ y extends to an
algebra morphism S(M ⊕ N ) → S(M ) ⊗ S(N ). Then it is easy to prove that in
this way two reciprocal isomorphisms have been constructed.
The definition of a functor of several variables can be easily guessed, and we
will present at once the evident example of the functor HomC from C × C to the
category of all sets, which is contravariant in the first variable and covariant in the
second variable. The set HomC (M, N ) has been defined in 1.2; now let us consider
two morphisms in C: f1 : M → M and f2 : N → N ; from them we derive a
mapping denoted by HomC (f1 , f2 ) :
HomC (M, N ) −→ HomC (M , N ), u −→ f2 uf1 .
Thus we have got a functor because

HomC (f1 g1 , g2 f2 ) = HomC (g1 , g2 ) ◦ HomC (f1 , f2 ) whenever f1 g1 and g2 f2
exist,
HomC (idM , idN ) is the identity mapping of HomC (M, N ).
By fixing the object N , we get a contravariant functor HomC (. . . , N ) of the first
variable (whence the notation HomC (f1 , N ) meaning HomC (f1 , idN )), and by fixing
M we get a covariant functor HomC (M, . . . ) of the second variable (whence the
notation HomC (M, f2 ) meaning HomC (idM , f2 )). When C is the category Mod(K),
the functor HomK takes its values in the category Mod(K).
Here we are also interested in the functor ⊗ which is a twice covariant functor
from Mod(K) × Mod(K) to Mod(K). The notations M ⊗ N and v ⊗ w have
been explained in 1.3, and it is easy to verify that they correspond to a functor;
the notations v ⊗ N and v ⊗ idN are synonymous. This functor has also a nice
behaviour relatively to direct sums, even infinite direct sums; as explained above,
the universal property of direct sums implies the existence of morphisms like this
one:
(M ⊗ Nj ) −→ M ⊗ Nj ;
j∈J
j∈J
to get a reciprocal morphism, it suffices to consider the evident bilinear mapping

M× Nj −→ (M ⊗ Nj );
j∈J
j∈J
because of the universal property of tensor products, we derive from it the recip-
rocal isomorphism.
If N is a free module, it is a direct sum of submodules all isomorphic to K,

and since M ⊗ K is canonically isomorphic to M , the tensor product M ⊗ N is
isomorphic to a direct sum of submodules all isomorphic to M . If M and N are
free modules, M ⊗ N is also a free module, and when x runs through a basis of M
and y through a basis of N , the tensor products x ⊗ y constitute a basis of M ⊗ N .
1.6 Exact sequences

Let us consider a sequence of two linear mappings u and v:
M −→ M −→ M ;
u v
it is called an exact sequence if Im(u) = Ker(v). The inclusion Im(u) ⊂ Ker(v) is

equivalent to the equality vu = 0; when we must prove the exactness of a sequence,
in most cases the equality vu = 0 will be obvious and the converse inclusion
Im(u) ⊃ Ker(v) will be the core of the proof. A sequence of several mappings, or
an infinite sequence of mappings, is said to be exact if all the subsequences of two
consecutive mappings are exact; for instance the sequence
0 −→ M −→M −→M −→ 0
u v
(1.6.1)
is exact if u is injective, v surjective, and Im(u) = Ker(v) .

An exact sequence like (1.6.1) is called a splitting exact sequence if the median
module M is the direct sum of M1 = Im(u) = Ker(v) and some submodule
M2 ; since the injection u determines an isomorphism M → M1 , and since the
surjection v then determines an isomorphism M2 → M , the splitting of the
exact sequence (1.6.1) makes M become isomorphic to M ⊕ M . Still under the
assumption that the sequence (1.6.1) is exact and splits, there exist two mappings
u : M → M and v : M → M such that
M2 = Ker(u ) = Im(v ), u u = idM and vv = idM ;
all this implies idM = uu + v v and u v = 0.

Conversely if the sequence (1.6.1) is exact and if there exists u : M → M
such that u u = idM , it is easy to prove that M is the direct sum of Im(u) and
Ker(u ), and this means that this exact sequence splits. In an analogous way the
exact sequence (1.6.1) is also splitting if there exists v : M → M such that
vv = idM , because this equality implies M = Im(v ) ⊕ Ker(v).
Besides, when the sequence (1.6.1) is no longer assumed to be exact, and
when the only hypothesis is the existence of four mappings u, v, u , v satisfying
the five equalities
vu = 0, u v = 0, u u = idM , vv = idM and uu + v v = idM , (1.6.2)

1.6. Exact sequences 13
then we get two splitting exact sequences:
0 −→ M −→M −→M −→ 0,
u v
u v
0 ←− M ←−M ←−M ←− 0.
Let F be an additive functor from the category Mod(K) into itself; the
additiveness of F means that for every couple of modules (M, N ) the mapping
u −→ F (u) is a group morphism from Hom(M, N ) into Hom(F (M ), F (N )) if F
is covariant, into Hom(F (N ), F (M )) if F is contravariant. With this hypothesis,
any equality vu = 0 implies F (v)F (u) = 0 or F (u)F (v) = 0 ; but this is not
sufficient to conclude that every exact sequence is transformed by F into an exact
sequence; F is called an exact functor if it transforms every exact sequence into
an exact sequence. It is not difficult to prove that F is exact if and only if it
transforms every exact sequence like (1.6.1) into an exact sequence.
Every splitting exact sequence is transformed into a splitting exact sequence,
because four mappings u, v, u , v satisfying the five equalities (1.6.2) are trans-
formed by F into four mappings satisfying the analogous five equalities that prove
the exactness and the splitting of the transformed sequence. Therefore if a functor
is not exact, its lack of exactness can be observed only on exact sequences that do
not split.
Unfortunately the most usual additive functors Hom and ⊗ are not exact for
all rings K; the former is only left exact (for both variables), and the latter is right
exact. This means that for all modules P and all exact sequences
0 −→ M −→ M −→ M ,
N −→ N −→ N −→ 0 ,
we get these exact sequences:
0 −→ Hom(N , P ) −→ Hom(N, P ) −→ Hom(N , P ) ,

0 −→ Hom(P, M ) −→ Hom(P, M ) −→ Hom(P, M ) ,
P ⊗ N −→ P ⊗N −→ P ⊗ N −→ 0.
From the right exactness of the functor ⊗ the following statement can be
immediately derived: the exactness of the two sequences
M −→ M −→ M −→ 0 and N −→ N −→ N −→ 0
implies the exactness of the sequence
(M ⊗ N ) ⊕ (M ⊗ N ) −→ M ⊗ N −→ M ⊗ N −→ 0; (1.6.3)
this is proved in Exercise (1.ex.10).

1.7 Projective modules and flat modules

A K-module P is called injective if the functor Hom(. . . , P ) is exact; it is called
projective if the functor Hom(P, . . . ) is exact; and it is called flat if the functor
P ⊗ · · · is exact. Because of the left exactness of the functor Hom and the right
exactness of the functor ⊗ (see 1.6), we get at once the following statements:
P is injective if and only if the mapping Hom(N, P ) → Hom(N , P ) is surjec-
tive whenever N → N is injective; P is projective if and only if the mapping
Hom(P, M ) → Hom(P, M ) is surjective whenever M → M is surjective; and P
is flat if and only if the mapping P ⊗ N → P ⊗ N is injective whenever N → N
is injective.
Here we never need injective modules; therefore only classical properties of
projective or flat modules are recalled in this section.
A (finite or infinite) direct sum of modules is projective if and only if all the
components are projective. Since K itself is projective, this proves that every free
module is projective.
The following four statements are equivalent:
(a) P is projective;
(b) when the morphism M → M is surjective, every morphism P → M can
be factorized through M by means of some morphism P → M ;
(c) every exact sequence 0 → M → M → P → 0 is splitting if it contains P at
the fourth place;
(d) there exists a module P such that P ⊕ P is a free module.
Moreover if P is a finitely generated projective module, there exists a module P
such that P ⊕ P is a free module with finite bases.
A (finite or infinite) direct sum of modules is flat if and only if all components
are flat. Consequently, since K is flat, every projective module is flat.
Every tensor product of projective modules (resp. flat modules) is projective
(resp. flat). When M is projective, Hom(M, N ) is projective (resp. flat) whenever
N is projective (resp. flat).
A module P is called faithfully flat if it is flat and if every equality P ⊗ M = 0
implies M = 0.
When P is merely flat, every linear mapping f : M → N gives an exact
sequence
0 −→ P ⊗ Ker(f ) −→ P ⊗ M −→ P ⊗ N −→ P ⊗ Coker(f ) −→ 0
which proves that the kernel and cokernel of P ⊗f can be identified with P ⊗Ker(f )
and P ⊗ Coker(f ). When P is faithfully flat, then f is injective (resp. surjective)
if and only if P ⊗ f is injective (resp. surjective); indeed the vanishing of Ker(f )
is equivalent to the vanishing of P ⊗ Ker(f ), and the same for Coker(f ).
More generally, when P is faithfully flat, a sequence M → M → M is exact
if and only if the sequence P ⊗ M −→ P ⊗ M −→ P ⊗ M is exact. Indeed the
1.8. Finitely presented modules 15
mapping M −→ Ker(M → M ) is surjective if and only if its gives a surjective

mapping by the functor P ⊗ · · · .
Let N and N be submodules of M . When P is merely flat, P ⊗N and P ⊗N
can be identified with submodules of P ⊗ M , and
(P ⊗ N ) + (P ⊗ N ) = P ⊗ (N + N ),
(P ⊗ N ) ∩ (P ⊗ N ) = P ⊗ (N ∩ N );
indeed the former equality is obvious, and the latter one follows from it because
of the exact sequence
0 −→ N ∩ N −→ N ⊕ N −→ N + N −→ 0
in which the second arrow is c −→ (c, −c) and the third one is (a, b) −→ a + b.
Obviously every inclusion N ⊂ N in M gives a similar inclusion in P ⊗ M . When
P is faithfully flat, the inclusion N ⊂ N is equivalent to P ⊗ N ⊂ P ⊗ N . Indeed
these inclusions are respectively equivalent to the surjectiveness of the mappings
N → N + N and P ⊗ N → P ⊗ (N + N ).
The finitely generated projective modules afford the most convenient frame to
generalize the classical properties of vector spaces of finite dimension. For instance
we may associate with each module M the dual module M ∗ = Hom(M, K), and
there is a canonical mapping from M into the bidual module (M ∗ )∗ ; if P is finitely
generated and projective, the mapping P → (P ∗ )∗ is an isomorphism; indeed this
is obviously true when P is a free module with a finite basis; when P is not free,
there exists a module P such that P ⊕ P is free with a finite basis; whence an
isomorphism P ⊕ P → ((P ⊕ P )∗ )∗ ; but (P ⊕ P )∗ is canonically isomorphic to
P ∗ ⊕ P ∗ , and finally we get an isomorphism
P ⊕ P −→ (P ∗ )∗ ⊕ (P ∗ )∗ ;
since this isomorphism maps each of the two components on the left side into
the corresponding component on the right side, it gives two isomorphisms, among
which is the announced isomorphism.
1.8 Finitely presented modules

When M is a K-module, every subset S that spans M , provides a surjective mor-
phism N → M defined on the module N = K (S) freely generated by S (see 1.3); if
M is finitely generated, we can require S to be a finite subset, and N to have finite
bases. Unfortunately in many cases it is not sufficient that N contains finite bases;
we must also consider the kernel of the morphism N → M , which is called the
module of relations between the generators. A finite presentation of M is a finite
subset of generators that gives a finitely generated module of relations between
these generators. Nonetheless the existence of finite presentations is ensured by a

much weaker definition.
(1.8.1) Definition. A module M is called finitely presented if it is finitely generated
and if there exists a surjective morphism f : P → M such that P is projective
and Ker(f ) finitely generated.
According to this definition every finitely generated projective module is
finitely presented. Now the next theorem implies that in a finitely presented mod-
ule every finite subset of generators actually gives a finitely generated module of
relations.
(1.8.2) Theorem. Let M be a finitely presented module, and f : P → M a surjective
morphism from a projective module P onto M ; the kernel of f is finitely generated
if and only if P is finitely generated.
This theorem is a consequence of the following lemma.
(1.8.3) Schanuel’s lemma. Let f : P → M and f : P → M be surjective mor-
phisms from two projective modules P and P onto the same module M ; the mod-
ules P ⊕ Ker(f ) and P ⊕ Ker(f ) are isomorphic. Consequently if P and Ker(f )
are finitely generated, so are P and Ker(f ) (and conversely).
Proof of (1.8.3). Since P is projective and f surjective, there exists u : P →
P such that f = f u. Let Q be the submodule of P ⊕ P containing all pairs
(x, x ) such that f (x) = f (x ). We get reciprocal isomorphisms between Q and
P ⊕ Ker(f ) if we map every (x, x ) ∈ Q to (x, x − u(x)), and conversely every
(x, y ) ∈ P ⊕ Ker(f ) to (x, y + u(x)). In an analogous way we get a pair of
reciprocal isomorphisms between Q and P ⊕ Ker(f ).
Proof of (1.8.2). Since M is finitely generated, there exists a surjective mapping
f : P → M defined on a finitely generated projective module P ; and since M is
finitely presented, there exists a surjective mapping f : P → M such that P
is projective and Ker(f ) finitely generated. Now if P is finitely generated, since
Ker(f ) is also finitely generated, P and Ker(f ) are finitely generated because
of (1.8.3). Conversely if Ker(f ) is finitely generated, since P is finitely generated,
P and Ker(f ) are also finitely generated.
When the basic ring K is noetherian, every submodule of a finitely generated
module is also finitely generated; therefore every finitely generated module is also
finitely presented.
Suppose that M is finitely presented; every finite subset of generators gives a
surjective morphism f : M0 → M defined on a free module M0 with a finite basis;
every finite subset of generators of Ker(f ) gives a surjective morphism M1 →
Ker(f ) defined on a free module M1 with a finite basis; we can link together the
exact sequences
M1 −→ Ker(f ) −→ 0 and 0 −→ Ker(f ) −→ M0 −→ M −→ 0,
1.9. Changes of basic rings 17
in order to get the exact sequence that symbolizes the finite presentation of M :
M1 −→ M0 −→ M −→ 0.
A tensor product of finitely presented modules is a finitely presented module,

because from the two exact sequences
M1 −→ M0 −→ M −→ 0 and N1 −→ N0 −→ N −→ 0,
by means of (1.6.3) we can derive the exact sequence
(M1 ⊗ N0 ) ⊕ (M0 ⊗ N1 ) −→ M0 ⊗ N0 −→ M ⊗ N −→ 0.
Moreover Hom(P, M ) is a finitely presented module whenever P is a finitely

generated projective module, and M a finitely presented module; indeed from the
above exact sequence M1 → M0 → M → 0 we can derive the exact sequence
Hom(P, M1 ) −→ Hom(P, M0 ) −→ Hom(P, M ) −→ 0;
there exists P such that P ⊕ P is free with finite bases; thus Hom(P ⊕ P , M0 ) is
also free with finite bases, and consequently Hom(P, M0 ) is finitely generated and
projective; and the same for Hom(P, M1 ).
1.9 Changes of basic rings

When we meet additive groups that are modules over two commutative rings K
and L, it is necessary to use precise notations like HomL (M, N ), M ⊗L N , TL (M ),
SL (M ),. . . indicating which basic ring is referred to. Here we study what happens
when there is a ring morphism f : K → L; in this case every L-module M is also
a K-module: for all κ ∈ K and all x ∈ M the product κx is by definition f (κ)x.
Such a ring morphism f : K → L is called an extension of the ring K, even when
it is not injective. It may even occur that f is surjective; for instance if M is not
a faithful K-module, the ideal containing all κ ∈ K such that κM = 0 is not
reduced to 0, and M becomes a faithful module over the quotient of K by this
ideal.
First let us suppose that M and N are L-modules, and therefore also K-
modules; there is obviously a canonical injection HomL (M, N ) → HomK (M, N )
and a canonical surjection M ⊗K N → M ⊗L N ; the kernel of the latter is spanned
by the elements λx ⊗ y − x ⊗ λy with λ running through L and x and y through
M and N ; both mappings are isomorphisms when f is surjective. If we consider
tensor and symmetric algebras, we find canonical morphisms of graded algebras
TK (M ) −→ TL (M ) and SK (M ) −→ SL (M );
in degree 0 we find merely the ring morphism f : K → L ; in degree 1 we find

the identity mapping of M = T1K (M ) = T1L (M ) = S1K (M ) = S1L (M ), and for
each degree k ≥ 2 the mappings TkK (M ) → TkL (M ) and SkK (M ) → SkL (M ) are
surjective, and even bijective when f is surjective.
Let us now suppose that M is merely a K-module; we may derive from it
two L-modules, called the extensions of M , which are L ⊗K M and HomK (L, M );
an element µ of L multiplies an element λ ⊗ x of L ⊗K M or an element ξ of
HomK (L, M ) in this way:
µ(λ ⊗ x) = (µλ) ⊗ x and (µξ)(λ) = ξ(µλ).
There are canonical K-linear mappings
M −→ L ⊗K M and HomK (L, M ) −→ M
defined by x −→ 1L ⊗ x and ξ −→ ξ(1L ) ; the former is not always injective,
and the latter is not always surjective. When L = K, both are bijective and
usually K ⊗K M and HomK (K, M ) are identified with M . But in general these
two extensions are different, and isomorphisms can be found between them only
under restrictive hypotheses. The next lemma gives more details in a particular
case.
(1.9.1) Lemma. Let us suppose that f : K → L is surjective; let a be its kernel.
Then L ⊗K M is canonically isomorphic to the quotient M/aM , and HomK (L, M )
to the submodule of all elements x ∈ M such that ax = 0.
Proof. The right exactness of the functor ⊗ gives the exact sequence
a ⊗K M −→ K ⊗K M −→ L ⊗K M −→ 0;
when K ⊗K M is identified with M , the image of a ⊗K M in M is aM , and thus

L ⊗K M = M/aM . The left exactness of the functor Hom allows us to prove the
statement involving HomK (L, M ).
Here we shall only use extensions like L ⊗K M , but the other extension
HomK (L, M ) may appear in other contexts. Let us begin with these isomorphisms:
TL (L ⊗K M ) ∼
= L ⊗K TK (M ), (1.9.2)
∼ L ⊗K SK (M ).
SL (L ⊗K M ) = (1.9.3)
Most of the here mentioned isomorphisms are easy consequences of the definitions
and universal properties of the objects under consideration, and the following
explanations about (1.9.2) should be a sufficient model for all the others. Observe
that the L-algebras TL (L ⊗K M ) and L ⊗K TK (M ) are generated by the elements
which in each algebra are written 1L ⊗ x (with x ∈ M ). The mapping which
maps every λ ⊗ x in L ⊗K M to the corresponding λ ⊗ x in L ⊗K TK (M ), is
L-linear, and therefore extends to a morphism of L-algebras TL (L ⊗K M ) →
L ⊗K TK (M ). Conversely the mapping x −→ 1L ⊗ x extends to a morphism
1.9. Changes of basic rings 19
of K-algebras TK (M ) → TL (L ⊗K M ), and by combining it with the canonical

morphism L → TL (L⊗K M ), the image of which lies in the center of TL (L⊗K M ),
we get a morphism of K-algebras L ⊗K TK (M ) → TL (L ⊗K M ). This converse
morphism too is L-linear, and the behaviour of both morphisms on the elements
1L ⊗ x shows that they are reciprocal isomorphisms of L-algebras.
When M is a K-module as previously, and N an L-module, there are canon-
ical isomorphisms
HomK (M, N ) ∼
= HomL (L ⊗K M, N ), (1.9.4)
M ⊗K N ∼ = (L ⊗K M ) ⊗L N. (1.9.5)
To be complete, let us also mention the isomorphisms
HomK (N, M ) ∼
= HomL (N, HomK (L, M )).
The isomorphisms (1.9.4) and (1.9.5) lead to an easy proof of this statement: the
extension L ⊗K M is L-projective (resp. L-flat) whenever M is K-projective (resp.
K-flat). When M is injective, its injectiveness is inherited by the other extension
HomK (L, M ).
When both M and N are K-modules, we find the canonical isomorphisms
L ⊗K (M ⊗K N ) ∼
= (L ⊗K M ) ⊗L (L ⊗K N ). (1.9.6)
To be complete, let us also mention the isomorphisms
HomK (L, HomK (M, N )) ∼
= HomL (L ⊗K M, HomK (L, N )).
Sometimes we also need the following morphism which is not always bijective:
L ⊗K HomK (M, N ) −→ HomL (L ⊗K M, L ⊗K N ),
λ ⊗ g −→ (µ ⊗ x −→ λµ ⊗ g(x)).
(1.9.7) Proposition. When M is projective, the above canonical morphism from
L ⊗K HomK (M, N ) into HomL (L ⊗K M, L ⊗K N ) is an isomorphism.
Proof. When M is a direct sum, this morphism is bijective if and only if it is
bijective when M is successively replaced with each direct summand; consequently
it suffices to prove (1.9.7) when M = K; in this case it is obviously bijective.
The extension K → L is called flat (resp. faithfully flat) if L is a flat (resp.
faithfully flat) K-module. When the extensions K → L and L → L are flat (resp.
faithfully flat), it is clear that K → L is flat (resp. faithfully flat). Let K → K
and K → L be two extensions of K, and L = K ⊗K L; when the extension
K → L is flat (resp. faithfully flat), then K → L is also flat (resp. faithfully flat);
indeed, for every K -module M ,
L ⊗K M ∼
= (K ⊗K L) ⊗K M ∼
= L ⊗K K ⊗K M ∼
= L ⊗K M .
Other properties of flatness or faithful flatness are stated in 1.7.
(1.9.8) Proposition. When K → L is a faithfully flat extension and M a K-module,

M is finitely generated (resp. finitely presented) if and only if the L-module L⊗K M
is finitely generated (resp. finitely presented).
Proof. Without any hypothesis on K → L, every exact sequence M1 → M0 →
M → 0 is transformed by the functor L ⊗ · · · into an exact sequence, and this
proves that the mentioned properties of M are inherited by L ⊗K M . Conversely
if L ⊗K M is finitely generated, there exist x1 , x2 , . . . , xn ∈ M such that L ⊗ M
is generated by all 1L ⊗ xi with i = 1, 2, . . . , n; let M0 be a free K-module with
basis (e1 , e2 , . . . , en ), and f the K-linear mapping M0 → M such that f (ei ) = xi
for i = 1, 2, . . . , n; this mapping f is surjective because L ⊗ f is surjective and L
faithfully flat; therefore M is finitely generated. The kernel of L ⊗ f is canonically
isomorphic to L⊗Ker(f ) because L is flat; if L⊗M is finitely presented, this kernel
is finitely generated; consequently L ⊗ Ker(f ) is finitely generated; by a similar
argument Ker(f ) is also finitely generated; thus M is finitely presented.
(1.9.9) Proposition. When K → L is a flat extension, and M a finitely presented
K-module, then for each K-module N the canonical morphism
L ⊗K HomK (M, N ) −→ HomL (L ⊗K M, L ⊗K N )
is an isomorphism.
Proof. It is obviously an isomorphism when M is a free module with a finite basis.
When M is merely finitely presented, there is an exact sequence M1 → M0 →
M → 0 in which M1 and M0 are free modules with finite bases; it leads to the
following diagram:
0→ L ⊗ Hom(M, N ) → L ⊗ Hom(M0 , N ) → L ⊗ Hom(M1 , N )
↓ ↓ ↓
0 → HomL (L ⊗ M, L ⊗ N ) → HomL (L ⊗ M0 , L ⊗ N ) → HomL (L ⊗ M1 , L ⊗ N )
The two lines are exact because of the flatness of L and the left exactness of the
functors HomK and HomL ; the second and third vertical arrows are isomorphisms;
now it is easy to prove that the first vertical arrow too is an isomorphism.
(1.9.10) Proposition. When K → L is a faithfully flat extension and P a K-
module, P is a finitely generated projective K-module if and only if L ⊗K P is a
finitely generated projective L-module.
Proof. When P is finitely generated and projective over K, so is L ⊗K P over L
because of (1.9.4). The hypothesis about K → L is only needed when we conversely
suppose that L⊗K P is a finitely generated projective L-module; then according to
(1.9.8), P is finitely presented; the projectiveness of P is equivalent to the exactness
of the functor HomK (P, . . . ), and since L is faithfully flat, it is equivalent to the
exactness of the functor L⊗K HomK (P, . . . ); because of (1.9.9) it is also equivalent
to the exactness of the functor HomL (L ⊗K P, L ⊗K · · · ); now the projectiveness
of L ⊗K P implies the projectiveness of P .
1.10. Rings and modules of fractions 21
1.10 Rings and modules of fractions

A multiplicative subset of K (also called a multiplicatively closed subset) is a subset
S that contains 1 and all products of two elements of S. We consider the category
C of all ring morphisms f : K → L such that f (s) is invertible in L for all s ∈ S;
a morphism from f to f : K → L is a ring morphism u : L → L such that
f = uf . The morphism K → 0 is a final universal object in C; it is its unique
object (up to isomorphy) when S contains 0, and it is preferable also to accept
this quite degenerate case. If C contains an initial universal object f : K → U , it
is unique up to isomorphy (see (1.2.1)), and U is called the ring of fractions of K
with denominator in S and denoted by S −1 K.
Let us prove the existence of S −1 K. Two elements (λ, s) and (λ , s ) of K × S
are said to be equivalent if there exists t ∈ S such that t(s λ − sλ ) = 0 ; it is easy
to prove that an equivalence relation has been defined in this way; let S −1 K be
the set of equivalence classes; the image of (λ, s) in S −1 K is written as a fraction
λ/s. This set is given a ring structure with the following operations:
λ µ tλ + sµ λµ λµ
+ = and = ;
s t st s t st
it is easy to prove that this addition and this multiplication are well defined on the
set of equivalence classes and satisfy the required properties for S −1 K to be a ring;
the zero and unit elements are respectively the fractions 0/1 and 1/1. Moreover
the mapping f : K −→ S −1 K which maps every λ to the fraction λ/1 is a ring
morphism, and it is easy to prove that it is an initial universal object in C, as
stated in the following theorem.
(1.10.1) Theorem. For every ring morphism f : K → L such that f (s) is invertible
for all s ∈ S, there exists a unique ring morphism f : S −1 K → L such that
f (λ) = f (λ/1) for all λ ∈ K.
It is clear that f (λ/s) = f (λ)f (s)−1 for all fractions λ/s. Moreover an element
λ ∈ K belongs to the kernel of the canonical morphism f : K → S −1 K if and only
if there exists t ∈ S such that tλ = 0.
If S is a multiplicative subset containing S, the universal property of S −1 K
implies that the canonical mapping K → S −1 K can be factorized through S −1 K,
and thus we get a ring morphism S −1 K → S −1 K.
When K is an integral domain (a ring without divisors of zero, and not
reduced to 0), the set of all nonzero elements of K is a multiplicative subset; the
corresponding ring of fractions is a field, which is called the field of fractions of
K; all rings of fractions of K and K itself can be identified with subrings of this
field.
Now let M be any K-module, and D the category in which the objects are
the K-linear mappings g : M → P from M into any (S −1 K)-module P ; a mor-
phism from g to g : M → P is a (S −1 K)-linear mapping u : P → P such that
g = ug. If D contains an initial universal object fM : M −→ V , it is unique up to

isomorphy, V is called the module of fractions of M with denominator in S and
denoted by S −1 M .
Let us prove the existence of S −1 M . Two elements (x, s) and (x , s ) of M ×S
are said to be equivalent if there exists t ∈ S such that t(s x−sx ) = 0 ; it is easy to
prove that an equivalence relation has been defined; the set of equivalence classes
is the wanted module S −1 M and the image of (x, s) in S −1 M is by definition the
fraction x/s. It is easy to prove that the following operations are well defined and
make S −1 M become an (S −1 K)-module:
x y tx + sy λ y λy
+ = and = .
s t st s t st
Moreover the mapping fM : M −→ S −1 M which maps each x to x/1, is a K-linear

mapping, and it is easy to prove that it is an initial universal object in D, in other
words: for every K-linear mapping g : M → P into an (S −1 K)-module P , there
exists a unique (S −1 K)-linear mapping g : S −1 M → P such that g(x) = g (x/1)
for all x ∈ M .
An element x ∈ M belongs to the kernel of the canonical morphism x −→ x/1
if and only if there exists t ∈ S such that tx = 0.
The canonical morphism x −→ x/1 is an isomorphism from M onto S −1 M
if and only if the endomorphism x −→ sx is bijective from M onto M for all
s ∈ S; this condition is necessary and sufficient for M to have a structure of
(S −1 K)-module compatible with its structure of K-module.
If S is a multiplicative subset of K containing S, the universal property of
S M gives a canonical (S −1 K)-linear mapping S −1 M → S −1 M.
−1
Two kinds of multiplicative subsets will be used later. First from any element
s ∈ K we can derive the multiplicative subset S = {1, s, s2 , s3 , . . . } of all powers
of s; then the ring S −1 K is usually denoted by Ks ; it is reduced to 0 if and only
if s is nilpotent. Similarly Ms is the module of fractions of M with denominator a
power of s.
An ideal p of K is called a prime ideal when the following four equivalent
statements are true:
(a) p = K, and whenever p contains a product xy of elements of K, it contains
x or y;
(b) p = K, and whenever p contains a product ab of ideals of K, it contains
a or b;
(c) the quotient K/p is an integral domain (without divisors of zero and not
reduced to 0);
(d) the complementary subset S = K \ p is a multiplicative subset.
The corresponding ring S −1 K and modules S −1 M are denoted by Kp and Mp ,
and are called the localizations of K and M at the prime ideal p.
1.10. Rings and modules of fractions 23
Every ring K (not reduced to 0) contains prime ideals; indeed Zorn’s Lemma
implies the existence of maximal ideals, and maximal ideals are prime, because
the following two assertions are equivalent:
(a) the ideal m is maximal (it is contained in no ideal other than m and K, and
m = K);
(b) the quotient K/m is a field.
Zorn’s Lemma even implies that every ideal other than K is contained in a maximal
ideal.
Let us also recall that for any ring K the following two statements are equiv-
alent:
(a) K contains exactly one maximal ideal;
(b) K is not reduced to 0, and a sum of noninvertible elements is never invertible.
When these statements are true, K is called a local ring. In a local ring the unique
maximal ideal m is the subset of all noninvertible elements. The quotient K/m is
called the residue field of the local ring K.
When p is a prime ideal of K, the elements of Kp which are not invertible,
are the elements of pKp (this notation is meaningful since Kp is a K-module); this
proves that the localized ring Kp is a local ring.
Here is a first application of these notions.
(1.10.2) Theorem. The radical of K, which is the subset Rad(K) of all its nilpotent
elements, is also the intersection of all its prime ideals.
Proof. It is clear that every prime ideal contains all nilpotent elements. Conversely
we prove that when s is not nilpotent, there exists a prime ideal that does not
contain it. Indeed let us consider the ring Ks of fractions with denominator in
the set of powers of s, and the canonical morphism f : K → Ks . Since s is not
nilpotent, Ks is not reduced to 0. If m is a maximal ideal of Ks , f −1 (m) is a prime
ideal p of K because K/p is isomorphic to a subring of the field Ks /m. Since f (s)
is invertible and cannot belong to m, we are sure that s ∈ / p.
Every multiplicative subset S affords a functor from Mod(K) toward
Mod(S −1 K); with every K-module M we associate the (S −1 K)-module S −1 M ,
and with every K-linear mapping f : M −→ N we associate the (S −1 K)-linear
mapping S −1 f : S −1 M → S −1 N defined in this way: because of the universal
property of S −1 M , the mapping M −→ N → S −1 N can be factorized in a unique
way through S −1 M .
(1.10.3) Theorem. The functor M −→ S −1 M is exact. Moreover it is equivalent
to the functor M −→ S −1 K ⊗ M corresponding to the extension K → S −1 K of
the basic ring; in other words, for each K-module M there is a canonical (S −1 K)-
linear isomorphism S −1 M → S −1 K ⊗ M , and for each K-linear mapping f :
M −→ N the canonical isomorphisms S −1 M → S −1 K ⊗M and S −1 N → S −1 K ⊗
N intertwine S −1 f and S −1 K ⊗ f .
Proof. Let us consider an exact sequence M −→ M −→ M , and let us prove

u v
the exactness of S −1 M → S −1 M → S −1 M . It is clear that S −1 v ◦ S −1 u = 0;

therefore we must prove that every fraction x/s ∈ S −1 M that belongs to the
kernel of S −1 v, must belong to the image of S −1 u. Indeed there exists t ∈ S such
that tv(x) = 0, in other words, tx ∈ Ker(v); consequently there exists x ∈ M
such that u(x ) = tx, whence x/s = S −1 u(x /st) as desired.
The proof of the second part of (1.10.3) is still easier: there is a mapping
S −1 M → S −1 K ⊗ M resulting from the universal property of S −1 M , and there
is a converse mapping (λ/s) ⊗ x −→ (λx)/s resulting from the universal property
of the tensor product; obviously they are reciprocal isomorphisms. The statement
about S −1 f and S −1 K ⊗ f is also evident.
(1.10.4) Corollary. The ring extension K → S −1 K is flat.

Indeed the exactness of the functor M −→ S −1 K ⊗M is an immediate consequence
of the exactness of the functor M −→ S −1 M .
(1.10.5) Corollary. For all K-modules M and M there is a canonical isomorphism
S −1 (M ⊗K M ) ∼
= S −1 M ⊗S −1 K S −1 M .
Indeed, according to (1.9.6), there is a canonical isomorphism
S −1 K ⊗K (M ⊗K M ) ∼
= (S −1 K ⊗K M ) ⊗S −1 K (S −1 K ⊗K M ).
(1.10.6) Corollary. Let a be the ideal of K generated by the elements s1 , s2 , . . . , sn .

The direct product L of the rings Ksi (where i = 1, 2, . . . , n) is faithfully flat if
and only if a = K. When a = K, this ring L is called a Zariski extension of K.
Proof. Since L is a direct sum of flat modules, it is flat. Let us suppose that
L ⊗ M = 0 ; therefore for each x ∈ M there exists a positive integer k such
thatski x = 0 for i = 1, 2, . . . , n. If a = K, there exist λ1 , λ2 ,. . . ,λ
nn such that
n
1 = i=1 λi si ; let us set m = n(k − 1) + 1 ; from the equality 1 = n i=1 λ
( i si )
m
k
we can deduce the existence of µ1 , µ2 , . . . , µn such that 1 = i=1 µi si ; this
shows that the equalities ski x = 0 imply x = 0 ; therefore M = 0. Conversely if
a = K, the equality (K/a) ⊗ L = 0 (a consequence of (1.9.1)) shows that L is not
faithfully flat.
Among the consequences of Theorem (1.10.3) there is the fact that S −1 N

can be considered as a submodule of S −1 M whenever N is a submodule of M ;
indeed the sequence 0 −→ S −1 N −→ S −1 M is exact. The proof of the following
lemma is left to the reader.
(1.10.7) Lemma. When N and N are submodules of M , then
S −1 (N + N ) = S −1 N + S −1 N and S −1 (N ∩ N ) = S −1 N ∩ S −1 N .
1.11. Localization and globalization 25
Whereas the functor M −→ S −1 M has a nice behaviour relatively to tensor

products (see (1.10.5)), it requires more caution when the functor Hom is involved.
(1.10.8) Proposition. For all K-modules M and N there is a canonical morphism
S −1 HomK (M, N ) −→ HomS −1 K (S −1 M, S −1 N )
which is an isomorphism whenever M is a finitely presented module.
Each element g/s of S −1 Hom(M, N ) is mapped to x/t −→ g(x)/st. This proposi-
tion is an immediate corollary of (1.9.9) because the extension K → S −1 K is flat,
as stated in (1.10.4).
Let us come back to the ring extension f : K → S −1 K. Later it shall be nec-

essary to know the prime ideals of S −1 K. As explained above for the submodules
of any K-module M , to each ideal a of K corresponds an ideal S −1 a of S −1 K; it
is the ideal generated by f (a). Conversely with every ideal b of S −1 K we associate
the ideal f −1 (b) of K; it is clear that f −1 (b) ∩ S = ∅ whenever b = S −1 K. The
proof of the following lemma is left to the reader.
(1.10.9) Lemma. The mapping q −→ f −1 (q) is a bijection from the set of prime
ideals q of S −1 K onto the set of prime ideals p of K such that p ∩ S is empty; the
converse bijection is p −→ S −1 p.
More generally S −1 f −1 (b) = b for every ideal b of S −1 K, but if a is an ideal of

K, the ideal f −1 (S −1 a) may be larger than a, since it is the set of all λ ∈ K such
that sλ ∈ a for some s ∈ S.
1.11 Localization and globalization

The spectrum of the ring K, denoted by Spec(K), is the set of all its prime ideals.
With every ideal a of K we associate the subset V(a) of all p ∈ Spec(K) such that
p ⊃ a.
(1.11.1) Lemma. The mapping a −→ V(a) has the following properties:
(a) V(K) = ∅ and V(0) = Spec(K);
(b) an inclusion a ⊂ b implies V(a) ⊃ V(b);
(c) when (aj )j∈J is any family of ideals of K, then
V(aj ) = V( aj );
j∈J j∈J
(d) for all ideals a and b of K, we can write
V(a) ∪ V(b) = V(a ∩ b) = V(ab).

All these statements are evident except perhaps the last one, a consequence of the
inclusions
V(a) ∪ V(b) ⊂ V(a ∩ b) ⊂ V(ab) ⊂ V(a) ∪ V(b).
Lemma (1.11.1) proves that there is a topology on Spec(K) for which the closed
subsets are the subsets V(a); it is called the Zariski topology of Spec(K). For every
p ∈ Spec(A) the topological closure of {p} is V(p); thus the point {p} is closed if
and only if p is a maximal ideal; this topology is almost never a Hausdorff topology.
As explained in 1.10, with each p ∈ Spec(K) is associated a localized ring Kp
with maximal ideal pKp , and a residue field Kp /pKp. The kernel of the ring mor-
phism K → Kp → Kp /pKp is exactly p. Therefore there is an injective morphism
from the integral domain K/p into the residue field, which extends to a morphism
from the field of fractions of K/p into the residue field. This morphism is surjec-
tive, consequently bijective (since every field morphism is injective). Therefore the
residue field Kp /pKp can be identified with the field of fractions of K/p. At every
maximal ideal m the residue field is K/m.
Every module M gives a localized module Mp at the point p, and a vector
space Fp ⊗K M over the residue field Fp = Kp /pKp . The associativity of tensor
products allows us to write
Fp ⊗K M = Fp ⊗Kp Kp ⊗K M = Fp ⊗Kp Mp ;
and then (1.9.1) allows us to identify Fp ⊗ M with Mp /pMp . For a maximal ideal
m we can write Fm = K/m and Fm ⊗ M = M/mM .
Let f : K → L be any ring morphism. When q is a prime ideal of L, then
f −1 (q) is a prime ideal of K, because the quotient K/f −1 (q) is isomorphic to a
subring of L/q. This defines a mapping Spec(f ) from Spec(L) into Spec(K), and
it is easy to prove that it is continuous. Thus we have constructed a contravariant
functor from the category Com(Z) of all commutative rings to the category of
topological spaces.
(1.11.2) Example. Let F be an algebraically closed field, and K = F [x1 , x2 , . . . , xn ]
a ring of polynomials over F ; K can be identified with the ring of polynomial
functions on F n and its field of fractions F (x1 , x2 , . . . , xn ) with the field of rational
functions on F n ; every ring of fractions of K is a subring of this field. At every
point a = (a1 , a2 , . . . , an ) ∈ F n there is a valuation morphism K → F defined
by f −→ f (a); if ma is its kernel, the quotient K/ma is isomorphic to F and
consequently ma is a maximal ideal; a theorem of Hilbert (Nullstellensatz) proves
that the image of the mapping a −→ ma is the set of all maximal ideals. We
identify the point a ∈ F n with the point ma ∈ Spec(K). The localized ring Ka
at the maximal ideal ma is the subring of all rational functions g that are defined
at the point a, and the image of g in the residue field (which is isomorphic to
K/ma = F ) can be identified with the value g(a) of g at the point a. Now let p
be any prime ideal of K, and V (p) the subset of all a ∈ F n such that f (a) = 0
for all f ∈ p; a subset like V (p) is called an irreducible algebraic submanifold of
1.11. Localization and globalization 27
F n , and its points are the maximal ideals belonging to V(p). The localized ring
at p is the subring of all rational functions that are defined at least at one point
of V (p) (consequently at almost all points of V (p)). The ideal p is the kernel of
the ring morphism that maps every polynomial function f ∈ K to its restriction
to the subset V (p), and thus K/p is identified with a ring of functions on V (p),
the so-called regular functions. The residue field at p is the field of fractions of
K/p and its elements are called the rational functions on V (p). If g belongs to the
localized ring Kp , its image in the residue field is its restriction to V (p).
Let s be a nonzero element in the previous ring K = F [x1 , . . . , xn ]. It is (in an
essentially unique way) a product of prime elements (each one generates a prime
ideal in K); with each prime divisor of s is associated an irreducible algebraic
hypersurface in F n , and V (Ks) (which is the subset of all a ∈ F n such that
s(a) = 0) is the union of all these hypersurfaces. Let Us be the complementary
subset of V (Ks) in F n . The elements of Ks are the rational functions on F n
that are defined at all points of Us ; their restrictions to Us are called the regular
functions on Us . According to Lemma (1.10.9) the mapping Spec(Ks ) → Spec(K)
is a bijection from Spec(Ks ) onto the subset Us of all prime ideals of K that do
not contain s; this subset Us is open because the complementary subset is V(Ks).
The elements of Us correspond to the irreducible algebraic submanifolds V (p)
contained in Us .
Let us examine what happens when Spec(K) is not a connected topological
space.
(1.11.3) Theorem. When K is a direct sum n of ideals K1 , K2 , . . . , Kn , each
Ki is generated by an idempotent ei and i=1 ei = 1 ; moreover the mapping
Spec(Ki ) → Spec(K) associated with the projection K → Ki induces a bijection
from Spec(Ki ) onto an open subset Ui of Spec(K), and Spec(K) is the disjoint
union of the open subsets U1 , U2 , . . . ,Un . Conversely when Spec(K) is a dis-
joint union of open subsets U1 , U2 , . . . , Un , then K is a direct sum of ideals K1 ,
K2 , . . . , Kn in such a way that each open subset Ui is the image of the mapping
Spec(Ki ) → Spec(K).
n
Proof. Let us suppose that K = i=1 Ki and let ei be the component of 1 in Ki
for i = 1, 2, . . . , n. The equality λ = ni=1 λei holds for every λ ∈ K, and proves
that λei is the projection of λ in Ki ; in particular e2i = ei and Ki = Kei . A prime
ideal cannot contain all the n idempotents ei because their sum is 1; if it does
not contain ei , it must contain all ej such that j = i, because ei ej = 0. Therefore
each prime ideal contains all the n idempotents ei except one, and Spec(K) is the
disjoint union of the n subsets Ui defined in this way: Ui is the set of all prime
ideals containing ej whenever j = i. Since Ui is equal both to V(K(1 − ei )) and
to the subset complementary to V(Kei ), it is open and closed. The continuous
mapping Spec(Ki ) → Spec(K) maps each prime ideal pi of Ki to the prime ideal
of K which is the direct sum of pi and all Kj with j = i; thus we get a bijection
Spec(Ki ) → Ui .
Random documents with unrelated
content Scribd suggests to you:
And at these encouraging words Piglet felt quite happy again, and
decided not to be a Sailor after all. So when Christopher Robin had
helped them out of the Gravel Pit, they all went off together hand-in-
hand.
And two days later Rabbit happened to meet Eeyore in the Forest.
"Hallo, Eeyore," he said, "what are you looking for?"
"Small, of course," said Eeyore. "Haven't you any brain?"
"Oh, but didn't I tell you?" said Rabbit. "Small was found two days
ago."
There was a moment's silence.
"Ha-ha," said Eeyore bitterly. "Merriment and what-not. Don't
apologize. It's just what would happen."
CHAPTER IV
IN WHICH It Is Shown That Tiggers Don't Climb Trees
One day when Pooh was thinking, he thought he would go and see
Eeyore, because he hadn't seen him since yesterday. And as he
walked through the heather, singing to himself, he suddenly
remembered that he hadn't seen Owl since the day before yesterday,
so he thought that he would just look in at the Hundred Acre Wood on
the way and see if Owl was at home.
Well, he went on singing, until he came to the part of the stream
where the stepping-stones were, and when he was in the middle of
the third stone he began to wonder how Kanga and Roo and Tigger
were getting on, because they all lived together in a different part of
the Forest. And he thought, "I haven't seen Roo for a long time, and if
I don't see him today it will be a still longer time." So he sat down on
the stone in the middle of the stream, and sang another verse of his
song, while he wondered what to do.
The other verse of the song was like this:
I could spend a happy morning

Seeing Roo,
Being Pooh.
For it doesn't seem to matter,
If I don't get any fatter
(And I don't get any fatter),
What I do.
The sun was so delightfully warm, and the stone, which had been
sitting in it for a long time, was so warm, too, that Pooh had almost
decided to go on being Pooh in the middle of the stream for the rest
of the morning, when he remembered Rabbit.
"Rabbit," said Pooh to himself. "I like talking to Rabbit. He talks about
sensible things. He doesn't use long, difficult words, like Owl. He uses
short, easy words, like 'What about lunch?' and 'Help yourself, Pooh.'
I suppose really, I ought to go and see Rabbit."
Which made him think of another verse:
Oh, I like his way of talking,

Yes, I do.
It's the nicest way of talking
Just for two.
And a Help-yourself with Rabbit
Though it may become a habit,
Is a pleasant sort of habit
For a Pooh.
So when he had sung this, he got up off his stone, walked back
across the stream, and set off for Rabbit's house.
But he hadn't got far before he began to say to himself:
"Yes, but suppose Rabbit is out?"
"Or suppose I get stuck in his front door again, coming out, as I did
once when his front door wasn't big enough?"
"Because I know I'm not getting fatter, but his front door may be
getting thinner."
"So wouldn't it be better if——"
And all the time he was saying things like this he was going more and
more westerly, without thinking ... until suddenly he found himself at
his own front door again.
And it was eleven o'clock.
Which was Time-for-a-little-something....
Half an hour later he was doing what he had always really meant to
do, he was stumping off to Piglet's house. And as he walked, he
wiped his mouth with the back of his paw, and sang rather a fluffy
song through the fur. It went like this:

Seeing Piglet.
And I couldn't spend a happy morning
Not seeing Piglet.
And it doesn't seem to matter
If I don't see Owl and Eeyore
(or any of the others),
And I'm not going to see Owl or Eeyore
(or any of the others)
Or Christopher Robin.
Written down, like this, it doesn't seem a very good song, but coming
through pale fawn fluff at about half-past eleven on a very sunny
morning, it seemed to Pooh to be one of the best songs he had ever
sung. So he went on singing it.
Piglet was busy digging a small hole in the ground outside his house.
"Hallo, Piglet," said Pooh.
"Hallo, Pooh," said Piglet, giving a jump of surprise. "I knew it was
you."
"So did I," said Pooh. "What are you doing?"
"I'm planting a haycorn, Pooh, so that it can grow up into an oak-tree,
and have lots of haycorns just outside the front door instead of having
to walk miles and miles, do you see, Pooh?"
"Supposing it doesn't?" said Pooh.
"It will, because Christopher Robin says it will, so that's why I'm
planting it."
"Well," said Pooh, "if I plant a honeycomb outside my house, then it
will grow up into a beehive."
Piglet wasn't quite sure about this.
"Or a piece of a honeycomb," said Pooh, "so as not to waste too
much. Only then I might only get a piece of a beehive, and it might be
the wrong piece, where the bees were buzzing and not hunnying.
Bother."
Piglet agreed that that would be rather bothering.
"Besides, Pooh, it's a very difficult thing, planting unless you know
how to do it," he said; and he put the acorn in the hole he had made,
and covered it up with earth, and jumped on it.
"I do know," said Pooh, "because Christopher Robin gave me a

mastershalum seed, and I planted it, and I'm going to have
mastershalums all over the front door."
"I thought they were called nasturtiums," said Piglet timidly, as he
went on jumping.
"No," said Pooh. "Not these. These are called mastershalums."
When Piglet had finished jumping, he wiped his paws on his front,
and said, "What shall we do now?" and Pooh said, "Let's go and see
Kanga and Roo and Tigger," and Piglet said, "Y-yes. L-lets"—
because he was still a little anxious about Tigger, who was a Very
Bouncy Animal, with a way of saying How-do-you-do, which always
left your ears full of sand, even after Kanga had said, "Gently, Tigger
dear," and had helped you up again. So they set off for Kanga's
house.
Now it happened that Kanga had felt rather motherly that morning,
and Wanting to Count Things—like Roo's vests, and how many
pieces of soap there were left, and the two clean spots in Tigger's
feeder; so she had sent them out with a packet of watercress
sandwiches for Roo and a packet of extract-of-malt sandwiches for
Tigger, to have a nice long morning in the Forest not getting into
mischief. And off they had gone.
And as they went, Tigger told Roo (who wanted to know) all about the
things that Tiggers could do.
"Can they fly?" asked Roo.
"Yes," said Tigger, "they're very good flyers, Tiggers are. Stornry good
flyers."
"Oo!" said Roo. "Can they fly as well as Owl?"
"Yes," said Tigger. "Only they don't want to."
"Why don't they want to?"
"Well, they just don't like it, somehow."
Roo couldn't understand this, because he thought it would be lovely
to be able to fly, but Tigger said it was difficult to explain to anybody
who wasn't a Tigger himself.
"Well," said Roo, "can they jump as far as Kangas?"
"Yes," said Tigger. "When they want to."
"I love jumping," said Roo. "Let's see who can jump farthest, you or
me."
"I can," said Tigger. "But we mustn't stop now, or we shall be late."
"Late for what?"
"For whatever we want to be in time for," said Tigger, hurrying on.
In a little while they came to the Six Pine Trees.
"I can swim," said Roo. "I fell into the river, and I swimmed. Can
Tiggers swim?"
"Of course they can. Tiggers can do everything."
"Can they climb trees better than Pooh?" asked Roo, stopping under
the tallest Pine Tree, and looking up at it.
"Climbing trees is what they do best," said Tigger. "Much better than
Poohs."
"Could they climb this one?"
"They're always climbing trees like that," said Tigger. "Up and down
all day."
"Oo, Tigger, are they really?"
"I'll show you," said Tigger bravely, "and you can sit on my back and
watch me." For of all the things which he had said Tiggers could do,
the only one he felt really certain about suddenly was climbing trees.
"Oo, Tigger, oo, Tigger, oo, Tigger!" squeaked Roo excitedly.
So he sat on Tigger's back and up they went.
And for the first ten feet Tigger said happily to himself, "Up we go!"
And for the next ten feet he said:
"I always said Tiggers could climb trees."
"Not that it's easy, mind you."
"Of course, there's the coming-down too. Backwards."
And then he said:
"Which will be difficult ...
"Unless one fell ...
"when it would be ...
"EASY."
And at the word "easy" the branch he was standing on broke
suddenly, and he just managed to clutch at the one above him as he
felt himself going ... and then slowly he got his chin over it ... and then
one back paw ... and then the other ... until at last he was sitting on it,
breathing very quickly, and wishing that he had gone in for swimming
instead.
Roo climbed off, and sat down next to him.
"Oo, Tigger," he said excitedly, "are we at the top?"
"No," said Tigger.
"Are we going to the top?"
"No," said Tigger.
"Oh!" said Roo rather sadly. And then he went on hopefully: "That
was a lovely bit just now, when you pretended we were going to fall-
bump-to-the-bottom, and we didn't. Will you do that bit again?"
"NO," said Tigger.
Roo was silent for a little while, and then he said, "Shall we eat our
sandwiches, Tigger?" And Tigger said, "Yes, where are they?" And
Roo said, "At the bottom of the tree." And Tigger said, "I don't think
we'd better eat them just yet." So they didn't.
By and by Pooh and Piglet came along. Pooh was telling Piglet in a
singing voice that it didn't seem to matter, if he didn't get any fatter,
and he didn't think he was getting any fatter, what he did; and Piglet
was wondering how long it would be before his haycorn came up.
"Look, Pooh!" said Piglet suddenly. "There's something in one of the
Pine Trees."
"So there is!" said Pooh, looking up wonderingly. "There's an Animal."
Piglet took Pooh's arm, in case Pooh was frightened.
"Is it One of the Fiercer Animals?" he said, looking the other way.
Pooh nodded.
"It's a Jagular," he said.
"What do Jagulars do?" asked Piglet, hoping that they wouldn't.
"They hide in the branches of trees, and drop on you as you go
underneath," said Pooh. "Christopher Robin told me."
"Perhaps we better hadn't go underneath, Pooh. In case he dropped
and hurt himself."
"They don't hurt themselves," said Pooh. "They're such very good
droppers."
Piglet still felt that to be underneath a Very Good Dropper would be a
Mistake, and he was just going to hurry back for something which he
had forgotten when the Jagular called out to them.
"Help! Help!" it called.
"That's what Jagulars always do," said Pooh, much interested. "They
call 'Help! Help!' and then when you look up, they drop on you."
"I'm looking down," cried Piglet loudly, so as the Jagular shouldn't do
the wrong thing by accident.
Something very excited next to the Jagular heard him, and squeaked:
"Pooh and Piglet! Pooh and Piglet!"
All of a sudden Piglet felt that it was a much nicer day than he had
thought it was. All warm and sunny——
"Pooh!" he cried. "I believe it's Tigger and Roo!"
"So it is," said Pooh. "I thought it was a Jagular and another Jagular."
"Hallo, Roo!" called Piglet. "What are you doing?"
"We can't get down, we can't get down!" cried Roo. "Isn't it fun? Pooh,
isn't it fun, Tigger and I are living in a tree, like Owl, and we're going
to stay here for ever and ever. I can see Piglet's house. Piglet, I can
see your house from here. Aren't we high? Is Owl's house as high up
as this?"
"How did you get there, Roo?" asked Piglet.

"On Tigger's back! And Tiggers can't climb downwards, because their
tails get in the way, only upwards, and Tigger forgot about that when
we started, and he's only just remembered. So we've got to stay here
for ever and ever—unless we go higher. What did you say, Tigger?
Oh, Tigger says if we go higher we shan't be able to see Piglet's
house so well, so we're going to stop here."
"Piglet," said Pooh solemnly, when he had heard all this, "what shall
we do?" And he began to eat Tigger's sandwiches.
"Are they stuck?" asked Piglet anxiously.
Pooh nodded.
"Couldn't you climb up to them?"
"I might, Piglet, and I might bring Roo down on my back, but I couldn't
bring Tigger down. So we must think of something else." And in a
thoughtful way he began to eat Roo's sandwiches, too.
Whether he would have thought of anything before he had finished

the last sandwich, I don't know, but he had just got to the last but one
when there was a crackling in the bracken, and Christopher Robin
and Eeyore came strolling along together.
"I shouldn't be surprised if it hailed a good deal tomorrow," Eeyore
was saying. "Blizzards and what-not. Being fine today doesn't Mean
Anything. It has no sig—what's that word? Well, it has none of that.
It's just a small piece of weather."
"There's Pooh!" said Christopher Robin, who didn't much mind what it
did tomorrow, as long as he was out in it. "Hallo, Pooh!"
"It's Christopher Robin!" said Piglet. "He'll know what to do."
They hurried up to him.
"Oh, Christopher Robin," began Pooh.
"And Eeyore," said Eeyore.
"Tigger and Roo are right up the Six Pine Trees, and they can't get
down, and——"
"And I was just saying," put in Piglet, "that if only Christopher Robin
——"
"And Eeyore——"
"If only you were here, then we could think of something to do."
Christopher Robin looked up at Tigger and Roo, and tried to think of
something.
"I thought," said Piglet earnestly, "that if Eeyore stood at the bottom of
the tree, and if Pooh stood on Eeyore's back, and if I stood on Pooh's
shoulders——"
"And if Eeyore's back snapped suddenly, then we could all laugh. Ha
ha! Amusing in a quiet way," said Eeyore, "but not really helpful."
"Well," said Piglet meekly, "I thought——"
"Would it break your back, Eeyore?" asked Pooh, very much
surprised.
"That's what would be so interesting, Pooh. Not being quite sure till
afterwards."
Pooh said "Oh!" and they all began to think again.
"I've got an idea!" cried Christopher Robin suddenly.
"Listen to this, Piglet," said Eeyore, "and then you'll know what we're
trying to do."
"I'll take off my tunic and we'll each hold a corner, and then Roo and
Tigger can jump into it, and it will be all soft and bouncy for them, and
they won't hurt themselves."
"Getting Tigger down," said Eeyore, "and Not hurting anybody. Keep
those two ideas in your head, Piglet, and you'll be all right."
But Piglet wasn't listening, he was so agog at the thought of seeing
Christopher Robin's blue braces again. He had only seen them once
before, when he was much younger, and, being a little over-excited
by them, had had to go to bed half an hour earlier than usual; and he
had always wondered since if they were really as blue and as bracing
as he had thought them. So when Christopher Robin took his tunic
off, and they were, he felt quite friendly to Eeyore again, and held the
corner of the tunic next to him and smiled happily at him. And Eeyore
whispered back: "I'm not saying there won't be an Accident now, mind
you. They're funny things, Accidents. You never have them till you're
having them."
When Roo understood what he had to do, he was wildly excited, and
cried out: "Tigger, Tigger, we're going to jump! Look at me jumping,
Tigger! Like flying, my jumping will be. Can Tiggers do it?" And he
squeaked out: "I'm coming, Christopher Robin!" and he jumped—
straight into the middle of the tunic. And he was going so fast that he
bounced up again almost as high as where he was before—and went
on bouncing and saying, "Oo!" for quite a long time—and then at last
he stopped and said, "Oo, lovely!" And they put him on the ground.
"Come on, Tigger," he called out. "It's easy."
But Tigger was holding on to the branch and saying to himself: "It's all
very well for Jumping Animals like Kangas, but it's quite different for
Swimming Animals like Tiggers." And he thought of himself floating
on his back down a river, or striking out from one island to another,
and he felt that that was really the life for a Tigger.
"Come along," called Christopher Robin. "You'll be all right."
"Just wait a moment," said Tigger nervously. "Small piece of bark in
my eye." And he moved slowly along his branch.
"Come on, it's easy!" squeaked Roo. And suddenly Tigger found how
easy it was.
"Ow!" he shouted as the tree flew past him.
"Look out!" cried Christopher Robin to the others.
There was a crash, and a tearing noise, and a confused heap of
everybody on the ground.
Christopher Robin and Pooh and Piglet picked themselves up first,
and then they picked Tigger up, and underneath everybody else was
Eeyore.
"Oh, Eeyore!" cried Christopher Robin. "Are you hurt?" And he felt
him rather anxiously, and dusted him and helped him to stand up
again.
Eeyore said nothing for a long time. And then he said: "Is Tigger
there?"
Tigger was there, feeling Bouncy again already.
"Yes," said Christopher Robin. "Tigger's here."
"Well, just thank him for me," said Eeyore.
CHAPTER V
IN WHICH Rabbit Has a Busy Day, and We Learn What Christopher
Robin Does in the Mornings
It was going to be one of Rabbit's busy days. As soon as he woke up
he felt important, as if everything depended upon him. It was just the
day for Organizing Something, or for Writing a Notice Signed Rabbit,
or for Seeing What Everybody Else Thought About It. It was a perfect
morning for hurrying round to Pooh, and saying, "Very well, then, I'll
tell Piglet," and then going to Piglet, and saying, "Pooh thinks—but
perhaps I'd better see Owl first." It was a Captainish sort of day, when
everybody said, "Yes, Rabbit" and "No, Rabbit," and waited until he
had told them.
He came out of his house and sniffed the warm spring morning as he
wondered what he would do. Kanga's house was nearest, and at
Kanga's house was Roo, who said "Yes, Rabbit" and "No, Rabbit"
almost better than anybody else in the Forest; but there was another
animal there nowadays, the strange and Bouncy Tigger; and he was
the sort of Tigger who was always in front when you were showing
him the way anywhere, and was generally out of sight when at last
you came to the place and said proudly "Here we are!"
"No, not Kanga's," said Rabbit thoughtfully to himself, as he curled his
whiskers in the sun; and, to make quite sure that he wasn't going
there, he turned to the left and trotted off in the other direction, which
was the way to Christopher Robin's house.
"After all," said Rabbit to himself, "Christopher Robin depends on Me.
He's fond of Pooh and Piglet and Eeyore, and so am I, but they
haven't any Brain. Not to notice. And he respects Owl, because you
can't help respecting anybody who can spell TUESDAY, even if he
doesn't spell it right; but spelling isn't everything. There are days
when spelling Tuesday simply doesn't count. And Kanga is too busy
looking after Roo, and Roo is too young and Tigger is too bouncy to
be any help, so there's really nobody but Me, when you come to look
at it. I'll go and see if there's anything he wants doing, and then I'll do
it for him. It's just the day for doing things."
He trotted along happily, and by-and-by he crossed the stream and
came to the place where his friends-and-relations lived. There
seemed to be even more of them about than usual this morning, and
having nodded to a hedgehog or two, with whom he was too busy to
shake hands, and having said, "Good morning, good morning,"
importantly to some of the others, and "Ah, there you are," kindly, to
the smaller ones, he waved a paw at them over his shoulder, and was
gone; leaving such an air of excitement and I-don't-know-what behind
him, that several members of the Beetle family, including Henry Rush,
made their way at once to the Hundred Acre Wood and began
climbing trees, in the hope of getting to the top before it happened,
whatever it was, so that they might see it properly.
Rabbit hurried on by the edge of the Hundred Acre Wood, feeling
more important every minute, and soon he came to the tree where
Christopher Robin lived. He knocked at the door, and he called out
once or twice, and then he walked back a little way and put his paw
up to keep the sun out, and called to the top of the tree, and then he
turned all round and shouted "Hallo!" and "I say!" "It's Rabbit!"—but
nothing happened. Then he stopped and listened, and everything
stopped and listened with him, and the Forest was very lone and still
and peaceful in the sunshine, until suddenly a hundred miles above
him a lark began to sing.
"Bother!" said Rabbit. "He's gone out."
He went back to the green front door, just to make sure, and he was
turning away, feeling that his morning had got all spoilt, when he saw
a piece of paper on the ground. And there was a pin in it, as if it had
fallen off the door.
"Ha!" said Rabbit, feeling quite happy again. "Another notice!"
This is what it said:
GON OUT
BACKSON
BISY
BACKSON.
C. R.
"Ha!" said Rabbit again. "I must tell the others." And he hurried off
importantly.
The nearest house was Owl's, and to Owl's House in the Hundred
Acre Wood he made his way. He came to Owl's door, and he knocked
and he rang, and he rang and he knocked, and at last Owl's head
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
Let us accompany you on the journey of exploring knowledge and

personal growth!
ebookfinal.com

10248

Uploaded by

Copyright:

Available Formats

10248

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

10248

Uploaded by

Copyright:

Available Formats

Download the full version of the ebook at

Quadratic mappings and Clifford algebras 1st

Explore and download more ebook at https://ebookfinal.com

Clifford Algebras Applications to Mathematics Physics and

Quasiconformal Mappings Riemann Surfaces and Teichmuller

Handbook of Conformal Mappings and Applications 1st

Hyperbolic Manifolds and Holomorphic Mappings An

Elements of the Representation Theory of Associative

Re Mapping Archaeology Critical Perspectives Alternative

Monomial Algebras Second Edition Rafael Villarreal

Neuroradiology cases 1st Edition Clifford J. Eskey

Jacques Helmstetter Artibano Micali

2000 Mathematical Subject Classiﬁcation: 15A66, 15A69, 15A63, 15A75, 16H05

Library of Congress Control Number: 2007942636

ISBN 978-3-7643-8605-4 Birkhäuser Verlag AG, Basel - Boston - Berlin

© 2008 Birkhäuser Verlag AG

ISBN 978-3-7643-8605-4 e-ISBN 978-3-7643-8606-1

6.6 Graded central simple algebras over a ﬁeld . . . . . . . . . . . . . . 353

Readers are assumed already to know elementary algebra (rings, ﬁelds,

(f ∧ g)(x ∧ y) = −f (x) g(y) + f (y) g(x) ;

Lipschitz, the forgotten pioneer

cliffordian ideas to difficult problems involving differential operators (the “holo-

There is no systematic presentation of Weyl algebras in this book, which

Jacques Helmstetter Artibano Micali

1.1 Some general notation and deﬁnitions

that the equality f (1A ) = 1B is also required. By deﬁnition a subring of A must

1.2 Universal objects in a category

source) of a morphism in C ∗ is by deﬁnition its source (resp. its target) in C.

1.3 Examples of universal objects

Freely generated modules

es (s) = 1 and es (t) = 0 whenever s = t. It is clear that K (S) is a free K-module

Tensor products of modules

(1.3.2) Proposition. For any bilinear mapping f : M × N → P there exists a

it is valid when M is a K-module, P an L-module, and N a module over K and

Tensor products of algebras

Direct products and direct sums

of linear mappings fj : P → Mj determines a

The objects of C are the families (fj ) of linear mappings Mj → P with

1.4 Tensor algebras and symmetric algebras

f (x1 ⊗ x2 ⊗ · · · ⊗ xk ) = f (x1 )f (x2 ) · · · f (xk ).

Since there is an algebra freely generated by the module M in the category

same way by means of commutative polynomials. As explained above, we can also

x1 ⊗ x2 ⊗ · · · ⊗ xj−1 ⊗ (xj ⊗ xj+1 − xj+1 ⊗ xj ) ⊗ xj+2 ⊗ · · · ⊗ xk

with j ∈ {1, 2, . . . , k − 1} ; if the k-linear mapping f is symmetric, the mapping

(1.5.1) Theorem. The symmetric algebra of the module M ⊕ N is canonically

HomC (M, N ) −→ HomC (M , N ), u −→ f2 uf1 .

Thus we have got a functor because

to get a reciprocal morphism, it suﬃces to consider the evident bilinear mapping

If N is a free module, it is a direct sum of submodules all isomorphic to K,

1.6 Exact sequences

it is called an exact sequence if Im(u) = Ker(v). The inclusion Im(u) ⊂ Ker(v) is

is exact if u is injective, v surjective, and Im(u) = Ker(v) .

M2 = Ker(u ) = Im(v ), u u = idM and vv = idM ;

all this implies idM = uu + v v and u v = 0.

vu = 0, u v = 0, u u = idM , vv = idM and uu + v v = idM , (1.6.2)

then we get two splitting exact sequences:

we get these exact sequences:

0 −→ Hom(N , P ) −→ Hom(N, P ) −→ Hom(N , P ) ,

implies the exactness of the sequence

this is proved in Exercise (1.ex.10).

es (s) = 1 and es (t) = 0 whenever s = t. It is clear that K (S) is a free K-module

HomC (M, N ) −→ HomC (M , N ), u −→ f2 uf1 .

Whereas the functor M −→ S −1 M has a nice behaviour relatively to tensor