Universitext

Jean-ClaudeHausmann

Mod Two
Homology
and Cohomology
Universitext
Universitext

Series editors
Sheldon Axler
San Francisco State University, San Francisco, CA, USA
Vincenzo Capasso
Universit degli Studi di Milano, Milan, Italy
Carles Casacuberta
Universitat de Barcelona, Barcelona, Spain

Angus MacIntyre
Queen Mary University of London, London, UK
Kenneth Ribet
University of California, Berkeley, CA, USA
Claude Sabbah
CNRS, cole polytechnique Centre de mathmatiques, Palaiseau, France

Endre Sli
University of Oxford, Oxford, UK
Wojbor A. Woyczynski
Case Western Reserve University, Cleveland, OH, USA

Universitext is a series of textbooks that presents material from a wide variety of

mathematical disciplines at masters level and beyond. The books, often well
class-tested by their author, may have an informal, personal, even experimental
approach to their subject matter. Some of the most successful and established
books in the series have evolved through several editions, always following the
evolution of teaching curricula, into very polished texts.

Thus as research topics trickle down into graduate-level teaching, first textbooks
written for new, cutting-edge courses may make their way into Universitext.

More information about this series at [Link]

Jean-Claude Hausmann

Mod Two Homology

and Cohomology

123
Jean-Claude Hausmann
University of Geneva
Geneva
Switzerland

ISSN 0172-5939 ISSN 2191-6675 (electronic)

ISBN 978-3-319-09353-6 ISBN 978-3-319-09354-3 (eBook)
DOI 10.1007/978-3-319-09354-3

Library of Congress Control Number: 2014944717

JEL Classification Code: 55-01, 55N10, 55N91, 57R91, 57R20, 55U10, 55U25, 55S10, 57R19, 55R91

Springer Cham Heidelberg New York Dordrecht London

Springer International Publishing Switzerland 2014

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the
work. Duplication of this publication or parts thereof is permitted only under the provisions of
the Copyright Law of the Publishers location, in its current version, and permission for use must
always be obtained from Springer. Permissions for use may be obtained through RightsLink at the
Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.

Printed on acid-free paper

Springer is part of Springer Science+Business Media ([Link])

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Simplicial (Co)homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Simplicial Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Definitions of Simplicial (Co)homology. . . . . . . . . . . . . . . . . 11
2.3 Kronecker Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 First Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.1 Reduction to Components . . . . . . . . . . . . . . . . . . . . . 20
2.4.2 0-Dimensional (Co)homology. . . . . . . . . . . . . . . . . . . 20
2.4.3 Pseudomanifolds. . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.4 Poincar Series and Polynomials . . . . . . . . . . . . . . . . 23
2.4.5 (Co)homology of a Cone . . . . . . . . . . . . . . . . . . . . . . 23
2.4.6 The Euler Characteristic . . . . . . . . . . . . . . . . . . . . . . 25
2.4.7 Surfaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 The Homomorphism Induced by a Simplicial Map . . . . . . . . . 30
2.6 Exact Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.7 Relative (Co)homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.8 Mayer-Vietoris Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.9 Appendix A: An Acyclic Carrier Result. . . . . . . . . . . . . . . . . 49
2.10 Appendix B: Ordered Simplicial (Co)homology . . . . . . . . . . . 50
2.11 Exercises for Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3 Singular and Cellular (Co)homologies . . . . . . . . . . . . . . . . . . . . . 59

3.1 Singular (Co)homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.1.2 Relative Singular (Co)homology. . . . . . . . . . . . . . . . . 66
3.1.3 The Homotopy Property . . . . . . . . . . . . . . . . . . . . . . 73
3.1.4 Excision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.1.5 Well Cofibrant Pairs . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.1.6 Mayer-Vietoris Sequences . . . . . . . . . . . . . . . . . . . . . 87

v
vi Contents

3.2 Spheres, Disks, Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

3.3 Classical Applications of the mod 2 (Co)homology . . . . . . . . . 94
3.4 CW-Complexes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.5 Cellular (Co)homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.6 Isomorphisms Between Simplicial and Singular
(Co)homology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.7 CW-Approximations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.8 Eilenberg-MacLane Spaces. . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.9 Generalized Cohomology Theories . . . . . . . . . . . . . . . . . . . . 123
3.10 Exercises for Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

4 Products. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.1 The Cup Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.1.1 The Cup Product in Simplicial Cohomology . . . . . . . . 127
4.1.2 The Cup Product in Singular Cohomology. . . . . . . . . . 131
4.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2.1 Disjoint Unions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2.2 Bouquets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2.3 Connected Sum(s) of Closed Topological Manifolds . . . 135
4.2.4 Cohomology Algebras of Surfaces . . . . . . . . . . . . . . . 137
4.3 Two-Fold Coverings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.3.1 H 1 , Fundamental Group and 2-Fold Coverings. . . . . . . 139
4.3.2 The Characteristic Class . . . . . . . . . . . . . . . . . . . . . . 142
4.3.3 The Transfer Exact Sequence of a 2-Fold Covering . . . 144
4.3.4 The Cohomology Ring of RPn . . . . . . . . . . . . . . . . . . 146
4.4 Nilpotency, Lusternik-Schnirelmann Categories
and Topological Complexity. . . . . . . . . . . . . . . . . . . . . . . . . 147
4.5 The Cap Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.6 The Cross Product and the Knneth Theorem. . . . . . . . . . . . . 156
4.7 Some Applications of the Knneth Theorem . . . . . . . . . . . . . 165
4.7.1 Poincar Series and Euler Characteristic of a Product . . 165
4.7.2 Slices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.7.3 The Cohomology Ring of a Product of Spheres . . . . . . 166
4.7.4 Smash Products and Joins . . . . . . . . . . . . . . . . . . . . . 167
4.7.5 The Theorem of Leray-Hirsch . . . . . . . . . . . . . . . . . . 172
4.7.6 The Thom Isomorphism . . . . . . . . . . . . . . . . . . . . . . 180
4.7.7 Bundles Over Spheres . . . . . . . . . . . . . . . . . . . . . . . . 189
4.7.8 The Face Space of a Simplicial Complex . . . . . . . . . . 194
4.7.9 Continuous Multiplications on KZ2 ; m . . . . . . . . . . . 196
4.8 Exercises for Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Contents vii

5 Poincar Duality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

5.1 Algebraic Topology and Manifolds . . . . . . . . . . . . . . . . . . . . 201
5.2 Poincar Duality in Polyhedral Homology Manifolds . . . . . . . 202
5.3 Other Forms of Poincar Duality. . . . . . . . . . . . . . . . . . . . . . 211
5.3.1 Relative Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . 211
5.3.2 Manifolds with Boundary . . . . . . . . . . . . . . . . . . . . . 215
5.3.3 The Intersection Form . . . . . . . . . . . . . . . . . . . . . . . . 217
5.3.4 Non Degeneracy of the Cup Product. . . . . . . . . . . . . . 219
5.3.5 Alexander Duality. . . . . . . . . . . . . . . . . . . . . . . . . . . 220
5.4 Poincar Duality and Submanifolds . . . . . . . . . . . . . . . . . . . . 221
5.4.1 The Poincar Dual of a Submanifold . . . . . . . . . . . . . 221
5.4.2 The Gysin Homomorphism . . . . . . . . . . . . . . . . . . . . 225
5.4.3 Intersections of Submanifolds. . . . . . . . . . . . . . . . . . . 227
5.4.4 The Linking Number. . . . . . . . . . . . . . . . . . . . . . . . . 231
5.5 Exercises for Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236

6 Projective Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 239

6.1 The Cohomology Ring of Projective SpacesHopf
Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
6.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
6.2.1 The Borsuk-Ulam Theorem . . . . . . . . . . . . . . . . . . . . 245
6.2.2 Non-singular and Axial Maps. . . . . . . . . . . . . . . . . . . 246
6.3 The Hopf Invariant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
6.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
6.3.2 The Hopf Invariant and Continuous Multiplications. . . . 250
6.3.3 Dimension Restrictions . . . . . . . . . . . . . . . . . . . . . . . 252
6.3.4 Hopf Invariant and Linking Numbers . . . . . . . . . . . . . 253
6.4 Exercises for Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

7 Equivariant Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

7.1 Spaces with Involution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
7.2 The General Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
7.3 Localization Theorems and Smith Theory . . . . . . . . . . . . . . . 283
7.4 Equivariant Cross Products and Knneth Theorems. . . . . . . . . 288
7.5 Equivariant Bundles and Euler Classes . . . . . . . . . . . . . . . . . 298
7.6 Equivariant Morse-Bott Theory. . . . . . . . . . . . . . . . . . . . . . . 309
7.7 Exercises for Chapter 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321

8 Steenrod Squares . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325

8.1 Cohomology Operations. . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
8.2 Properties of Steenrod Squares . . . . . . . . . . . . . . . . . . . . . . . 330
8.3 Construction of Steenrod Squares . . . . . . . . . . . . . . . . . . . . . 333
8.4 Adem Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
viii Contents

8.5 The Steenrod Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346

8.6 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
8.7 Exercises for Chapter 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354

9 Stiefel-Whitney Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355

9.1 Trivializations and Structures on Vector Bundles . . . . . . . . . . 355
9.2 The Class w1 Orientability . . . . . . . . . . . . . . . . . . . . . . . . . 363
9.3 The Class w_ 2 Spin Structures . . . . . . . . . . . . . . . . . . . . . . . 367
9.4 Definition and Properties of Stiefel-Whitney Classes . . . . . . . . 372
9.5 Real Flag Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
9.5.1 Definitions and Morse Theory . . . . . . . . . . . . . . . . . . 376
9.5.2 Cohomology Rings . . . . . . . . . . . . . . . . . . . . . . . . . . 381
9.5.3 Schubert Cells and Stiefel-Whitney Classes . . . . . . . . . 389
9.6 Splitting Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
9.7 Complex Flag Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
9.8 The Wu Formula. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411
9.8.1 Wus Classes and Formula. . . . . . . . . . . . . . . . . . . . . 411
9.8.2 Orientability and Spin Structures . . . . . . . . . . . . . . . . 415
9.8.3 Applications to 3-Manifolds. . . . . . . . . . . . . . . . . . . . 418
9.8.4 The Universal Class for Double Points . . . . . . . . . . . . 420
9.9 Thoms Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
9.9.1 Representing Homology Classes by Manifolds . . . . . . . 425
9.9.2 Cobordism and Stiefel-Whitney Numbers. . . . . . . . . . . 429
9.10 Exercises for Chapter 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432

10 Miscellaneous Applications and Developments . . . . . . . . . . . . . . . 433

10.1 Actions with Scattered or Discrete Fixed Point Sets . . . . . . . . 433
10.2 Conjugation Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
10.3 Chain and Polygon Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 444
10.3.1 Definitions and Basic Properties . . . . . . . . . . . . . . . . . 444
10.3.2 Equivariant Cohomology . . . . . . . . . . . . . . . . . . . . . . 449
10.3.3 Non-equivariant Cohomology. . . . . . . . . . . . . . . . . . . 458
10.3.4 The Inverse Problem . . . . . . . . . . . . . . . . . . . . . . . . . 464
10.3.5 Spatial Polygon Spaces and Conjugation Spaces. . . . . . 468
10.4 Equivariant Characteristic Classes . . . . . . . . . . . . . . . . . . . . . 470
10.5 The Equivariant Cohomology of Certain Homogeneous
Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 475
10.6 The Kervaire Invariant. . . . . . . . . . . . . . . . . . . . . . . . ..... 485
10.7 Exercises for Chapter 10 . . . . . . . . . . . . . . . . . . . . . . ..... 498
Contents ix

11 Hints and Answers for Some Exercises . . . . . . . . . . . . . . . . . . . . 501

11.1 Exercises for Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
11.2 Exercises for Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
11.3 Exercises for Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
11.4 Exercises for Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
11.5 Exercises for Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
11.6 Exercises for Chapter 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
11.7 Exercises for Chapter 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
11.8 Exercises for Chapter 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
11.9 Exercises for Chapter 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . 514

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
Chapter 1
Introduction

Mod 2 homology first occurred in 1908 in a paper of Tietze [196] (see also
[40, pp. 4142]). Several results were first established using this mod 2 approach, like
the linking number for submanifolds in Rn (see Sect. 5.4.4), as well as Alexander
duality [7]. One argument in favor of the choice of the mod 2 homology was its
simplicity, as J.W. Alexander says in his introduction: The theory of connectivity
[homology] may be approached from two different angles depending on whether
or not the notion of sense [orientation] is developed and taken into consideration.
We have adopted the second and somewhat simpler point of view in this discussion
in order to condense the necessary preliminaries as much as possible. A treatment
involving the idea of sense would be somewhat more complicated but would follow
along much the same lines.
Besides being simpler than its integral counterpart, mod 2 homology sometimes
gives new theorems. The first historical main example is the generalization of
Poincar duality to all closed manifolds, whether orientable or not, a result obtained
by Veblen and Alexander in 1913 [200]. As a consequence, the Euler characteristic
of a closed odd-dimensional manifold vanishes.
The discoveries of Stiefel-Whitney classes in 19361938 and of Steenrod squares
in 19471950 gave mod 2 cohomology the status of a major tool in algebraic topol-
ogy, providing for instance the theory of spin structures and Thoms work on the
cobordism ring.
These notes are an introduction, at graduate students level, of mod 2 (co)homology
(there will be essentially no other). They include classical applications (Brouwer fixed
point theorem, Poincar duality, Borsuk-Ulam theorem, Smith theory, etc) and less
classical ones (face spaces, topological complexity, equivariant Morse theory, etc).
The cohomology of flag manifolds is treated in details, including for Grassmanni-
ans the relationship between Stiefel-Whitney classes and Schubert calculus. Some
original applications are given in Chap. 10.
Our approach is different than that of classical textbooks, in which mod 2 (co)
homology is just a particular case of (co)homology with arbitrary coefficients. Also,
most authors start with a full account of homology before approaching cohomology.
In these notes, mod 2 (co)homology is treated as a subject by itself and we start
Springer International Publishing Switzerland 2014 1
J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_1
2 1 Introduction

with cohomology and homology together from the beginning. The advantages of
this approach are the following.
The definition of a (co)chain is simple and intuitive: an (say, simplicial) m-cochain
is a set of m-simplexes; an m-chain is a finite set of m-simplexes. The concept of
cochain is simpler than that of chain (one less word in the definition. . .), more
flexible and somehow more natural. We thus tend to consider cohomology as the
main concept and homology as a (useful) tool for some arguments.
Working with Z2 and its standard linear algebra is much simpler than working
with Z. For instance, the Kronecker pairing has an intuitive geometric interpreta-
tion occurring at the beginning which shows in an elementary way that cohomology
is the dual of homology. Several computations, like the homology of surfaces, are
quite easy and come early in the exposition. Also, the cohomology ring is com-
mutative. The cup square a  a  a is a linear map and may be also non-trivial
in odd degrees, leading to important invariants.
The absence of sign and orientation considerations is an enormous technical sim-
plification (even of importance in computer algorithms computing homology).
With much lighter computations and technicalities, the ideas of proofs are more
apparent.

We hope that these notes will be, for students and teachers, a complement or
companion to textbooks like those of Hatcher [82] or Munkres [155]. From our
teaching experience, starting with mod 2 (co)homology and taking advantage of
its above mentioned simplicity is a great help to grasp the ideas of the subject.
The technical difficulties of signs and orientations for finer theories, like integral
(co)homology, may then be introduced afterwards, as an adaptation of the more
intuitive mod 2 (co)homology.
Not in this book The following tools are not used in these notes.
Augmented (co)chain complexes. The reduced cohomology H (X ) is defined as
coker (H ( pt) H (X )) for the unique map X pt.
Simplicial approximation.
Spectral sequences (except in the proof of Proposition 7.2.17).
Also, we do not use advanced homotopy tools, like spectra, completions, etc.
Because of this, some prominent problems using mod 2 cohomology are only briefly
surveyed, like the work by Adams on the Hopf-invariant-one problem (p. 353), the
Sullivans conjecture (pp. 240 and 353) and the Kervaire invariant (Sect. 10.6).
Prerequisites The reader is assumed to have some familiarity with the following
subjects:
general point set topology (compactness, connectedness, etc).
elementary language of categories and functors.
simple techniques of exact sequences, like the five lemma.
1 Introduction 3

elementary facts about fundamental groups, coverings and higher homotopy

groups (not much used).
elementary techniques of smooth manifolds.

Acknowledgments A special thank is due to Volker Puppe who provided several valuable
suggestions and simplifications. Michel Zisman, Pierre de la Harpe, Samuel Tinguely and Matthias
Franz have carefully read several sections of these notes. The author is also grateful for useful com-
ments to Jim Davis, Rebecca Goldin, Andr Haefliger, Tara Holm, Allen Knutson, Jrme Scherer,
Dirk Schtz, Andras Szenes, Vladimir Turaev, Paul Turner, Claude Weber and Sad Zarati.
Chapter 2
Simplicial (Co)homology

Simplicial homology was invented by Poincar in 1899 [162] and its mod 2 version,
presented in this chapter, was introduced in 1908 by Tietze [196]. It is the simplest
homology theory to understand and, for finite complexes, it may be computed algo-
rithmically. The mod 2 version permits rapid computations on easy but non-trivial
examples, like spheres and surfaces (see Sect. 2.4).
Simplicial (co)homology is defined for a simplicial complex, but is an invariant
of the homotopy type of its geometric realization (this result will be obtained in
different ways using singular homology: see Sect. 3.6). The first section of this chapter
introduces classical techniques of (abstract) simplicial complexes. Since simplicial
homology was the only existing (co)homology theory until the 1930s, simplicial
complexes played a predominant role in algebraic topology during the first third of the
12th century (see the Introduction of Sect. 5.1). Later developments of (co)homology
theories, defined directly for topological spaces, made this combinatorial approach
less crucial. However, simplicial complexes remain an efficient way to construct
topological spaces, also largely used in computer science.

2.1 Simplicial Complexes

In this section we fix notations and recall some classical facts about (abstract) sim-
plicial complexes. For more details, see [179, Chap. 3].
A simplicial complex K consists of
a set V (K ), the set of vertices of K .
a set S(K ) of finite non-empty subsets of V (K ) which is closed under inclusion:
if S(K ) and , then S(K ). We require that {v} S(K ) for all
v V (K ).
An element of S(K ) is called a simplex of K (simplexes and simplices are
admitted as plural of simplex; we shall use simplexes, in analogy with com-
plexes). If () = m + 1, we say that is of dimension m or that is an m-simplex.
The set of m-simplexes of K is denoted by Sm (K ). The set S0 (K ) of 0-simplexes is
Springer International Publishing Switzerland 2014 5
J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_2
6 2 Simplicial (Co)homology

in bijection with V (K ), and we usually identify v V (K ) with {v} S0 (K ). We

say that K is of dimension n if Sm (K ) = for m > n, and that K is of dimen-
sion n (or n-dimensional) if it is of dimension n but not of dimension n 1.
A simplicial complex of dimension 1 is called a simplicial graph. A simplicial
complex K is called finite if V (K ) is a finite set.
If S(K ) and , we say that is a face of . As S(K ) is closed under
inclusion, it is determined by it subset Smax (K ) of maximal simplexes (if K is finite
dimensional). A subcomplex L of K is a simplicial complex such that V (L) V (K )
and S(L) S(K ). If S S(K ) we denote by S the subcomplex generated by S,
i.e. the smallest subcomplex of K such that S S( S). The m-skeleton K m of K is
the subcomplex of K generated by the union of Sk (K ) for k m.
Let S(K ). We denote by the subcomplex of K formed by and all its faces
({} in the above notation). The subcomplex of generated by the proper faces of
is called the boundary of .

2.1.1 Geometric realization. The geometric realization |K | of a simplicial complex

K is, as a set, defined by

|K | := { : V (K ) [0, 1]  vV (K ) (v) = 1 and 1 ((0, 1]) S(K )} .

We can thus see |K | as the set of probability measures on V (K ) which are supported
by the simplexes (this language is just used for comments and only in this section).
There is a distance on |K | defined by
 
d(, ) = [(v) (v)]2
vV (K )

which defines the metric topology on |K |. The set |K | with the metric topology is
denoted by |K |d . For instance, if Sm (K ), then ||
d is isometric
 to the standard
Euclidean simplex m = {(x0 , . . . , xm ) Rm+1 | xi 0 and xi = 1}.
However, a more used topology for |K | is the weak topology, for which A |K |
d is closed in ||
is closed if and only if A || d for all S(K ). The notation
|K | stands for the set |K | endowed with the weak topology. A map f from |K |
to a topological space X is then continuous if and only if its restriction to || d is
continuous for each S(K ). In particular, the identity |K | |K |d is continuous,
which implies that |K | is Hausdorff. The weak and the metric topology coincide if
and only if K is locally finite, that is each vertex is contained in a finite number
of simplexes. When K is not locally finite, |K | is not metrizable (see e.g. [179,
Theorem 3.2.8]).
When a simplicial complex K is locally finite, has countably many vertices and
is finite dimensional, it admits a Euclidean realization, i.e. an embedding of |K |
into some Euclidean space R N which is piecewise affine. A map f : |K | R N is
piecewise affine if, for each S(K ), the restriction of f to ||
is an affine map.
Thus, for each simplex , the image of || is an affine simplex of R N . If dim K n,
such a realization exists in R2n+1 (see e.g. [179, Theorem 3.3.9]).
2.1 Simplicial Complexes 7

If S(K ) then || |K |. We call || the geometric simplex associated to .

Its boundary is ||.
The space || || is the geometric open simplex associated
to . Observe that |K | is the disjoint union of its geometric open simplexes.
There is a natural injection i : V (K )  |K | sending v to the Dirac measure with
value 1 on v. We usually identify v with i(v), seeing a simplex v as a point of |K |
(a geometric vertex). In this way, a point |K | may be expressed as a convex
combination of (geometric) vertices:

= (v)v . (2.1.1)
vV (K )

2.1.2 Let K and L be simplicial complexes. Their join is the simplicial complex
K L defined by
V (L).
(1) V (K L) = V (K )
(2) S(K L) = S(K ) S(L) { | S(K ) and S(L)}.

Observe that, if Sr (K ) and Ss (L), then Sr +s+1 (K L). Also,

= and |K L| the topological join of |K | and |L| (see p. 171).

2.1.3 Stars, links, etc. Let K be a simplicial complex and S(K ). The star St()
of is the subcomplex of K generated by all the simplexes containing . The link
Lk() of is the subcomplex of K formed by the simplexes S(K ) such that
= and S(K ). Thus, Lk() is a subcomplex of St() and

St() = Lk() .

More generally, if L is a subcomplex of K , the star St(L) of L is the subcomplex

of K generated by all the simplexes containing a simplex of L. The link Lk(L) of
L is the subcomplex of K formed by the simplexes S(St(L)) S(L). One has
St(L) = L Lk(L). The open star Ost(L) of L is the open neighbourhood of |L| in
|K | defined by

Ost(L) = { |K | | (v) > 0 if v V (L)} .

This is the interior of |St(L)| in |K |.

2.1.4 Simplicial maps. Let K and L be two simplicial complexes. A simplicial map
f : K L is a map f : V (K ) V (L) such that f () S(L) if S(K ), i.e.
the image of a simplex of K is a simplex of L. Simplicial complexes and simplicial
maps form a category, the simplicial category, denoted by Simp.
A simplicial map f : K L induces a continuous map | f |: |K | |L| defined,
for w V (L), by

| f |()(w) = (v) .
v f 1 (w)
8 2 Simplicial (Co)homology

In other words, | f |() is the pushforward of the probability measure on |L|. The
geometric realization is thus a covariant functor from the simplicial category Simp
to the topological category Top of topological spaces and continuous maps.

2.1.5 Components. Let K be a simplicial complex. We define an equivalence relation

on V (K ) by saying that v v  if there exists x0 , . . . , xm V (K ) with x0 = v,
xm = v  and {xi , xi+1 } S(K ). A maximal subcomplex L of K such that V (L) is an
equivalence class is called a component of K . The set of components of K is denoted
by 0 (K ). As the vertices of a simplex are all equivalent, K is the disjoint union
of its components and 0 (K ) is in bijection with V (K )/ . The relationship with
0 (|K |), the set of (path)-components of the topological space |K |, is the following.

Lemma 2.1.6 The natural injection j: V (K ) |K | descends to a bijection j:

0 (K ) 0 (|K |).

Proof The definition of the relation makes clear that j descends to a map j :
0 (K ) 0 (|K )|. Any point of |K | is joinable by a continuous path to some vertex
j (v). Hence, j is surjective. To check the injectivity of j,
let v, v  V (K ) with

j(v) ). There exists then a continuous path c: [0, 1] |K | with c(0) = j (v)
= j(v 

and c(1) = j (v  ). Consider the open cover {Ost(w) | w V (K )} of |K |. By

compactness of [0, 1], there exists n N and vertices v0 , . . . , vn1 V (K ) such
that c([k/n, (k + 1)/n]) Ost(vk ) for all k = 0, . . . , n 1. As c(0) = j (v) and
c(1) = j (v  ), one deduces that v0 = v and vn1 = v  . For 0 < k n 1, one
has c(k/n) Ost(vk1 ) Ost(vk ). This implies that {vk1 , vk } S(K ) for all
k = 1, . . . , n 1, proving that v v  . 

A simplicial complex is called connected if it is either empty or has one component.

Note that |K | is locally path-connected for any simplicial complex K . Indeed, any
point has a neighborhood of the form |St(v)| for some vertex v, and |St(v)| path-
connected. Therefore, |K | is path-connected if and only if |K | is connected. Using
Lemma 2.1.6, this proves the following lemma.

Lemma 2.1.7 Let K be a simplicial complex. Then K is connected if and only if

|K | is a connected space.

Finally, we note the functoriality of 0 . Let f : K L be a simplicial map.

If v v  for v, v  V (K ), then f (v) f (v  ), so f descends to a map 0 f :
0 (K ) 0 (L). If f : K L and g: L M are two simplicial maps, then
0 (g f ) = 0 g 0 f . Also, 0 id K = id0 (K ) . Thus, 0 is a covariant functor from
the simplicial category Simp to the category Set of sets and maps.

2.1.8 Simplicial order. A simplicial order on a simplicial complex L is a partial

order on V (L) such that each simplex is totally ordered. For example, a total order
on V (L), as in examples where vertices are labeled by integers, is a simplicial order.
A simplicial order always exists, as a consequence of the well-ordering theorem.
2.1 Simplicial Complexes 9

2.1.9 Triangulations. A triangulation of a topological space X is a homeomorphism

h: |K | X , where K is a simplicial complex. A topological space is triangulable
if it admits a triangulation. It will be useful to have a good process to triangulate
some subspaces of Rn . A compact subspace A of Rn is a convex cell if it is the set
of solutions of families of affine equations and inequalities

f i (x) = 0, i = 1, . . . , r and g j (x) 0, j = 1, . . . , s .

A face B of A is a convex cell obtained by replacing some of the inequalities g j 0

by the equations g j = 0. The dimension of B is the dimension of the smallest affine
subspace of Rn containing B. A vertex of A is a cell of dimension 0. By induction
on the dimension, one proves that a convex cell is the convex hull of its vertices (see
e.g. [138, Theorem 5.2.2]).
A convex-cell complex P is a finite union of convex cells in Rn such that:
(i) if A is a cell of P, so are the faces of A;
(ii) the intersection of two cells of P is a common face of each of them.
The dimension of P is the maximal dimension of a cell of P. The r -skeleton P r
is the subcomplex formed by the cells of dimension r . The 0-skeleton coincides
with the set V (P) of vertices of P.
A partial order on V (P) is an affine order for P if any subset R V (P) formed
by affinely independent points is totally ordered. For instance, a total order on V (P)
is an affine order. The following lemma is a variant of [104, Lemma 1.4].

Lemma 2.1.10 Let P be a convex-cell complex. An affine order for P determines

a triangulation h : |L |
P, where L is a simplicial complex with V (L ) =
V (P). The homeomorphism h is piecewise affine and is a simplicial order
on L .

Proof The order being chosen, we drop it from the notations. For each subcom-
plex Q of P, we shall construct a simplicial complex L(Q) and a piecewise affine
homeomorphism h Q : |L(Q)| Q such that,
(i) V (L(Q)) = V (Q);
(ii) if Q  Q, then L(Q  ) L(Q) and h Q  is the restriction of h Q to |L(Q  )|.
The case Q = P will prove the lemma. The construction is by induction on the
dimension of Q, setting L(Q) = Q and h Q = id if dim Q = 0.
Suppose that L(Q) and h Q have been constructed, satisfying (i) and (ii) above, for
each subcomplex Q of P of dimension k 1. Let A be a k-cell of K with minimal
vertex a. Then A is the topological cone, with cone-vertex a, of the union B of faces
of A not containing a. The triangulation h B : |L(B)| |B| being constructed by
induction hypothesis, define L(A) to be the join L(B) {a} and h A to be the unique
piecewise affine extension of h B . Observe that, if C is a face of A, then h C is the
restriction to L(C) of h A . Therefore, this process may be used for each k-cell of P
to construct h Q : |L(Q)| Q for each subcomplex Q of P with dim Q k. 
10 2 Simplicial (Co)homology

2.1.11 Subdivisions. Let Z be a set and A be a family of subsets of Z . A simplicial

complex L such that
(a) V (L) Z ;
(b) for each S(L) there exists A A such that A;
is called a (Z , A)-simplicial complex, or a Z -simplicial complex supported by A.
Let K be a simplicial complex. Let N be a (|K |, GS(K ))-simplicial complex,
where
GS(K ) = {|| | S(K )}

is the family of geometric simplexes of K . A continuous map j: |N | |K | is

associated to N , defined by

j () = (w)w .
wV (N )

In other word, j is the piecewise affine map sending each vertex of N to to

the corresponding point of |K |. A subdivision of a simplicial complex K is a
(|K |, GS(K ))-simplicial complex N for which the associated map j: |N | |K | is
a homeomorphism (in other words, j is a triangulation of |K |).
Let N be a (|K |, GS(K ))-simplicial complex for a simplicial complex K . If L is
a subcomplex of K , then

N L = { S(N )| |L|}

is a (|L|, GS(L))-simplicial complex. Its associated map jL : |N L | |L| is the

restriction of j to |L|. The following Lemma is useful to recognize a subdivision
(compare [179, Chap. 3, Sect. 3, Theorem 4]).

Lemma 2.1.12 Let N be a (|K |, GS(K ))-simplicial complex. Then N is a subdivi-

sion of K if and only if, for each S(K ), the simplicial complex N is finite and
j : |N | | | is bijective.

Proof If N is a subdivision of K , then j is bijective since j is a homeomorphism.

Also, |N | = j 1 (| |) is compact, so N is finite.
Conversely, The fact that j is bijective for each S(K ) implies that the
continuous map j is bijective. If N is finite, then j is a continuous bijection between
compact spaces, hence a homeomorphism. This implies that the map j 1 , restricted
to each geometric simplex, is continuous. Therefore, j 1 is continuous since K is
endowed with the weak topology. 

Seeing V (K ) as a subset of |K |, we get the following corollary.

Corollary 2.1.13 Let N be a subdivision of K . Then V (K ) V (N ).

2.1 Simplicial Complexes 11

A useful systematic subdivision process is the barycentric subdivision. Let

Sm (K ) be an m-simplex of a simplicial complex K . The barycenter |K | of is
defined by

1 
= v.
m + 1 v

The barycentric subdivision K  of K is the (|K |, GS(K ))-simplicial complex where

V (K  ) = { |K | | S(K )};
{ 0 , . . . , m } Sm (K  ) whenever 0 m (i = j if i = j).
Using Lemma 2.1.12, the reader can check that K  is a subdivision of K . Observe
that the partial order defined by

(2.1.2)

is a simplicial order on K  .

2.2 Definitions of Simplicial (Co)homology

Let K be a simplicial complex. In this section, we give the definitions of the homology
H (K ) and cohomology H (K ) of K under the various and peculiar forms available
when the coefficients are in the field Z2 = {0, 1}.

Definition 2.2.1 (subset definitions)

(a) An m-cochain is a subset of Sm (K ).
(b) An m-chain is a finite subset of Sm (K ).

The set of m-cochains of K is denoted by C m (K ) and that of m-chains by Cm (K ).

By identifying Sm (K ) with the singleton {}, we see Sm (K ) as a subset of both
Cm (K ) and C m (K ). Each subset A of Sm (K ) is determined by its characteristic
function A : Sm (K ) Z2 , defined by

1 if A
A () =
0 otherwise.

This gives a bijection between subsets of Sm (K ) and functions from Sm (K ) to Z2 .

We see such a function as a colouring (0 = white and 1 = black). The following
colouring definition is equivalent to the subset definition:

Definition 2.2.2 (colouring definitions)

(a) An m-cochain is a function a: Sm (K ) Z2 .
(b) An m-chain is a function : Sm (K ) Z2 with finite support.
12 2 Simplicial (Co)homology

The colouring definition is used in low-dimensional graphical examples to draw

(co)chains in black (bold lines for 1-(co)chains).

Definition 2.2.2 endow C m (K ) and Cm (K ) with a structure of a Z2 -vector space. The

singletons provide a basis of Cm (K ), in bijection with Sm (K ). Thus,
Definition 2.2.2b is equivalent to

Definition 2.2.3 Cm (K ) is the Z2 -vector space with basis Sm (K ):


Cm (K ) = Z2 .
Sm (K )

We shall pass from one of Definitions 2.2.1, 2.2.2 or 2.2.3 to another without notice;
the context usually prevents ambiguity. We consider C (K ) = mN Cm (K ) and
C (K ) = mN C m (K ) as graded Z2 -vector spaces. The convention C1 (K ) =
C 1 (K ) = 0 is useful.
We now define the Kronecker pairing on (co)chains

,
C m (K ) Cm (K ) Z2

by the equivalent formulae

a,  = 
(a ) (mod 2) using Definition 2.2.1a and b
=  a() using Definitions 2.2.1a and 2.2.2b (2.2.1)
= Sm (K ) a()() using Definitions 2.2.2a and b.

Lemma 2.2.4 The Kronecker pairing is bilinear and the map a  a,  is an
isomorphism between C m (K ) and Cm (K ) = hom(Cm (K ), Z2 ).

Proof The bilinearity is obvious from the third line of Eq. (2.2.1). Let 0 = a
C m (K ). This means that, as a subset of Sm (K ), a is not empty. If a, then
a,  = 0, which proves the injectivity of a  a, . As for its surjectivity, let
h hom(Cm (K ), Z2 ). Using the inclusion Sm (K )  Cm (K ) given by  { },
define

a = { Sm (K ) | h( ) = 1} .

For each Sm (K ) the equation h() = a,  holds true. As Sm (K ) is a basis of

Cm (K ), this implies that h = a, . 
2.2 Definitions of Simplicial (Co)homology 13

We now define the boundary and coboundary operators. The boundary operator
: Cm (K ) Cm1 (K ) is the Z2 -linear map defined by

() = {(m 1)-faces of } = Sm1 (),

Sm (K ). (2.2.2)

Formula (2.2.2) is written in the language of Definition 2.2.1b. Using Definition

2.2.3, we get

() = . (2.2.3)
Sm1 ()

The coboundary operator : C m (K ) C m+1 (K ) is defined by the equation

a,  = a,  . (2.2.4)

The last equation indeed defines by Lemma 2.2.4 and may be seen as the Kronecker
adjoint of . In particular, if Sm (K ) and Sm1 (K ) then

() ( ) . (2.2.5)

The first equivalence determines the operator since Sm (K ) is a basis for Cm (K ).

The second equivalence determines if Sm1 (K ) is finite. Note that the definition
of may also be given as follows: if a C m (K ), then

(a) = { Sm+1 (K ) |  (a ( )) is odd} .

Let Sm (K ). Each Sm2 (K ) with belongs to the boundary of

exactly two (m 1)-simplexes of . Using Eq. (2.2.3), this implies that = 0.
By Eq. (2.2.4) and Lemma 2.2.4, we get = 0. We define the Z2 -vector spaces
Z m (K ) = ker( : Cm (K ) Cm1 (K )), the m-cycles of K .
Bm (K ) = image ( : Cm+1 (K ) Cm (K )), the m-boundaries of K .
Z m (K ) = ker( : C m (K ) C m+1 (K )), the m-cocycles of K .
B m (K ) = image ( : C m1 (K ) C m (K )), the m-coboundaries of K .
For example, Fig. 2.1 shows a triangulation K of the plane, with V (K ) = Z Z.
The bold line is a cocycle a which is a coboundary: a = B, with B = {{(m, n)} |
(m, n) V (K ) and m 0}, drawn in bold dots.
Since = 0 and = 0, one has Bm (K ) Z m (K ) and B m (K ) Z m (K ).
We form the quotient vector spaces
Hm (K ) = Z m (K )/Bm (K ), the mth -homology vector space of K .
H m (K ) = Z m (K )/B m (K ), the mth -cohomology vector space of K .
As for the (co)chains, the notations H (K ) = mN Hm (K ) and H (K ) =
mN H m (K ) stand for the (co)homology seen as graded Z2 -vector spaces. By
14 2 Simplicial (Co)homology

Fig. 2.1 A triangulation K

of the plan, with
V (K ) = Z Z

convention, H1 (K ) = H 1 (K ) = 0. Also, the homology and the cohomology

are in duality via the Kronecker pairing:
Proposition 2.2.5 (Kronecker duality) The Kronecker pairing on (co)chains induces
a bilinear map

,
H m (K ) Hm (K ) Z2 .

Moreover, the correspondence a  a,  is an isomorphism

k
H m (K ) hom(Hm (K ), Z2 ) .

Proof Instead of giving a direct proof, which the reader may do as an exercise, we
will take advantage of the more general setting of Kronecker pairs, developed in the
next section. In this way, Proposition 2.2.5 follows from Proposition 2.3.5. 

2.3 Kronecker Pairs

All the vector spaces in this section are over an arbitrary fixed field F. The dual of a
vector space V is denoted by V  .
A chain complex is a pair (C , ), where

C is a graded vector space C = mN Cm . We add the convention that C1 = 0.
: C C is a linear map of degree 1, i.e. (Cm ) Cm1 , satisfying
= 0. The operator is called the boundary of the chain complex.
A cochain complex is a pair (C , ), where

C is a graded vector space C = mN C m . We add the convention that C 1 = 0.
: C C is a linear map of degree +1, i.e. (C m ) C m+1 , satisfying
= 0. The operator is called the coboundary of the cochain complex.
2.3 Kronecker Pairs 15

A Kronecker pair consists of three items:

(a) a chain complex (C , ).
(b) a cochain complex (C , ).
(c) a bilinear map
,
C m Cm F

satisfying the equation

a,  = a,  . (2.3.1)

for all a C m and Cm+1 and all m N. Moreover, we require that the map

k: C m Cm , given by k(a) = a, , is an isomorphism.

Example 2.3.1 Let K be a simplicial complex. Its simplicial (co)chain complexes

(C (K ), ), (C (K ), ), together with the pairing  ,  of Sect. 2.2 is a Kronecker
pair, with F = Z2 , as seen in Lemma 2.2.4 and Eq. (2.2.4).

Example 2.3.2 Let (C , ) be a chain complex. One can define a cochain complex

(C , ) by C m = Cm and =  and then get a bilinear map (pairing) , by the
evaluation: a,  = a(). These constitute a Kronecker pair. Actually, via the map
k, any Kronecker pair is isomorphic to this one. The reader may use this fact to
produce alternative proofs of the results of this section.

We first observe that, as the Kronecker pairing is non-degenerate, chains and

cochains mutually determine each other:

Lemma 2.3.3 Let (C , ), (C , ),  ,  be a Kronecker pair.

(a) Let a, a  C m . Suppose that a,  = a  ,  for all Cm . Then a = a  .
(b) Let ,  Cm . Suppose that a,  = a,   for all a C m . Then =  .
(c) Let Sm be a basis for Cm and let f : Sm F be a map. Then, there is a unique
a C m such that a,  = f () for all Sm .

Proof In Point (a), the hypotheses imply that k(a) = k(a  ). As k is injective, this
shows that a = a  .
In Point (b), suppose that =  . Let A (Cm ) such that A(  ) = 0. Then,
a,  = a,   for a = k1 (A) C m .


Finally, the condition a() = f () for all Sm defines a unique a Cm and
1
a = k (a). 

As is Sect. 2.2, we consider the Z2 -vector spaces

Zm = ker( : Cm Cm1 ), the m-cycles (of C ).
Bm = image ( : Cm+1 Cm ), the m-boundaries.
Zm = ker( : C m C m+1 ), the m-cocycles.
Bm = image ( : C m1 C m ), the m-coboundaries.
16 2 Simplicial (Co)homology

Since = 0 and = 0, one has Bm Z m and B m Z m . We form the quotient

vector spaces
Hm = Z m /Bm , the mth-homology group (or vector space).
H m = Z m /B m , the mth-cohomology group (or vector space).
We consider the (co)homology as graded vector spaces: H = mN Hm and H =
mN H m .
The cocycles and coboundaries may be detected by the pairing:

Lemma 2.3.4 Let a C m . Then

(i) a Z m if and only if a, Bm  = 0.
(ii) a B m if and only if a, Z m  = 0.

Proof Point (i) directly follows from Eq. (2.3.1) and the fact that k is injective. Also,
if a B m , Eq. (2.3.1) implies that a, Z m  = 0. It remains to prove the converse
(this is the only place in this lemma where we need vector spaces over a field instead
just module over a ring). We consider the exact sequence

0 Z m Cm Bm1
0. (2.3.2)


Let a C m such that a, Z m  = 0. By (2.3.2), there exists a1 Bm1 such that
a,  = a1 . As we are dealing with vector spaces, Bm1 is a direct summand

of Cm1 . We can thus extend a1 to a2 Cm1 . As k is surjective, there exists
a3 C m1 such that a3 ,  = a2 . For all Cm , one then has

a3 ,  = a3 ,  = a2 () = a1 () = a,  .

As k is injective this implies that a = a3 B m . 

Let us restrict the pairing  ,  to Z m Z m . Formula (2.3.1) implies that

Z m , Bm  = B m , Z m  = 0 .

,
Hence, the pairing descends to a bilinear map H m Hm F, giving rise to a

linear map k: H m Hm , called the Kronecker pairing on (co)homology. We see

H and H as (co)chain complexes by setting = 0 and = 0.

Proposition 2.3.5 (H , H ,  , ) is a Kronecker pair.

Proof Equation (2.3.1) holds trivially since and both vanish. It remains to show

that k: H m Hm is bijective.

Let a0 Hm . Pre-composing a0 with the projection Z m Hm produces a1
 
Z m . As Z m is a direct summand in Cm , one can extend a1 to a2 Cm . Since
(C , C ,  , ) is a Kronecker pair, there exists a C m such that a,  = a2 . The
2.3 Kronecker Pairs 17

cochain a satisfies a, Bm  = a2 (Bm ) = 0 which, by Lemma 2.3.4, implies that

a Z m . The cohomology class [a] H m of a then satisfies [a],  = a0 . Thus, k
is surjective.
For the injectivity of k, let b H m with b, Hm  = 0. Represent b by b Z m ,
which then satisfies b, Z m  = 0. By Lemma 2.3.4, b B m and thus b = 0. 
be two chain complexes. A map : C C is a
Let (C , ) and (C , )
morphism of chain complexes or a chain map if it is linear map of degree 0 (i.e.
(Cm ) C m ) such that = . This implies that (Z m ) Z m and (Bm )
B m . Hence, induces a linear map H : Hm H m for all m.
be two cochain complexes. A linear map
In the same way, let (C , ) and (C , )
: C C of degree 0 is a morphism of cochain complexes or a cochain map if
= . Hence, induces a linear map H : H m H m for all m.
C , ,
Let P = (C , , C , ,  , ) and P = (C , ,  ,  ) be two Kronecker
pairs. A morphism of Kronecker pairs, from P to P, consists of a pair (, ) where
: C C is a morphism of chain complexes and : C C is a morphism of
cochain complexes such that

a, () = (a),  . (2.3.3)

Eq. (2.3.3) is equivalent to the commutativity of

Using the isomorphisms k and k,
the diagram

C / C

k k . (2.3.4)
 

C
 / C

Lemma 2.3.6 Let P and P be Kronecker pairs as above. Let : C C be a

morphism of chain complex. Define : C C by Eq. (2.3.3) (or Diagram (2.3.4)).
Then the pair (, ) is a morphism of Kronecker pairs.

Proof Obviously, is a linear map of degree 0 and Eq. (2.3.3) is satisfied. It remains
to show that is a morphism of cochain-complexes. But, if b Cm ( K ) and
Cm+1 (K ), one has

(b),  = (b),  = b, () = b, ()

() = (b),
= b,  ,

which proves that (b) = (b). 
18 2 Simplicial (Co)homology

A morphism (, ) of Kronecker pairs determines a morphism of Kronecker pairs

(H , H ) from (H , H ,  , ) to ( H , H ,  ,  ). This process is functorial:

Lemma 2.3.7 Let (1 , 1 ) be a morphism of Kronecker pairs from P to P and let

(2 , 2 ) be a morphism of Kronecker pairs from P to P.
Then

(H 2 H 1 , H 1 H 2 ) = (H (2 1 ), H (2 1 ))

Proof That H 2 H 1 = H (2 1 ) is a tautology. For the cohomology equality,

we use that

H 1 H 2 (a),  = H 2 (a), H 1 () = a, H 2 H 1 ()

= a, H (2 1 )() = H (2 1 ))(a), 

holds for all a H and all H . 

We finish this section with some technical results which will be used later.

Lemma 2.3.8 Let f : U V and g: V W be two linear maps between vector

spaces. Then, the sequence

f g
U
V
W (2.3.5)

is exact at V if and only if the sequence

f g
U  V 
W (2.3.6)

is exact at V  .

Proof As f  g  = (g f ) , then f  g  = 0 if and only if g f = 0.

On the other hand, suppose that ker g image f . We shall prove that ker f 
image g  . Indeed, let a ker f  . Then, a(image f ) = 0 and, using the inclusion
ker g image f , we deduce that a(ker g) = 0. Therefore, a descends to a linear
map a: V / ker g F. The quotient space V / ker g injects into W , so there exists
b W  such that a = b g = g  (b), proving that a image g  .
Finally, suppose that ker g  image f . Then there exists a V  such that
a(image f ) = 0, i.e., a ker f  , and a(ker g) = 0, i.e. a / image g  . This proves

that ker f  image g .  

Lemma 2.3.9 Let (, ) be a morphism of Kronecker pairs from P = (C , , C ,

C , ,
,  , ) to P = (C , ,  ,  ). Then the pairings  ,  and  ,  induce bilinear
maps

, ,
coker ker F and ker coker F
2.3 Kronecker Pairs 19

such that the induced linear maps

k k
(ker ) and ker
coker (coker )

are isomorphisms.

Proof Equation (2.3.3) implies that (C ), ker  = 0 and ker , (C ) = 0,

whence the induced pairings. Consider the exact sequence

0 / ker / C / C / coker / 0.

By Lemma 2.3.8, passing to the dual preserves exactness. Using Diagram (2.3.4),
one gets a commutative diagram


0 o (ker ) o C o

C o

(coker ) o 0
O O O O
k k k k . (2.3.7)

0 o coker o H o C o ker o 0

By diagram-chasing, the two extreme up-arrows are bijective (one can also invoke
the famous five-lemma: see e.g. [179, Chap. 4, Sect. 5, Lemma 11]). 

Corollary 2.3.10 Let (, ) be a morphism of Kronecker pairs from (C , , C , ,

C , ,
 , ) to (C , ,  ,  ). Then the pairings  ,  and  ,  on (co)homology induce
bilinear maps

, ,
coker H ker H F and ker H coker H F

such that the induced linear maps

k k
coker H
(ker H ) and ker H
(coker H )

are isomorphisms.

Proof The morphism (, ) induces a morphism of Kronecker pairs (H , H )

from (H , H ,  , ) to ( H , H ,  ,  ). Corollary 2.3.10 follows then from
Lemma 2.3.9 applied to (H , H ). 

Corollary 2.3.10 implies the following

Corollary 2.3.11 Let (, ) be a morphism of Kronecker pairs from (C , , C ,

C , ,
,  , ) to (C , ,  ,  ). Then
20 2 Simplicial (Co)homology

(a) H is surjective if and only if H is injective.

(b) H is injective if and only if H is surjective.
(c) H is bijective if and only if H is bijective.

2.4 First Computations

2.4.1 Reduction to Components

Let K be a simplicial complex. We have seen in 2.1.5 that K is the disjoint

 union of its
components, whose set is denoted by 0 (K ). Therefore, Sm (K ) = L0 (K ) Sm (L)
which, by Definition 2.2.3, gives a canonical isomorphism

Cm (L) Cm (K ) .
L0 (K )

This direct sum decomposition commutes with the boundary operators, giving a
canonical isomorphism

H (L) H (K ) . (2.4.1)
L0 (K )

As for the cohomology, seeing an m-cochain as a map : Sm (K ) Z2

(Definition 2.2.2) the restrictions of to Sm (L) for all L 0 (K ) gives an
isomorphism


C m (K )
C m (L)
L0 (K )

commuting with the coboundary operators. This gives an isomorphism


H (K )
H (L) . (2.4.2)
L0 (K )

The isomorphisms of (2.4.1) and (2.4.2) permit us to reduce (co)homology com-

putations to connected simplicial complexes. They are of course compatible with the
Kronecker duality (Proposition 2.2.5). A formulation of these isomorphisms using
simplicial maps is given in Proposition 2.5.3.

2.4.2 0-Dimensional (Co)homology

Let K be a simplicial complex. The unit cochain 1 C 0 (K ) is defined by 1 = S0 (K ),

using the subset definition. In the language of colouring, one has 1(v) = 1 for all
2.4 First Computations 21

v V (K ) = S0 (K ), that is all vertices are black. If = {v, w} S1 (K ), then

1,  = 1,  = 1(v) + 1(w) = 0 ,

which proves that (1) = 0 by Lemma 2.2.4. Hence, 1 is a cocycle, whose cohomol-
ogy class is again denoted by 1 H 0 (K ).

Proposition 2.4.1 Let K be a non-empty connected simplicial complex. Then,

(i) H 0 (K ) = Z2 , generated by 1 which is the only non-vanishing 0-cocycle.
(ii) H0 (K ) = Z2 . Any 0-chain is a cycle, which represents the non-zero element
of H0 (K ) if and only if  is odd.

Proof If K is non-empty the unit cochain does not vanish. As C 1 (K ) = 0, this

implies that 1 = 0 in H 0 (K ).
Let a C 0 (K ) with a = 0, 1. Then there exists v, v  V (K ) with a(v) = a(v  ).
Since K is connected, there exists x0 , . . . , xm V (K ) with x0 = v, xm = v  and
{xi , xi+1 } S(K ). Therefore, there exists 0 k < m with a(xk ) = a(xk+1 ). This
implies that {xk , xk+1 } a, proving that a = 0. We have thus proved (i).
Now, H0 (K ) = Z2 since H 0 (K ) H0 (K ) . Any C0 (K ) is a cycle since
C1 (K ) = 0. It represents the non-zero homology class if and only if 1,  = 1,
that is if and only if  is odd. 
(K )
Corollary 2.4.2 Let K be a simplicial complex. Then H 0 (K ) Z2 0 .
(K )
Here, Z2 0 denotes the set of maps from 0 (K ) to Z2 . The isomorphism of
Corollary 2.4.2 is natural for simplicial maps (see Corollary 2.5.6).

Proof By Proposition 2.4.1 and its proof, H 0 (K ) = Z 0 (K ) is the set of maps from
V (K ) to Z2 which are constant on each component. Such a map is determined by a
map from 0 (K ) to Z2 and conversely. 

2.4.3 Pseudomanifolds

An n-dimensional pseudomanifold is a simplicial complex M such that

(a) every simplex of M is contained in an n-simplex of M.
(b) every (n 1)-simplex of M is a face of exactly two n-simplexes of M.
(c) for any ,  Sn (M), there exists a sequence = 0 , . . . , m =  of n-
simplexes such that i and i+1 have an (n 1)-face in common for i 1 < m.

Example 2.4.3 (1) Let m be an integer with m 3. The polygon Pm is the 1-

dimensional pseudomanifold for which V (Pm ) = {0, 1, . . . , m 1} = Z/mZ and
S1 (Pm ) = {{k, k + 1} | k V (Pm )}. It can be visualized in the complex plane as
the equilateral m-gon whose vertices are the mth roots of the unity.
22 2 Simplicial (Co)homology

(2) Consider the triangulation of S 2 given by an icosahedron. Choose one pair of

antipodal vertices and identify them in a single point. This gives a quotient simplicial
complex K which is a 2-dimensional pseudomanifold. Observe that |K | is not a
topological manifold.

Pseudomanifolds have been introduced in 1911 by Brouwer [22, p. 477], for his
work on the degree and on the invariance of the dimension. They are also called
n-circuits in the literature. Proposition 2.4.4 below and its proof, together with
Proposition 2.4.1, shows that n-dimensional pseudomanifolds satisfy Poincar dual-
ity in dimensions 0 and n.
Let M be a finite n-dimensional pseudomanifold. The n-chain [M] = Sn (M)
Cn (M) is called the fundamental cycle of M (it is a cycle by Point (b) of the above def-
inition). Its homology class, also denoted by [M] Hn (M) is called the fundamental
class of M.

Proposition 2.4.4 Let M be a finite non-empty n-dimensional pseudomanifold.

Then,
(i) Hn (M) = Z2 , generated by [M] which is the only non-vanishing n-cycle.
(ii) H n (M) = Z2 . Any n-cochain a is a cocycle, and [a] = 0 in H n (M) if and only
if a is odd.

Proof We define a simplicial graph L with V (L) = Sn (M) by setting {,  }

S1 (L) if and only if and  have an (n 1)-face in common. The identification
Sn (M) = V (L) produces isomorphisms

Fn : Cn (M)
C 0 (L) and F n : C n (M)
C0 (L) . (2.4.3)

(As M is finite, so is L and C (L) is equal to C (L), using Definition 2.2.2) On the
other hand, by Point (b) of the definition of a pseudomanifold, one gets a bijection

Sn1 (M)
F: S1 (L). It gives rise to isomorphisms

Fn1 : Cn1 (M)
C 1 (L) and F n1 : C n1 (M)
C1 (L) . (2.4.4)

The isomorphisms of (2.4.3) and (2.4.4) satisfy

Fn1 = Fn and F n1 = F n .

Since Cn+1 (M) = 0 by Point (a) of the definition of a pseudomanifold, the above
isomorphisms give rise to isomorphisms

H 0 (L) and F : H n (M)
F : Hn (M) H0 (L)

with F ([M]) = 1. By Point (c) of the definition of a pseudomanifold, the graph L

is connected. Therefore, Proposition 2.4.4 follows from Proposition 2.4.1. 
2.4 First Computations 23

The proof of Proposition 2.4.4 actually gives the following result.

Proposition 2.4.5 Let M be a finite non-empty simplicial complex satisfying Con-

ditions (a) and (b) of the definition of an n-dimensional pseudomanifold. Then, M is
a pseudomanifold if and only if Hn (M) = Z2 .

2.4.4 Poincar Series and Polynomials


A graded Z2 -vector space A = iN Ai is of finite type if Ai is finite dimensional
for all i N. In this case, the Poincar series of A is the formal power series
defined by

Pt (A ) = dim Ai t i N[[t]].
iN

When dim A < , the series Pt (A ) is a polynomial, also called the Poincar
polynomial of A .
A simplicial complex K is of finite (co)homology type if H (K ) (or, equivalently,
H (K )) is of finite type. In this case, the Poincar series of K is that of H (K ).
The (co)homology of a simplicial complex of finite (co)homology type is, up to
isomorphism, determined by its Poincar series, which is often the shortest way to
describe it. The number dim Hm (K ) is called the m-th Betti number of K . The vector
space C (K ) is endowed with the basis S(K ) for which the matrix of the boundary
operator is given explicitly. Thus, the Betti numbers may be effectively computed by
standard algorithms of linear algebra.

2.4.5 (Co)homology of a Cone

The simplest non-empty simplicial complex is a point whose (co)homology is

obviously

0 if m > 0
H ( pt) Hm ( pt)
m
(2.4.5)
Z2 if m = 0 .

In terms of Poincar polynomial: Pt ( pt) = 1. Let L be a simplicial complex. The

cone on L is the simplicial complex C L defined by V (C L) = V (L) {} and

Sm (C L) = Sm (L) { {} | Sm1 (L)} .

Note that C L is the join C L L {}.

24 2 Simplicial (Co)homology

Proposition 2.4.6 The cone C L on a simplicial complex L has its (co)homology

isomorphic to that of a point. In other words, Pt (C L) = 1.

Proof By Kronecker duality, it is enough to prove the result on homology. The cone
C L is obviously connected and non-empty (it contains ), so H0 (C L) = Z2 .
Define a linear map D: Cm (C L) Cm+1 (C L) by setting, for Sm (C L):

{} if
/
D() =
0 if .

Hence, D D = 0. If
/ , the formula

D() = D() + (2.4.6)

holds true in Cm (C L) (and has a clear geometrical interpretation). Suppose that

and dim 1. Then = D( ) with = {}. Using Formula (2.4.6)
and that D D = 0, one has

D() + = D( D( )) + = D(D( ) + ) + D( ) = 0 .

Therefore, Formula (2.4.6) holds also true if , provided dim 1. This

proves that

D() = D() + for all Cm (C L) with m 1 . (2.4.7)

Now, if Cm (C L) satisfies = 0, Formula (2.4.7) implies that = D(),

which proves that Hm (C L) = 0 if m 1. 

As an application of Proposition 2.4.6, let A be a set. The full complex F A on A

is the simplicial complex for which V (F A) = A and S(F A) is the family of all
finite non-empty subsets of A. If A is finite and non-empty, then F A is isomorphic
to a simplex of dimension A 1. Denote by F A the subcomplex of F A generated
by the proper (i.e. = A) subsets of A. For instance, F A = F A if A is infinite.

Corollary 2.4.7 Let A be a non-empty set. Then

(i) F A has its (co)homology isomorphic to that of a point, i.e. Pt (F A) = 1.
(ii) If 3 A , then Pt (F A) = 1 + t A1 .
(iii) If A = 2, then Pt (F A) = 2.

Proof As A is not empty, F A is isomorphic to the cone over F A deprived of one

of its elements. Point (i) then follows from Proposition 2.4.6. Let n = A 1. The
chain complex of F A looks like a sequence

n n1
0 Cn (F A)
Cn1 (F A) C0 (F A) 0 ,
2.4 First Computations 25

which, by (i), is exact except at C0 (F A). One has Cn (F A) = Z2 , generated by the

A Sn (F A). Hence, ker n1 Z2 . As the chain complex C (F A) is the same as
that of F A with Cn replaced by 0, this proves (ii). If A = 2, then F A consists of
two 0-simplexes and Point (iii) follows from (2.4.5) to (2.4.1). 

2.4.6 The Euler Characteristic

Let K be a finite simplicial complex. Its Euler characteristic (K ) is defined as

(K ) = (1)m  Sm (K ) Z .
mN

Proposition 2.4.8 Let K be a finite simplicial complex. Then

 
(K ) = (1)m dim Hm (K ) = (1)m dim H m (K ) .
mN m

As in the definition of the Poincar polynomial, the number dim Hm (K ) is the

dimension of Hm (K ) as a Z2 -vector space. In other words, dim Hm (K ) is the m-th
Betti number of K . Proposition 2.4.8 holds true for the (co)homology with coeffi-
cients in any field F, though the Betti numbers depend individually on F.
Proof By Kronecker duality, only the first equality requires a proof. Let cm , z m , bm
and h m be the dimensions of Cm (K ), Z m (K ), Bm (K ) and Hm (K ). Elementary linear
algebra gives the equalities

cm = z m + bm1
z m = bm + h m .

We deduce that
   
(K ) = (1)m cm = (1)m h m + (1)m bm + (1)m bm1 .
mN mN mN mN

As b1 = 0, the last two sums cancels each other, proving Proposition 2.4.8. 
Corollary 2.4.9 Let K be a finite simplicial complex. Then

(K ) = Pt (K )t=1 .

The following additive formula for the Euler characteristic is useful.

Lemma 2.4.10 Let K be a simplicial complex. Let K 1 and K 2 be two subcomplexes
of K such that K = K 1 K 2 . Then,

(K ) = (K 1 ) + (K 2 ) (K 1 K 2 ) .
26 2 Simplicial (Co)homology

Proof The formula follows directly from the equations Sm (K ) = Sm (K 1 )Sm (K 2 )

and Sm (K 1 K 2 ) = Sm (K 1 ) Sm (K 2 ). 

2.4.7 Surfaces

A surface is a manifold of dimension 2. In this section, we give examples of trian-

gulations of surfaces and compute their (co)homology. Strictly speaking, the results
would hold only for the given triangulations, but we allow us to formulate them in
more general terms. For this, we somehow admit that
a connected surface is a pseudomanifold of dimension 2. This will be established
rigorously in Corollary 5.2.7 but the reader may find a proof as an exercise and
this is easy to check for the particular triangulations given below.
up to isomorphism, the (co)homology of a simplicial complex K depends only of
the homotopy type of |K |. This will be proved in Sect. 3.6. In particular, the Euler
characteristic of two triangulations of a surface coincide.

The 2-Sphere

The 2-sphere S 2 being homeomorphic to the boundary of a 3-simplex, it follows

from Corollary 2.4.7 that:

Pt (S 2 ) = 1 + t 2 .

The Projective Plane

The projective plane RP 2 is the quotient of S 2 by the antipodal map. The triangulation
of S 2 as a regular icosahedron being invariant under the antipodal map, it gives a
triangulation of RP 2 given in Fig. 2.2. Note that the border edges appear twice,

Fig. 2.2 A triangulation 1

of R P 2

a
2 5
3 4

0
4 3

5 2
2.4 First Computations 27

showing as expected that RP 2 is the quotient of a 2-disk modulo the antipodal

involution on its boundary.
Being a quotient of an icosahedron, the triangulation of Fig. 2.2 has 6 vertices, 15
edges and 10 facets, thus (RP 2 ) = 1. Using that RP 2 is a connected 2-dimensional
pseudomanifold, we deduce that

Pt (RP 2 ) = 1 + t + t 2 . (2.4.8)

To identify the generators of H 1 (RP 2 ) Z2 and H1 (RP 2 ), we define

 
a = = {1, 2}, {2, 3}, {3, 4}, {4, 5}, {5, 1} S1 (RP 2 ) . (2.4.9)

We see a C 1 (RP 2 ) and C1 (RP 2 ). The cochain a is drawn in bold on Fig. 2.2,
where it looks as the set of border edges, since each of its edges appears twice on the
figure. It is easy to check that (a) = 0 and () = 0. As  = 5 is odd, one has
a,  = 1, showing that a is the generator of H 1 (RP 2 ) = Z2 and is the generator
of H1 (RP 2 ) = Z2 .

The 2-Torus

The 2-torus T 2 = S 1 S 1 is the quotient of a square whose opposite sides are identi-
fied. A triangulation of T 2 is described (in two copies) in Fig. 2.3. This triangulation
has 9 vertices, 27 edges and 18 facets, which implies that (T 2 ) = 0. Since T 2 is a
connected 2-dimensional pseudomanifold, we deduce that

Pt (T 2 ) = (1 + t)2 .

In Fig. 2.3 are drawn two chains , C1 (T 2 ) given by

   
= {3, 8}, {8, 9}, {9, 3} and = {5, 7}, {7, 9}, {9, 5} .

1 2 3 1 1 2 3 1

6 8 6 8
4 4 4 4
a
7 9 7 9
5 5 5 5


1 2 3 1 1 2 3 1
b

Fig. 2.3 Two copies of a triangulation of the 2-torus T 2 , showing generators of H 1 (T 2 ) and H1 (T 2 )
28 2 Simplicial (Co)homology

We also drew two cochains a, b C 1 (T 2 ) defined as

 
a = {4, 5}, {5, 6}, {6, 7}, {7, 8}, {8, 9}, {9, 4}

and
 
b = {2, 3}, {3, 6}, {6, 8}, {8, 7}, {7, 9}, {9, 2} .

One checks that = = 0 and that a = b = 0. Therefore, they represent

classes a, b H 1 (T 2 ) and , H1 (T 2 ). The equalities

a,  = 1 , a,  = 0 , b,  = 0 , b,  = 1

imply that a, b is a basis of H 1 (T 2 ) and , is a basis of H1 (T 2 ).

If we consider a and b as 1-chains (call them a and b), we also have a = b = 0.
Note that

= 1 , a, a
a, b = 0 , b, a
= 0 , b, b =1

This proves that a = and b = in H1 (T 2 ).

The Klein Bottle

A triangulation of the Klein bottle K is pictured in Fig. 2.4. As the 2-torus, the
Klein bottle is the quotient of a square with opposite side identified, one of these
identifications reversing the orientation. One checks that (K ) = 0. Since K is
a connected 2-dimensional pseudomanifold, the (co)homology of K is abstractly
isomorphic to that of T 2 :

Pt (K ) = (1 + t)2

(In Chap. 3, H (T 2 ) and H (K ) will be distinguished by their cup product: see

p. 138). In Fig. 2.4 the dotted lines show two 1-chains , C1 (K ) given by
   
= {3, 8}, {8, 9}, {9, 3} and = {5, 7}, {7, 9}, {9, 5} . (2.4.10)

The bold lines describe two 1-cochains a, b C 1 (K ) defined as

 
a = {4, 5}, {5, 6}, {6, 7}, {7, 8}, {8, 9}, {9, 5} (2.4.11)

and
 
b = {2, 3}, {3, 6}, {6, 8}, {8, 7}, {7, 9}, {9, 2} . (2.4.12)
2.4 First Computations 29

1 2 3 1 1 2 3 1

6 8 6 8
4 5 4 5
a
7 9 7 9
5 4 5 4


1 2 3 1 1 2 3 1
b
Fig. 2.4 Two copies of a triangulation of the Klein bottle K , showing generators of H 1 (K ) and
H1 (K )

One checks that = = 0 and that a = b = 0. Therefore, they represent

classes a, b H 1 (K ) and , H1 (K ). The equalities

a,  = 1 , a,  = 1 , b,  = 0 , b,  = 1

imply that a, b is a basis of H 1 (K ) and , is a basis of H1 (K ).

Here
As in the case of T 2 , we may regard a and b as 1-chains (call them a and b).

b = 0 but a = {4} + {5} = 0.

Other Surfaces

Let K 1 and K 2 be two simplicial complexes such that |K 1 | and |K 2 | are surfaces. A
simplicial complex L with |L| homeomorphic to the connected sum |K 1 ||K 2 | may
be obtained in the following way: choose 2-simplexes 1 K 1 and 2 K 2 . Let
L i = K i i and let L be obtained by taking the disjoint union of L 1 and L 2 and
identifying 1 with 2 . Thus, L = L 1 L 2 and L 0 = L 1 L 2 is isomorphic to the
boundary of a 2-simplex.
By Lemma 2.4.10, one has

(L) = (L 1 ) + (L 2 ) (L 0 )
= (K 1 ) 1 + (K 2 ) 1 0
= (K 1 ) + (K 2 ) 2 . (2.4.13)

The orientable surface g of genus g is defined as the connected sum of g copies

of the torus T 2 . By Formula (2.4.13), one has

(g ) = 2 2g . (2.4.14)
30 2 Simplicial (Co)homology

As g is a 2-dimensional connected pseudomanifold, one has

Pt (g ) = 1 + 2gt + t 2 .

The nonorientable surface  g of genus g is defined as the connected sum of g

1 = RP 2 and 
copies of RP 2 . For instance,  2 is the Klein bottle. Formula (2.4.13)
implies

g) = 2 g .
( (2.4.15)

g is a 2-dimensional connected pseudomanifold, one has

As 

g ) = 1 + gt + t 2 .
Pt (

2.5 The Homomorphism Induced by a Simplicial Map

Let f : K L be a simplicial map between the simplicial complexes K and L. Recall

that f is given by a map f : V (K ) V (L) such that f () S(L) if S(K ), i.e.
the image of an m-simplex of K is an n-simplex of L with n m. We define C f :
C (K ) C (L) as the degree 0 linear map such that, for all Sm (K ), one has

f () if f () Sm (L) (i.e. if f | is injective)
C f () = (2.5.1)
0 otherwise.

We also define C f : C (L) C (K ) by setting, for a C m (L),

 
C f (a) = Sm (K ) | f () a . (2.5.2)

In the following lemma, we use the same notation for the (co)boundary operators
and and the Kronecker product  , , both for K of for L.
Lemma 2.5.1 Let f : K L be a simplicial map. Then
(a) C f = C f .
(b) C f = C f .
(c) C f (b),  = b, C f () for all b C (L) and all C (K ).
In other words, the couple (C f, C f ) is a morphism of Kronecker pairs.
Proof To prove (a), let Sm (K ). If f restricted to is injective, it is straightfor-
ward that C f () = C f (). Otherwise, we have to show that C f () = 0.
Let us label the vertices v0 , v1 , . . . , vm of in such a way that f (v0 ) = f (v1 ).
Then, C f () is a sum of two terms: C f () = C f (0 ) + C f (1 ), where
0 = {v1 , v2 , . . . , vm } and 1 = {v0 , v2 , . . . , vm }. As C f (0 ) = C f (1 ), one has
2.5 The Homomorphism Induced by a Simplicial Map 31

C f () = 0. Thus, Point (a) is established. Point (c) can be easily deduced from
Definitions (2.5.1) and (2.5.2), taking for a simplex of K . Point (b) then follows
from Points (a) and (c), using Lemma 2.3.6 and its proof. 
By Lemma 2.5.1 and Proposition 2.3.5, the couple (C f, C f ) determines linear
maps of degree zero

H f : H (K ) H (L) and H f : H (L) H (K )

such that

H f (a),  = a, H f () for all a H (L) and H (K ) . (2.5.3)

Lemma 2.5.2 (Functoriality) Let f : ZK L and g: L M be simplicial maps.

Then H (g f ) = H g H f and H (g f ) = H f H g. Also H id K = id H (K )
and H id K = id H (K )
In other words, H and H are functors from the simplicial category Simp to the
category GrV of graded vector spaces and degree 0 linear maps. The cohomology
is contravariant and the homology is covariant.
Proof For S(K ), the formula C (g f )() = C g C f () follows directly
from Definition (2.5.1). Therefore C (g f ) = C g C f and then H (g f ) =
H g H f . The corresponding formulae for cochains and cohomology follow from
Point (c) of Lemma 2.5.1. The formulae for id K is obvious. 
Simplicial maps and components. Let K be a simplicial complex. For each
component L 0 (K ) of K , the inclusion i L : L K is a simplicial map. The
results of Sect. 2.4.1 may be strengthened as follows.
Proposition 2.5.3 Let K be a simplicial complex. The family of simplicial maps i L :
L K for L 0 (K ) gives rise to isomorphisms

(H i L ) 
H (K ) / L0 (K ) H (L)

and

 H i L
L0 (K ) H (L)
/ H (K ) .

The homomorphisms H 0 f and H0 f . We use the same notation 1 H 0 (K ) and

1 H 0 (L) for the classes given by the unit cochains.
Lemma 2.5.4 Let f : K L be a simplicial map. Then H 0 f (1) = 1.
Proof The formula C 0 f (1) = 1 in C 0 (K ) follows directly from Definition
(2.5.2). 
32 2 Simplicial (Co)homology

Corollary 2.5.5 Let f : K L be a simplicial map with K and L connected. Then

H 0 f : Z2 = H 0 (L) H 0 (K ) = Z2

and

H0 f : Z2 = H0 (K ) H0 (L) = Z2

are the identity isomorphism.

Proof By Proposition 2.4.1, the generator of H 0 (L) (or H 0 (K )) is the unit cocycle
1. By Lemma 2.5.4, this proves the cohomology statement. The homology statement
also follows from Proposition 2.4.1, since H0 (K ) and H0 (L) are generated by a cycle
consisting of a single vertex. 
(L) (K )
More generally, one has H 0 (L) Z2 0 and H 0 (K ) Z2 0 by Corollary 2.4.2.
Using this and Lemma 2.5.4, one gets the following corollary.
(L) (K )
Corollary 2.5.6 Let f : K L be a simplicial map. Then H 0 f : Z2 0 Z2 0
is given by H 0 f () = 0 f .
The degree of a map. Let f : K L be a simplicial map between two finite
connected n-dimensional pseudomanifolds. Define the degree deg( f ) Z2 by

0 if H n f = 0
deg( f ) = (2.5.4)
1 otherwise.

By Proposition 2.4.4, H n (K ) H n (L) Z2 . Thus, deg( f ) = 1 if and only if

H n f is the (only possible) isomorphism between H n (K ) and H n (L). By Kronecker
duality, the homomorphism Hn f may be used instead of H n f in the definition
of deg( f ). Our degree is sometimes called the mod 2 degree, since, for oriented
pseudomanifolds, it is the mod 2 reduction of a degree defined in Z (see, e.g. [179,
Exercises of Chap. 4]).
Let f : K L be a simplicial map between two finite n-dimensional pseudo-
manifolds. For Sn (L), define

d( f, ) = { Sn (K ) | f ( ) = } N. (2.5.5)

As an example, let K = L = P4 , the polygon of Example 2.4.3 with 4 edges. Let

f be defined by f (0) = 0, f (1) = 1, f (2) = 2, f (3) = 1. Then, d( f, {0, 1}) =
d( f, {1, 2}) = 2, d( f, {2, 3}) = d( f, {3, 0}) = 0 and deg( f ) = 0. This example
illustrates the following proposition.
Proposition 2.5.7 Let f : K L be a simplicial map between two finite n-
dimensional pseudomanifolds which are connected. For any Sn (L), one has

deg( f ) = d( f, ) mod 2 .
2.5 The Homomorphism Induced by a Simplicial Map 33

Proof By Proposition 2.4.4, H n (L) = Z2 is generated by the cocycle formed by the

singleton and C n f () represents the non-zero element of H n (K ) if and only if
C n f () = d( f, ) is odd. 
The interest of Proposition 2.5.7 is 2-fold: first, it tells us that deg( f ) may be
computed using any Sm (L) and, second, it asserts that d( f, ) is independent of
. Proposition 2.5.7 is the mod 2 context of the identity between the degree introduced
by Brouwer in 1910, [22, p.419], and its homological interpretation due to Hopf in
1930, [98, Sect. 2]. For a history of the notion of the degree of a map, see [40,
pp.169175].
Example 2.5.8 Let f : T 2 K be the two-fold cover of the Klein bottle K by the
for i = 1, . . . , 9.
2-torus T 2 , given in Fig. 2.5. In formulae: f (i) = i = f (i)
2
The 1-dimensional (co)homology vector spaces of T and K admit the bases:
(i) V = {[a], H 1 (T 2 ), where a is drawn in Fig. 2.5 and
[b]}
 
b = {2, 3}, {3, 6}, {6, 8}, {8, 7}, {7, 9}, {9, 2} .

= {[],
(ii) W H1 (T 2 ), where is drawn in Fig. 2.5 and
[]}
 
= {5, 7}, {7, 9}, {9, 4},
{4,
7},
{7,
9},
{9,
5} .

(iii) V = {[a], [b]} H 1 (K ), where a and b are defined in Eqs. (2.4.11) and (2.4.12)
(a drawn in Fig. 2.5).
(iv) W = {[], []} H1 (K ), where and are defined in Eq. (2.4.10) ( drawn
in Fig. 2.5).
The matrices for C f and C f in these bases are
   
10 10
C f = and C f = .
00 00

H1 () , the bases V and V are dual
Note that, under the isomorphism k: H 1 ()
and W; therefore, the matrix of C f is the transposed of that of C f .
of W

2 3
1
2
3 2 3
1 1 1 1

6 8
5
6
8 6 8
4 4 4 5
a
f a
7 9
4
7
9 7 9
5 5 5 4



1 2 3
1
2
3 1 1 2 3 1

Fig. 2.5 Two-fold cover f : T 2 K over the triangulation K of the Klein bottle given in Fig. 2.4
34 2 Simplicial (Co)homology

Now, T 2 and K are 2-dimensional pseudomanifolds and d( f, ) = 2 for each

S2 (K ). By Proposition 2.5.7, deg( f ) = 0 and both H f : H 2 (K ) H 2 (T 2 )
and H f : H2 (T 2 ) H2 (K ) vanish.
Contiguous maps. Two simplicial maps f, g: K L are called contiguous if
f () g() S(L) for all S(K ). We denote by () the subcomplex of L
generated by the simplex f ()g() S(L). For example, the inclusion K  C K
of a simplicial complex K into its cone and the constant map of K onto the cone
vertex of C K are contiguous.
Proposition 2.5.9 Let f, g: K L be two simplicial maps which are contiguous.
Then H f = H g and H f = H g.
Proof By Kronecker Duality, using Diagram (2.3.4), it is enough to prove that H f =
H g. By induction on m, we shall prove the following property:
Property H(m): there exists a linear map D: Cm (K ) Cm+1 (L) such that:
(i) D() + D() = C f () + C g() for each Cm (K ).
(ii) for each Sm (K ), D() Cm+1 ( ()) Cm+1 (L).
We first prove that Property H(m) for all m implies that H f = H g. Indeed, we
would then have a linear map D: C (K ) C+1 (L) satisfying

C f + C g = D + D . (2.5.6)

Such a map D is called a chain homotopy from C f to C g. Let Z (K ). By

Eq. (2.5.6), one has C f () + C g() = D() which implies that H f ([]) +
H g([]) in H (L).
We now prove that H(0) holds true. We define D: C0 (K ) C1 (L) as the unique
linear map such that, for v V (K ):

{ f (v), g(v)} = ({v}) if f (v) = g(v)
D({v}) =
0 otherwise.

Formula (i) being true for any {v} S0 (K ), it is true for any C0 (K ). Formula
(ii) is obvious.
Suppose that H(m 1) holds true for m 1. We want to prove that H(m) also
holds true. Let Sm (K ). Observe that D() exists by H(m 1). Consider the
chain Cm (L) defined by

= C f () + C g() + D()

Using H(m 1), one has

= C f () + C g() + D()
= C f () + C g() + D() + C f () + C g()
= 0.
2.5 The Homomorphism Induced by a Simplicial Map 35

On the other hand, Cm ( ()). As m 1, Hm ( ()) = 0 by Corollary 2.4.7.

There exists then Cm+1 ( ()) such that = . Choose such an and set
D() = . This defines D: Cm (K ) Cm+1 (L) which satisfies (i) and (ii), proving
H(m). 

Remark 2.5.10 The chain homotopy D in the proof of Proposition 2.5.9 is not explic-
itly defined. This is because several of these exist and there is no canonical way to
choose one (see [155, p. 68]). The proof of Proposition 2.5.9 is an example of the
technique of acyclic carriers which will be developed in Sect. 2.9.

Remark 2.5.11 Let f, g: K L be two simplicial maps which are contiguous. Then
| f |, |g|: |K | |L| are homotopic. Indeed, the formula F(, t) = (1 t)| f |() +
t|g|() (t [0, 1]) makes sense and defines a homotopy from | f | to |g|.

2.6 Exact Sequences

In this section, we develop techniques to obtain long (co)homology exact sequences

from short exact sequences of (co)chain complexes. The results are used in several
forthcoming sections. All vector spaces in this section are over a fixed arbitrary
field F.
Let (C1 , 1 ), (C2 , 2 ) and (C , ) be cochain complexes of vector spaces, giving
rise to cohomology graded vector spaces H1 , H2 and H . We consider morphisms
of cochain complexes J : C1 C and I : C C2 so that
J I
0 C1
C
C2 0 (2.6.1)

is an exact sequence. We call (2.6.1) a short exact sequence of cochain complexes.

Choose a GrV-morphism S: C2  C which is a section of I . The section S cannot
be assumed in general to be a morphism of cochain complexes. The linear map S:
C2m C m+1 satisfies

I S(a) = 2 I S(a) = 2 (a) ,

thus S(Z 2m ) J (C1m+1 ). We can then define a linear map : Z 2m C1m+1 by

the equation

J = S . (2.6.2)

If a Z 2m , then J 1 ( (a)) = (S(a)) = 0. Therefore, (Z 2m ) Z 1m+1 .

Moreover, if b C2m1 and a = 2 (b), then

I S(b) = 2 I S(b) = a ,
36 2 Simplicial (Co)homology

whence S(b) = S(a) + J (c) for some c C1m . Therefore (a) = 1 (c), which
shows that (B2 ) B1 . Hence, induces a linear map

: H2 H1+1

which is called the cohomology connecting homomorphism for the short exact
sequence (2.6.1).

Lemma 2.6.1 The connecting homomorphism : H2 H1+1 does not depend

on the linear section S.

Proof Let S  : C2m C m be another section of I , giving rise to  : Z 2m Z 1m+1 ,

via the equation J  = S  . Let a Z 2m . Then

S  (a) = S(a) + J (u)

for some u C1m . Therefore, the equations

J  (a) = (S(a)) + (J (u)) = (S(a)) + J (1 (u))

hold in C m+1 . This implies that  (a) = (a) + 1 (u) in Z 1m+1 , and then  (a) =
(a) in H1m+1 . 

Proposition 2.6.2 The long sequence

H J HI H J
H1m+1
H1m H m H2m

is exact.

The exact sequence of Proposition 2.6.2 is called the cohomology exact sequence,
associated to the short exact of cochain complexes (2.6.1).

Proof The proof involves 6 steps.

1. H I H J = 0 As H I H J = H (I J ), this comes from that I J = 0.
2. H I = 0 Let b Z m . Then I (b + S(I (b))) = 0. Hence, b + S(I (b)) = J (c)
for some c C1m . Therefore,

J I (b) = (S(I (b)) = (b + J (c)) = (b) + J 1 (c) = J 1 (c) ,

which proves that I (b) = 1 (c), and then H I = 0 in H1 .

3. H J =0 Let a Z 2m . Then, J (a) = (S(a)) B m+1 , so H J ([a]) =
0 in H m+1 (K ).
4. ker H J Image Let a Z 1m+1 representing [a] ker H J . This means
that J (a) = (b) for some b C m . Then, I (b) Z 2m and S(I (b)) = b + J (c)
2.6 Exact Sequences 37

for some c C1m . Therefore,

S I (b) = (b) + (J (c)) = J (a) + J (1 (c)) .

As J is injective, this implies that (I (c)) = a + 1 (c), proving that

([I (c)]) = [a].
5. ker H I Image H J Let a Z m representing [a] ker H I . This means that
I (a) = 2 (b) for some b C2m1 . Let c = (S(b)) C m . One has I (a +c) = 0,
so a + c = J (e) for some e C1m . As (a + c) = 0 and J is injective, the cochain
e is in Z 1m . As c B m , H J ([e]) = [a] in H m .
6. ker Image H I Let a Z 2m representing [a] ker . This means that
(a) = 1 (b) for some b C1m . In other words,

(S(a)) = J (1 (b)) = (J (b)) .

Hence, c = J (b) + S(a) Z m and H I ([c]) = [a]. 

We now prove the naturality of the connecting homomorphism in cohomology.
We are helped by the following intuitive interpretation of : first, we consider C1
a cochain subcomplex of C via the injection J . Second, a cocycle a Z 2m may be
C1 . Then, ([a]) = [(a)].
represented by a cochain in a C m such that (a)
More precisely:
Lemma 2.6.3 Let
J I
0 C1
C
C2 0

be a short exact sequence of cochain complexes. Then

(a) I 1 (Z 2m ) = {b C m | (b) J (C1m+1 )}.
(b) Let a Z 2m representing [a] H2m . Let b C m with I (b) = a. Then ([a]) =
[J 1 ((b))] in H1m+1 .
Proof Point (a) follows from the fact that I is surjective and from the equality 2 I =
I . For Point (b), choose a section S: C2m Cm of I . By Lemma 2.6.1, ([a]) =
[J 1 ((S(a))]. The equality I (b) = a implies that b = S(a) + J (c) for some
c C1m . Therefore,

[J 1 ((b))] = [J 1 ( S(a))] + [1 (c)] = ([a]) . 

Let us consider a commutative diagram

/ C J / C I / C / 0
0 1 2

F1 F F2 (2.6.3)
  
0 / C J / C I / C / 0
1 2
38 2 Simplicial (Co)homology

of morphisms of cochain complexes, where the horizontal sequences are exact. This
gives rise to two connecting homomorphisms : H 2 H 1+1 and : H2 H1+1 .

Lemma 2.6.4 (Naturality of the cohomology exact sequence) The diagram

/ H m H J / H m H I / H m / H m+1 H J /
1 2 1

H F1 HF H F2 H F1
   
... / Hm H J / Hm HI / Hm / H m+1 H J /
1 2 1

is commutative.

Proof The commutativity of two of the square diagrams follows from the func-
toriality of the cohomology: H F H J = H J H F1 since F J = J F1
and H F2 H I = H I H F since F2 I = I F. It remains to prove that
H F1 = H F2 .
Let a Z 2m representing [a] H 2m . Let b C m with I(b) = a. Then, I F(b) =
F2 (a). Using Lemma 2.6.3, one has

H F2 ([a]) = [J 1 F(b)]

= [J 1 F (b)]

= [F1 J1 (b)]

= H F1 ([a]) . 

We are now interested in the case where the cochain complexes (Ci , i ) and
(C , )are parts of Kronecker pairs

P1 = (C1 , 1 ), (C,1 , 1 ),  , 1 , P2 = (C2 , 2 ), (C,2 , 2 ) , 2

and

P = (C , ), (C , ),  ,  .

Let us consider two morphism of Kronecker pairs, (J, j) from P to P1 and (I, i)
from P2 to P. We suppose that the two sequences

J I
0 C1
C
C2 0 (2.6.4)

and
i j
0 C,2
C
C,1 0 (2.6.5)
2.6 Exact Sequences 39

are exact sequences of (co)chain complexes. Note that, by Lemma 2.3.8, (2.6.4) is
exact if and only if (2.6.5) is exact. Exact sequence (2.6.4) gives rise to the coho-
mology connecting homomorphism : H2 H1+1 . We construct a homology
connecting homomorphism in the same way. Choose a linear section s: C,1 C
of j, not required to be a morphism of chain complexes. As in the cohomology
setting, one can defines : Z m+1,1 Z m,2 by the equation

i = s . (2.6.6)

We check that (Bm+1,1 ) Bm,2 . Hence induces a linear map

: H+1,1 H,2

called the homology connecting homomorphism for the short exact sequence (2.6.5).
Lemma 2.6.5 The connecting homomorphism : H+1,1 H,2 does not depend
on the linear section s.
Proof The proof is analogous to that of Lemma 2.6.1 and is left as an exercise to the
reader. 
Lemma 2.6.6 The connecting homomorphisms : H2m H1m+1 and : Hm+1,1
Hm,1 satisfy the equation

 (a), 1 = a, ()2

for all a H2m , Hm+1,1 and all m N. In other words, ( , ) is a morphism

of Kronecker pairs from (H1 , H,1 ,  , 1 ) to (H2 , H,2 ,  , 2 ).
Proof Let a Z 2m represent a and Z m+1,1 represent . Choose linear sections
S and s of I and j. Using Formulae (2.6.2) and (2.6.6), one has

 (a), 1 =  (a),
 1
=  (a),

j s()
1
= J (a),
s()

= S(a),
s()
i ()
= S(a),
= I S(a),
() 2
()
= a, 2 = a, ()2 . 

Proposition 2.6.7 The long sequence

H i H j H i
Hm,2 Hm Hm,1
Hm1,2

is exact.
40 2 Simplicial (Co)homology

The exact sequence of Proposition 2.6.7 is called the homology exact sequence asso-
ciated to the short exact of chain complexes (2.6.5). It can be established directly, in
an analogous way to that of Proposition 2.6.2. To make a change, we shall deduce
Proposition 2.6.7 from Proposition 2.6.2 by Kronecker duality.
Proof By our hypotheses couples (I, i) and (J, j) are morphisms of Kronecker pairs,
and so is ( , ) by Lemma 2.6.6. Using Diagram (2.3.4), we get a commutative
diagram

 
(H i) (H j)
o (Hm,2 ) o (Hm ) o (Hm,1 ) o Hm1,2 o


O O O O
k k k k .
HI H J
o H1m o Hm o H1m o H2m1 o

By Proposition 2.6.2, the bottom sequence of the above diagram is exact. Thus,
the top sequence is exact. By Lemma 2.3.8, the sequence of Proposition 2.6.7 is
exact. 
Let us consider commutative diagrams

/ C J / C I / C / 0
0 1 2

F1 F F2 (2.6.7)
  
0 / C J / C I / C / 0
1 2

and
j i
0 o C ,1 o C o C ,2 o 0
O O O
f1 f f2 (2.6.8)
j
0 o C,1 o C o C,2 o
i
0

such that the horizontal sequences are exact, Fi and F are morphisms of cochain
complexes and f i and f are morphisms of cochain complexes.
Lemma 2.6.8 (Naturality of the homology exact sequence) Suppose that (Fi , f i )
and (F, f ) are morphisms of Kronecker pairs. Then, the diagram

H j
/ Hm,2 H i
/ Hm / Hm,1 / Hm1,2 H i
/

H f 2 H f H f 1 H f 2
   
H i H j H i
... / H m,2 / H / H m,1 / H m1,2 /
m

is commutative.
2.6 Exact Sequences 41

Proof By functoriality of the homology, the square diagrams not involving com-
mute. It remains to show that H f 1 =H f 2 . As H F1 = H F2 by
Lemma 2.6.4, one has

a, H f 1 ()2 = H F1 (a), 1 =  H F2 (a), 1 = a, H f 1 ()2

for all a H 2m1 and Hm,1 . By Lemma 2.3.3, this implies that H f 1 =
H f 2 . 

2.7 Relative (Co)homology

A simplicial pair is a couple (K , L) where K is a simplicial complex and L is a

subcomplex of K . The inclusion i: L  K is a simplicial map. Let a C m (K ).
If, using Definition 2.2.1a of Sect. 2.2, we consider a as a subset of Sm (K ), then
C i(a) = a Sm (L). If we see a as a map a: Sm (K ) Z2 , then C i(a) is the
restriction of a to Sm (L). We see that C i: C (K ) C (L) is surjective. Define
C i

C m (K , L) = ker C m (K ) C m (L)

and C (K , L) = mN C m (K , L). This definition implies that

C m (K , L) is the set of subsets of Sm (K ) Sm (L);
if K is a finite simplicial complex, C m (K , L) is the vector space with basis
Sm (K ) Sm (L).
As C i is a morphism of cochain complexes, the coboundary : C (K ) C (K )
preserves C (K , L) and gives rise to a coboundary : C (K , L) C (K , L) so that
(C (K , L), ) is a cochain complex. The cocycles Z (K , L) and the coboundaries
B (K , L) are defined as usual, giving rise to the definition

H m (K , L) = Z m (K , L)/B m (K , L) .

The graded Z2 -vector space H (K , L) = mN H m (K , L) is the simplicial relative

cohomology of the simplicial pair (K , L).
When useful, the notations K , L and K ,L are used for the coboundaries of the
cochain complexes C (K ), C (L) and C (K , L). We denote by j the inclusion j :
C (K , L)  C (K ), which is a morphism of cochain complexes, and use the same
notation j for the induced linear map j : H (K , L) H (K ) on cohomology.
We also use the notation i for both C i and H i. We get thus a short exact sequence
of cochain complexes

j i
0 C (K , L)
C (K )
C (L) 0 . (2.7.1)
42 2 Simplicial (Co)homology

If a C m (L), any cochain a C m (K ) with i (a)

= a is called a extension of
a as a cochain in K . For instance, the 0-extension of a is defined by a = a
Sm (L) Sm (K ). Using Sect. 2.6, Exact sequence (2.7.1) gives rise to a (simplicial
cohomology) connecting homomorphism

: H (L) H +1 (K , L) .

It is induced by a linear map : Z m (L) Z m+1 (K , L) characterized by the

equation j = K S for some (or any) linear section S: C m (L) C m (K ) of i ,
not required to be a morphism of cochain complex. For instance, one can take S(a)
to be the 0-extension of a. Using that C (K , L) is a chain subcomplex of C (K ),
the following statement makes sense and constitutes a useful recipe for computing
the connecting homomorphism .

Lemma 2.7.1 Let a Z m (L) and let a C m (K ) be any extension of a as an

is an (m + 1)-cocycle of (K , L) representing (a).
m-cochain of K . Then, K (a)

Proof Choose a linear section S: C m (L) C m (K ) such that S(a) = a.

The
equation j = K S proves the lemma. 

We can now use Proposition 2.6.2 and get the following result.

Proposition 2.7.2 The long sequence

j i j
H m (K , L)
H m (K )
H m (L)
H m+1 (K , L)

is exact.

The exact sequence of Proposition 2.7.2 is called the simplicial cohomology exact
sequence, or just the simplicial cohomology sequence, of the simplicial pair (K , L).
We now turn our interest to homology. The inclusion L  K induces an inclu-
sion i : C (L)  C (K ) of chain complexes. We define Cm (K , L) as the quotient
vector space

Cm (K , L) = coker i : Cm (L)  Cm (K ) .

As i is a morphism of chain complexes, C (K , L) = mN Cm (K , L) inherits

a boundary operator = K ,L : C (K , L) C1 (K , L). The projection j :
C (K ) C (K , L) is a morphism of chain complexes and one gets a short exact
sequence of chain complexes

i j
0 C (L)
C (K )
C (K , L) 0 . (2.7.2)

The cycles and boundaries Z (K , L) and B (K , L) are defined as usual, giving rise
to the definition
2.7 Relative (Co)homology 43

Hm (K , L) = Z m (K , L)/Bm (K , L) .

The graded Z2 -vector space H (K , L) = mN Hm (K , L) is the relative homology

of the simplicial pair (K , L). As before, the notations K and L may be used for the
boundary operators in C (K ) and C (L) and i and j are also used for the induced
maps in homology.
Since the linear map i : C (L)  C (K ) is induced by the inclusion of bases
S(L)  S(K ), the quotient vector space C (K , L) may be considered as the vector
space with basis S(K ) S(L). This point of view provides a tautological linear
map s: C (K , L) C (K ), which is a section of j but not a morphism of chain
complexes.
The Kronecker pairings for K and L are denoted by  ,  K and  ,  L , both at the
levels of (co)chains and of (co)homology. As  j (K , L), i (L) K = 0, we get a
bilinear map

, K ,L
C m (K , L) Cm (K , L) Z2 .

The formula

a,  K ,L =  j (a), s() K (2.7.3)

holds for all a C m (K , L), Cm (K , L) and all m N. Observe also that the
formula

S(b), i () K = b,  L (2.7.4)

holds for all b C m (L), Cm (L) and all m N.

Lemma 2.7.3 C (K , L), K ,L , C (K , L), K ,L ,  ,  K ,L is a Kronecker pair.

Proof We first prove that  K ,L (a),  K ,L = a, K ,L () K ,L for all a C m (K , L)
and all Cm+1 (K , L) and all m N. Indeed, one has

 K ,L (a),  K ,L =  j K ,L (a), s() K

=  K j (a), s() K
=  j (a), K s() K (2.7.5)

Observe that j K s() = K ,L () and therefore K s() = s K ,L () + i (c)

for some c Cm (L). Hence, the chain of equalities in (2.7.5) may be continued

 K ,L (a),  K ,L =  j (a), K s() K

=  j (a), s K ,L () + i (c) K
=  j (a), s K ,L () K +  j (a), i (c) K
  
0
= a, K ,L () K ,L . (2.7.6)
44 2 Simplicial (Co)homology

It remains to prove that the linear map k: C (K , L) C (K , L) given by

k(a) = a,  is an isomorphism. As the inclusion i: L  K is a simplicial map, the
couple (C i, C i) is a morphism of Kronecker pairs by Lemma 2.5.1 and the result
follows from Lemma 2.3.9. 

Passing to homology then produces three Kronecker pairs with vanishing

(co)boundary operators:

P L = (H (L), H (L), ,  L ), P K = (H (K ), H (K ), ,  K )

and

P K ,L = (H (K , L), H (K , L), ,  K ,L ) .

Using Sect. 2.6, short exact sequence (2.7.2) gives rise to the (simplicial homology)
connecting homomorphism

: H (K , L) H1 (L) .

Z m (K , L) Z m1 (L) characterized by the equation

It is induced by a linear map :

j = K s ,

using the section s of j defined above (or any other one).

Lemma 2.7.4 The following couples are morphisms of Kronecker pairs:

(a) (i , i ), from P L to P K .
(b) ( j , j ), from P K to P K ,L .
(c) ( , ), from P K ,L to P L .

Proof As the inclusion L  K is a simplicial map, Point (a) follows from

Lemma 2.5.1. Point (c) is implied by Lemma 2.6.6. To prove Point (b), let a
C m (K , L) and Cm (K ). Observe that s( j ()) = +i () for some Cm (L)
and that  j (a), i () K = 0. Therefore:

a, j () K ,L =  j (a), s j () K =  j (a),  K 

Proposition 2.6.7 now gives the following result.

Proposition 2.7.5 The long sequence

i j i
Hm (L)
Hm (K )
H m (K , L)
Hm1 (L)

is exact.
2.7 Relative (Co)homology 45

The exact sequence of Proposition 2.7.5 is called the (simplicial) homology exact
sequence, or just the (simplicial) cohomology sequence, of the simplicial pair (K , L).
We now study the naturality of the (co)homology sequences. Let (K , L) and
(K  , L  ) be simplicial pairs. A simplicial map f of simplicial pairs from (K , L)
to (K  , L  ) is a simplicial map f K : K K  such that the restriction of f to L
is a simplicial map f L : L L  . The morphism C f K : C (K  ) C (K ) then
restricts to a morphism of cochain complexes C f : C (K  , L  ) C (K , L) and
the morphism C f K : C (K ) C (K  ) descends to a morphism of chain complexes
C f : C (K , L) C (K  , L  ). The couples (C f K , C f K ) and (C f L , C f L ) are
morphisms of Kronecker pairs by Lemma 2.5.1. We claim that (C f, C f ) is a
morphism of Kronecker pair from (C (K , L), . . . ) to (C (K  , L  ), . . . ). Indeed, let
a C m (K  , L  ) and Cm (K , L). One has

C f (a),  K ,L =  j C f (a), s() K

= C f K j  (a), s() K
=  j  (a), C f K s() K 
=  j  (a), C f K s() K  (2.7.7)

and

a, C f () K  ,L  =  j  (a), s  C f () K  (2.7.8)

The equation j s  C f () = j C f K s() = f () implies that s  C

f () = C f K s() + i  () for some Cm (L  ). As  j  (a), i  () K  = 0,
Equations (2.7.7) and (2.7.8) imply that C f (a),  K ,L = a, C f () K  ,L  .
Lemmas 2.6.4 and 2.6.8 then imply the following

Proposition 2.7.6 The cohomology and homology sequences are natural with
respect to simplicial maps of simplicial pairs. In other words, given a simplicial
map of simplicial pairs f : (K , L) (K  , L  ), the following diagrams

j  i   j 
/ H m (K  , L  ) / H m (K  ) / H m (L  ) / H m+1 (K  , L  ) /

H f H fK H fL H f
   
j i j
... / H m (K , L) / H m (K ) / H m (L) / H m+1 (K , L) /

and

i j i
/ Hm (L) / Hm (K ) / Hm (K , L) / Hm1 (L) /

H f L H f K H f H f L
 i 
 j
 
 i 
... / Hm (L  ) / H m (K  ) / Hm (K  , L  ) / Hm1 (L  ) /

are commutative.
46 2 Simplicial (Co)homology

We finish this section by the exact sequences for a triple. A simplicial triple is a
triplet (K , L , M) where K is a simplicial complex, L is a subcomplex of K and M
is a subcomplex of L. A simplicial map f of simplicial triples, from (K , L , M) to
(K  , L  , M  ) is a simplicial map f K : K K  such that the restrictions of f K to L
and M are simplicial maps f L : L L  and f M : M M  .
A simplicial triple T = (K , L , M) gives rise to pair inclusions

i j
(L , M)
(K , M)
(K , L)

and to a commutative diagram

j K ,L
iK ,L
0 / C (K , L) / C (K ) / C (L) / 0
O
C j = id i L ,M (2.7.9)
  
j K ,M
iK ,M
0 / C (K , M) / C (K ) / C (M) / 0

where the horizontal lines are exact sequences of cochain complexes. A diagram-
chase shows that the morphism i K ,L j K ,M , which sends C (K , M) to C (L), has
image C (L , M) and kernel the image of C j. This morphism coincides with C i.
We thus get a short exact sequence of cochain complexes

C j C i
0 C (K , L) C (K , M) C (L , M) 0 . (2.7.10)

The same arguments with the chain complexes gives a short exact sequence

C i C j
0 C (L , M) C (K , M) C (K , L) 0 . (2.7.11)

As above in this section, short exact sequences (2.7.10) and (2.7.11) produces con-
necting homomorphisms T : H (L , M) H +1 (K , L) and T : H (K , L)
C1 (L , M). They satisfy T (a),  = a, T () as well as following proposition.

Proposition 2.7.7 ((Co)homology sequences of a simplicial triple) Let T = (K , L , M)

be a simplicial triple. Then,
(a) the sequences

H j H i T H j
H m (K , L) H m (K , M) H m (L , M) H m+1 (K , L)

and
H i H j T H i
Hm (L , M) Hm (K , M) Hm (K , L) Hm1 (L , M)

are exact.
2.7 Relative (Co)homology 47

(b) the exact sequences of Point (a) are natural for simplicial maps of simplicial
triples.

Remark 2.7.8 As H () = 0, we get a canonical GrV-isomorphisms H (K , )
H (K ), etc. Thus, the (co)homology sequences for the triple (K , L , ) give back
those of the pair (K , L)

H j H i H j
H m (K , L) H m (K ) H m (L)
H m+1 (K , L)
(2.7.12)
and
H i H j H i
Hm (L) Hm (K ) H m (K , L)
Hm1 (L) (2.7.13)

where i: L K and j: (K , ) (K , L) denote the inclusions. This gives a more

precise description of the morphisms j and j of Propositions 2.7.2 and 2.7.5.

2.7.9 Historical note. The relative homology was introduced by S. Lefschetz in

1927 in order to work out the Poincar duality for manifolds with boundary
(see, e.g. [40, p. 58], [51, p. 47]). The use of exact sequences occurred in several parts
of algebraic topology after 1941 (see, e.g. [40, p. 86], [51, p. 47]). The (co)homology
exact sequences play an essential role in the axiomatic approach of Eilenberg-
Steenrod, [51].

2.8 Mayer-Vietoris Sequences

Let K be a simplicial complex with two subcomplexes K 1 and K 2 . We suppose that

K = K 1 K 2 (i.e. S(K ) = S(K 1 ) S(K 2 )). We call (K , K 1 , K 2 ) a simplicial
triad. Then, K 0 = K 1 K 2 is a subcomplex of K 1 , K 2 and K , with S(K 0 ) =
S(K 1 ) S(K 2 ). The Mayer-Vietoris sequences relate the (co)homology of X to that
of X i , generalizing Lemma 2.4.10. The various inclusions are denoted as follows

K0
i1
/ K1

i2 j1 (2.8.1)
 j2

K2 / K.

The notations i 1 , j1 , , stand for both C i 1 , C j1 , etc, and H i 1 , H j1 , etc. The

same holds for chains and homology: i 1 for both C i 1 and H i 1 , etc. Diagram (2.8.1)
induces two diagrams
48 2 Simplicial (Co)homology

j1
C (K 0 ) / / C (K 1 )
i 1
C (K ) / / C (K 1 )
 
j2 i 1 and i 2 j1
   
i 2 j2
C (K 2 ) / / C (K 0 ) C (K 2 ) / / C (K ) .

The cohomology diagram is Cartesian (pullback) and the homology diagram is co-
Cartesian (pushout). Therefore, the sequence

( j1 , j2 ) i 1 +i 2
0 C (K ) C (K 1 ) C (K 2 ) C (K 0 ) 0 (2.8.2)

is an exact sequence of cochain complexes and the sequence

(i 1 ,i 2 ) j1 + j2
0 C (K 0 ) C (K 1 ) C (K 2 ) C (K ) 0 (2.8.3)

is an exact sequence of chain complexes.

Consider the Kronecker pairs (C (K i ), C (K i ),  , i ) for i = 0, 1, 2, and the
Kronecker pair (C (K ), C (K ),  , ). A bilinear map
   
 ,  : C (K 1 ) C (K 2 ) C (K 1 ) C (K 2 ) Z2

is defined by

(a1 , a2 ), (1 , 2 ) = a1 , 1 1 + a2 , 2 2 .

We check that (C (K 1 ) C (K 2 ), C (K 1 ) C (K 2 ),  ,  ) is a Kronecker pair

and that the couples (( j1 , j2 ), j1 + j2 ) and (i 1 + i 2 , (i 1 , i 2 )) are morphisms of
Kronecker pairs. By Sect. 2.6, there exist linear maps M V : H (K 0 ) H +1 (K )
and M V : H (K ) H1 (K 0 ) which, by Propositions 2.6.2 and 2.6.7, give the
following proposition.

Proposition 2.8.1 (Mayer-Vietoris sequences) The long sequences

( j1 , j2 ) i 1 +i 2 M V
H m (K ) H m (K 1 ) H m (K 2 ) H m (K 0 ) H m+1 (K )

and
(i 1 ,i 2 ) j1 + j2 M V
Hm (K 0 ) Hm (K 1 ) Hm (K 2 ) Hm (K ) Hm1 (K 0 )

are exact.

The homomorphisms M V and M V are called the Mayer-Vietoris connecting

homomorphisms in (co)homology. By Lemma 2.6.6, they satisfy  M V (a),  =
2.8 Mayer-Vietoris Sequences 49

a, M V ()0 for all a H m (K 0 ), all Hm+1 (k) and all m N. To define the
connecting homomorphisms, one must choose a linear section S of i 1 + i 2 and s of
j1 + j2 . One can choose S(a) = (S1 (a), 0), where S1 : C (K ) C (K 1 ) is the
tautological section of i 1 given by the inclusion S(K 0 )  S(K 1 ) (see Sect. 2.7). A
choice of s is given, for S(K ), by

(, 0) if S(K 1 )
s() =
/ S(K 1 ) .
(0, 0) if

These choices produce linear maps M V : Z (K 0 ) Z +1 (K ) and M V : Z (K )

Z 1 (K 0 ), representing M V and M V and defined by the equations

( j1 , j2 ) M V = (1 , 2 ) S and (i 1 , i 2 ) M V = (1 , 2 ) s .

(The apparent asymmetry of the choices has no effect by Lemma 2.6.1 and its homol-
ogy counterpart: exchanging 1 and 2 produces other sections, giving rise to the same
connecting homomorphisms.)
Finally, the Mayer-Vietoris sequences are natural for maps of simplicial triads. If
T = (K , K 1 , K 2 ) and T  = (K  , K 1 , K 2 ) are simplicial triads and if f : K K 
is a simplicial map such that f (K i ) K i , then the Mayer Vietoris sequences of T
and T  are related by commutative diagrams, as in Proposition 2.7.6. This is a direct
consequence of Lemmas 2.6.4 and 2.6.8.

2.9 Appendix A: An Acyclic Carrier Result

The powerful technique of acyclic carriers was introduced by Eilenberg and MacLane
in 1953 [50], after earlier work by Lefschetz. Proposition 2.9.1 below is a very par-
ticular example of this technique, adapted to our needs. For a full development of
acyclic carriers, see, e.g., [155, Chap. 1,Sect. 13].
be two chain complexes and let : C C be a
Let (C , ) and (C , )
morphism of chain complexes. We suppose that Cm is equipped with a basis Sm
for each m and denote by S the union of all Sm . An acyclic carrier A for with
respect to the basis S is a correspondence which associates to each s S a subchain
complex A (s) of C such that
(a) (s) A (s).
(b) H0 (A (s)) = Z2 and Hm (A (s)) = 0 for m > 0.
(c) let s Sm and t Sm1 such that t occurs in the expression of s in the basis
Sm1 . Then A (t) is a subchain complex of A (s) and the inclusion A (t)
A (s) induces an isomorphism on H0 .
(d) if s S0 C0 = Z 0 , then H0 (s) = 0 in H0 (A (s)).
50 2 Simplicial (Co)homology

Proposition 2.9.1 Let and  be two morphisms of chain complexes from (C , )

Suppose that and  admit the same acyclic carrier A with respect to
to (C , ).
some basis S of C . Then H = H  .

Proof The proof is similar to that of Proposition 2.5.9. By induction on m, we shall

prove the following property:
Property H(m): there exists a linear map D: Cm C m+1 such that:
(i) D() + D() = () +  () for all Cm .
(ii) for each s Sm , D(s) Am+1 (s).

Property H(m) for all m implies that H = H  . Indeed, we then have a linear
map D: C C +1 satisfying

+  = D + D . (2.9.1)

Let Z . By Eq. (2.9.1), one has () + () = D() which implies that

H ([]) + H  ([]) in H .
Let us prove H(0). Let s S0 . In H0 (A (s)) = Z2 , one has H0 (s) = 0
and H0  (s) = 0. Therefore H (s) = H  (s) in H0 (A (s)). This implies that
(s) +  (s) = ( s ) for some s A1 (s). We set D(s) = s . This procedure,
for each s S0 , provides a linear map D: C0 C 1 , which, as C0 = 0, satisfies
(s) +  (s) = D() + D(()).
We now prove that H(m 1) implies H(m) for m 1. Let s Sm . The chain
D(s) exists in Am (s) by H(m 1). Let Am (s) defined by

= (s) +  (s) + D(s)

Using H(m 1), one checks that = 0. Since Hm (A (s)) = 0, there exists
Am+1 (s) such that = . Choose such an element and set D() = . This
defines D: Cm C m+1 which satisfies (i) and (ii), proving H(m). 

2.10 Appendix B: Ordered Simplicial (Co)homology

This technical section may be skipped in a first reading. It shows that simplicial
(co)homology may be defined using larger sets of (co)chains, based on ordered
simplexes. This will be used for comparisons between simplicial and singular
(co)homology (see 17) and to define the cup and cap products in Chap. 4.
Let K be a simplicial complex. Define

Sm (K ) = {(v0 , . . . , vm ) V (K )m+1 | {v0 , . . . , vm } S(K )} .

Observe that dim{v0 , . . . , vm } m and may be strictly smaller if there are repetitions
amongst the vi s. An element of Sm (K ) is an ordered m-simplex of K .
2.10 Appendix B: Ordered Simplicial (Co)homology 51

The definitions of ordered (co)chains and (co)homology are the same those for
the simplicial case (see Sect. 2.2), replacing the simplexes by the ordered simplexes.
We thus set

Definition 2.10.1 (subset definitions)

(a) An ordered m-cochain is a subset of Sm (K ).
(b) An ordered m-chain is a finite subset of Sm (K ).

The set of ordered m-cochains of K is denoted by C m (K ) and that of ordered

m-chains by C m (K ). As in Sect. 2.2, Definition 2.10.1 are equivalent to

Definition 2.10.2 (colouring definitions)

(a) An ordered m-cochain is a function a: Sm (K ) Z2 .
(b) An ordered m-chain is a function : Sm (K ) Z2 with finite support.

Definition 2.10.2 endow C m (K ) and C m (K ) with a structure of a Z2 -vector space.

The singletons provide a basis of C m (K ), in bijection with Sm (K ). Thus, Definition
2.10.2.b is equivalent to

Definition 2.10.3 C m (K ) is the Z2 -vector space with basis Sm (K ):


C m (X ) = Z2 .
Sm (X )

We consider the graded Z2 -vector spaces C (K ) = mN C m (K ) and C (K ) =

mN C m (K ). The Kronecker pairing on ordered (co)chains

,
C m (K ) C m (K ) Z2

is defined, using the various above definitions, by the equivalent formulae

a,  = 
(a ) (mod 2) using Definition 2.10.1a and b
=  a() using Definitions 2.10.1a and 2.10.2b (2.10.1)
= Sm (K ) a()() using Definitions 2.10.2a and b.

As in Lemma 2.2.4, we check that the map k: C m (K ) C m (K ) , given by k(a) =

a, , is an isomorphism.
The boundary operator : C m (K ) C m1 (K ) is the Z2 -linear map defined, for
(v0 , . . . , vm ) Sm (K ) by


m
0 , . . . , vm ) =
(v (v0 , . . . , vi , . . . , vm ) , (2.10.2)
i=0
52 2 Simplicial (Co)homology

where (v0 , . . . , vi , . . . , vm ) Sm1 is the m-tuple obtained by removing vi . The

coboundary operator : C m (K ) C m+1 (K ) is defined by the equation

 = a, 
a, . (2.10.3)

C (K ), ,
With these definition, (C (K ), ,  , ) is a Kronecker pair. We define
the vector spaces of ordered cycles Z (K ), ordered boundaries B (K ), ordered cocy-

cles Z (K ), ordered coboundaries B (K ), ordered homology H (K ) and ordered
cohomology H (K ) as in Sect. 2.3. By Proposition 2.3.5, the pairing on (co)chain
descends to a pairing
,
H m (K ) Hm (K ) Z2

so that the map k: H m H m , given by k(a) = a, , is an isomorphism (ordered
Kronecker duality).
Example 2.10.4 Let K = pt be a point. Then, Sm ( pt) contains one element for
each integer m, namely the (m + 1)-tuple ( pt, . . . , pt). Then, C m ( pt) = Z2 for all
m N and the chain complex looks like

C 2k+1 ( pt)
C 2k ( pt)
C 2k1 ( pt) C 1 ( pt)
C 0 ( pt) 0 .
0 0 0

Therefore,

0 if > 0
H ( pt) H ( pt)
Z2 if = 0 .

One sees that, for a simplicial complex reduced to a point, the ordered (co)homology
and the simplicial (co)homology are isomorphic.
Example 2.10.5 The unit cochain 1 C 0 (K ) is defined as 1 = S0 (K ). It is a cocycle
and defines a class 1 = H 0 (K ). If K is non-empty and connected, then H 0 (K ) Z2
generated by 1. Then H0 (K ) Z2 by Kronecker duality; one has Z 0 (K ) = C 0 (K )
and Z 0 (K ) represents the non-zero element of H0 (K ) if and only if  is odd.
The proofs are the same as for Proposition 2.4.1.
Example 2.10.6 Let L be a simplicial complex and C L be the cone on L. Then

0 if > 0
H (C L) H (C L)
Z2 if = 0 .

The proof is the same as for Proposition 2.4.6, even simpler, since D: C m (C L)
C m+1 (C L) is defined, for (v0 , . . . , vm ) Sm (C L) by the single line formula
D(v0 , . . . , vm ) = (, v0 , . . . , vm ).
Let f : L K be a simplicial map. We define C f : C (L) C (K ) as the
degree 0 linear map such that
2.10 Appendix B: Ordered Simplicial (Co)homology 53

C f (v0 , . . . , vm ) = ( f (v0 ), . . . , f (vm ))

for all (v0 , . . . , vm ) S(L). The degree 0 linear map C f : C (K ) C (L) is
defined by

C f (a),  = a, C f () .

By Lemma 2.3.6, (C f, C f ) is a morphism of Kronecker pairs.

We now construct a functorial isomorphism between the ordered and non-ordered
(co)homologies, its existence being suggested by the previous examples. Define :
C (K ) C (K ) by

{v0 , . . . , vm } if vi = v j for all i = j
((v0 , . . . , vm )) =
0 otherwise.

We check that is a morphism of chain complexes. We define : C (K ) C (K )

by requiring that the equation  (a),  = a, () holds for all a C (K )
and all C (K ). By Lemma 2.3.6, is a morphism of cochain complexes
and ( , ) is a morphism of Kronecker pairs between (C (K ), C (K )) and
(C (K ), C (K )). It thus defines a morphism of Kronecker pairs (H , H )
between ( H (K ), H (K )) and (H (K ), H (K )).
To define a morphism of Kronecker pairs in the other direction, choose a simplicial
order on K (see 2.1.8). Define : C (K ) C (K ) as the unique linear map
such that

({v0 , . . . , vm }) = (v0 , . . . , vm ) ,

where v0 v1 vm . We check that is a morphism of chain complexes

and define : C (K ) C (K ) by requiring that the equation  (a),  =
a, () holds for all a C (K ) and all C (K ). By Lemma 2.3.6,
( , ) is a morphism of Kronecker pairs between (C (K ), C (K )) and (C (K ),
C (K )). It then defines a morphism of Kronecker pairs (H , H ) between
(H (K ), H (K )) and ( H (K ), H (K )).

Proposition 2.10.7 H H = id H (K ) and H H = id H (K ) .

Proof As = idC (K ) , the first equality follows from Lemma 2.3.7. For
the second one, let (v0 , . . . , vm ) Sm (K ). Let = {v0 , . . . , vm } Sk (K ) with
k m. Clearly, (v0 , . . . , vm ) C ().
By what was seen in Examples 2.10.5
and 2.10.6, the correspondence (v0 , . . . , vm )  C ({v0 , . . . , vm }) is an acyclic car-
rier A , with respect to the basis S (K ), for both idC(K
) and . Therefore, the
equality H H = id H (K ) follows by Lemma 2.3.7 and Proposition 2.9.1.
54 2 Simplicial (Co)homology

Applying Kronecker duality to Proposition 2.10.7 gives the following

Corollary 2.10.8 H H = id H (K ) and H H = id H (K ) .

Corollary 2.10.9 H and H are isomorphisms.

Corollary 2.10.10 H and H are isomorphisms which do not depend on the
simplicial order .
Proof This follows from Proposition 2.10.7 and Corollary 2.10.8, since H and
H do not depend on . 
We shall see in Sect. 4.1 that H and H are isomorphisms of graded Z2 -
algebras. We now prove that they are also natural with respect to simplicial maps.
Let f : L K be a simplicial map. Let C f : C (L) C (K ) be the unique linear
map such that
C f ((v0 , . . . , vm )) = ( f (v0 ), . . . , f (vm ))

for each (v0 , . . . , vm ) Sm (K ). Doing this for each m N produces a GrV-

morphism C f : C (L) C (K ). The formula C f = C f is straightforward
(much easier than that for non-ordered chains). Hence, we get a GrV-morphism
H f : H (L) H (K ). A GrV-morphism C f : C (K ) C (L) is defined
by the equation C f (a),  = a, C f () required to hold for all a C m (L),
C m (K ) and all m N. It is a cochain map and induces a GrV-morphism H f :
H (K ) H (L), Kronecker dual to H f .
Proposition 2.10.11 Let f : L K be a simplicial map. Let be a simplicial
order on K and  be a simplicial order on L. Then the diagrams

H f H f
H (L) / H (K ) H (K ) / H (L)

K K O O
and 
H  H H H H H H H
 
H f

H (L)
H f
/ H (K ) H (K ) / H (L)

are commutative.
Proof By Kronecker duality, only the homology statement requires a proof. It is
enough to prove that H f H = H H f since the formula H f H  =
H H f will follow by Corollary 2.10.8. Finally, the formula C f C  =
C C f is straightforward. 
The above isomorphism results also work in relative ordered (co)homology. Let
(K , L) be a simplicial pair. Denote by i: L  K the simplicial inclusion. We define
the Z2 -vector space of relative ordered (co)chain by

C i

C m (K , L) = ker C m (K ) C m (L)
2.10 Appendix B: Ordered Simplicial (Co)homology 55

and

C m (K , L) = coker i : C m (L)  C m (K ) .

These inherit (co)boundaries : C (K , L) C (K , L) and = C (K , L)

C 1 (K , L) which give rise to the definition of relative ordered (co)homology
H (k, L) and H (K , L). Connecting homomorphisms : H (L) H +1 (K , L)
and : H (K , L) H 1 (L) are defined as in Sect. 2.7, giving rise to long exact
sequences. Our homomorphisms : C (K ) C (K ) and : C (K ) C (K )
satisfy (C (L)) C (L) and (C (L) C (L), giving rise to homo-
morphisms on relative (co)chains and relative (co)homology H : H (K , L)
H (K , L), etc. Proposition 2.10.7 and Corollary 2.10.8 and their proofs hold in
relative (co)homology. Hence, as for Corollaries 2.10.9 and 2.10.10, we get
Corollary 2.10.12 H : H (K , L) H (K , L) and H : H (K , L)
H (K , L) are isomorphisms.
Corollary 2.10.13 H : H (K , L) H (K , L) and H : H (K , L)
H (K , L) are isomorphisms which do not depend on the simplicial order .

2.11 Exercises for Chapter 2

2.1. Let Fn be the full complex on the set {0, 1, . . . , n} (see p. 24). What are the
2-simplexes of the barycentric subdivision F2 of F2 ? How many n-simplexes
does Fn contain?
2.2. Compute the Euler characteristic and the Poincar polynomial of the k-
skeleton Fnk of Fn .
2.3. Let X be a metric space and let > 0. The Vietoris-Rips complex X of X
is the simplicial complex whose simplexes are the finite non-empty subset of
X whose diameter is < (the diameter of A X is the least upper bound of
d(x, y) for x, y A). In particular, V (X ) = X .
(a) Describe |X | for various
when X isthe set of vertices of a cube of edge 1
in R3 . In particular, if 2 < 3, show that |X | is homeomorphic
to S 3 .
(b) Let X be the space n-th roots of unity, with the distance d(x, y) being the
minimal length of an arc of the unit circle joining x to y. Suppose that
4/n < 6/n.
(i) If n = 6, show that |X | is homeomorphic to S 2 .
(ii) If n 7 is odd, show that |X | is homeomorphic to a Mbius band.
(iii) If n 7 is even, show that |X | is homeomorphic to S 1 [0, 1].
Note: the complex X was introduced by Vietoris in 1927 [201]. After its re-
introduction by E. Rips for studying hyperbolic groups, it has been popularized
56 2 Simplicial (Co)homology

under the name of Rips complex. For some developments and applications, see
[84, 129] and Wikipedias page Vietoris-Rips complex.
2.4. Let  = (1 , . . . , n ) 
Rn>0 . A subset J of {1, . . . , n} is called -short (or just
short) if iJ i < i J / i . Show that short subsets are the simplexes of a
simplicial complex Sh() with V (Sh()) J (used in Sect. 10.3). Describe
Sh(1, 1, 1, 1, 3), Sh(1, 1, 3, 3, 3) and Sh(1, 1, 1, 1, 1). Compute their Euler
characteristics and their Poincar polynomials.
2.5. Let K be the simplicial complex with V (K ) = Z and S1 (K ) = {{r, r +1} | r
Z} (|K | R). Then S1 (K ) is a 1-cocycle. Find all the cochains a C 0 (K )
such that S1 (K ) = (a).
2.6. Find a simplicial pair (K , L) such that |K | is homeomorphic to S 1 I and
|L| = Bd |K |. In the spirit of Sect. 2.4.7, compute the simplicial cohomology
of K and of (K , L) and find (co)cycles generating H (K ), H (K , L), H (K )
and H (K , L). Write completely the (co)homology sequence of (K , L).
2.7. Same exercise as before with |K | the Mbius band and |L| = Bd |K |.
2.8. Let f : K L be a simplicial map between simplicial complexes. Suppose
that L is connected and K is non-empty. Show that H0 f is surjective.
2.9. Let m, n, q be positive integers. If m = nq, the quotient map Z Z/nZ
descends to a map Z/mZ Z/nZ, giving rise to a simplicial map f : Pm
Pn between the simplicial polygons Pm and Pn (see Example 2.4.3). Compute
H f .
2.10. Let M be an n-dimensional pseudomanifold. Let and  be two distinct
n-simplexes of M. Find a C n1 (M) such that (a) = {,  }.
2.11. Let M be a finite non-empty n-dimensional pseudomanifold. Let Z n1 (M)
which is a boundary. Prove that is the boundary of exactly two n chains.
2.12. Let f : M N be a simplicial map between finite n-dimensional pseudo-
manifolds. Show that the following two conditions are equivalent.
(a) Hn f = 0.
(b) There exists S(N ) such that  f 1 ({}) is odd.
2.13. Let {1} be the 0-dimensional simplicial complex with vertices 1 and 1. Let
K be a simplicial complex. The simplicial suspension  K is the join K {1}.
(a) Let P4 be the polygon complex with 4-edges (see Example 2.4.3). Show
that P4 K is isomorphic to the double suspension ( K ). [Hint: show
that the join operation is associative: (K L) M K (L M).]
(b) Prove that the suspension of a pseudomanifold is a pseudomanifold.
(c) Prove that the correspondence K   K gives a functor from Simp to
itself.
2.14. Let A be a finite set. Show that F A is a pseudomanifold.
2.15. Let M be an n-dimensional pseudomanifold which is infinite. What is Hn (M)?
2.16. Let (K , K 1 , K 2 ) be a simplicial triad. Suppose that K 1 and K 2 are connected
and that K 1 K 2 is not empty. Show that K is connected.
2.17. Let (K , K 1 , K 2 ) be a simplicial triad and let K 0 = K 1 K 2 .
2.11 Exercises for Chapter 2 57

(a) Prove that the homomorphism H (K 1 , K 0 ) H (K , K 2 ) induced by the

inclusion is an isomorphism (simplicial excision).
(b) Write the commutative diagram involving the homology sequences of
(K 1 , K 0 ) and (K , K 2 ). Using (a), construct out of this diagram the Mayer-
Vietoris sequence for the triad (K , K 1 , K 2 ).
2.18. Deduce the additivity formula for the Euler characteristic of Lemma 2.4.10
from the Mayer-Vietoris sequence.
2.19. Let M1 and M2 be two finite n-dimensional pseudomanifolds. Let i S(Mi )
and let h: 1 2 be a bijection. The simplicial connected sum M = M1  M2
(using h) is the simplicial complex defined by

V (M) = V (M1 ) V (M2 ) {v h(v) for v 1 }

and

S(M) = S(M1 ) {1 } S(M2 ) {2 } .

Prove that M is a pseudomanifold. Compute H (M) in terms of H (M1 ) and

H (M2 ).
Chapter 3
Singular and Cellular (Co)homologies

3.1 Singular (Co)homology

Singular (co)homology provides a functor associating to a topological space X a

graded Z2 -vector space, whose isomorphism class depends only on the homotopy
type of X . Such functors, from Top to categories of algebraic objects, constitute the
main subject of algebraic topology.
Invented by Eilenberg in 1944 [49] after earlier attempts by Lefschetz, singular
homology is formally akin to simplicial homology. However, in order to make
computations for non-trivial examples, we need to establish some properties, such as
homotopy and excision, which require some work. When K is a simplicial complex,
the simplicial homology of K and the singular cohomology of |K | are isomorphic in
several ways, some of them being functorial (see Sect. 3.6). Singular (co)homology
is especially powerful and relevant for spaces having the homotopy type of a
CW-complex, a notion introduced in Sect. 3.4. For such spaces, singular (co)homology
is isomorphic to other (co)homology theories (see Sect. 3.7) and the cohomol-
ogy functor H n is representable by the Eilenberg-MacLane space K (Z2 , n) (see
Sect. 3.8).

3.1.1 Definitions

The standard Euclidean m-simplex m is defined by


m = {(x0 , . . . , xm ) Rm+1 | xi 0 and xi = 1},

endowed with the induced topology from that of Rn+1 . In particular, m = if

m < 0. Let X be a topological space. A singular m-simplex of X is a continuous
map : m X . The set of singular m-simplexes of X is denoted by Sm (X ).

Springer International Publishing Switzerland 2014 59

J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_3
60 3 Singular and Cellular (Co)homologies

The definitions of singular (co)chains and (co)homology are copied from those for
the simplicial case (see Sect. 2.2), replacing simplicial simplexes by singular ones.
We thus set

Definition 3.1.1 (subset definitions)

(a) A singular m-cochain of X is a subset of Sm (X ).
(b) A singular m-chain of X is a finite subset of Sm (X ).
The set of singular m-cochains of X is denoted by C m (X ) and that of singular m-
chains by Cm (X ). As in Sect. 2.2, Definition 3.1.1 are equivalent to

Definition 3.1.2 (colouring definitions)

(a) A singular m-cochain is a function a : Sm (X ) Z2 .
(b) A singular m-chain is a function : Sm (X ) Z2 with finite support.
Definition 3.1.2 endow C m (X ) and Cm (X ) with a structure of a Z2 -vector space.
The singletons provide a basis of Cm (X ), in bijection with Sm (X ). Thus, Definition
3.1.2b is equivalent to

Definition 3.1.3 Cm (X ) is the Z2 -vector space with basis Sm (X ):


Cm (X ) = Z2 .
Sm (X )

We consider the graded vector spaces C (X ) = mN Cm (X ) and C (X ) =

mN C m (X ). By convention, C m (X ) = Cm (X ) = 0 if m < 0 (so the index m
could be taken in Z in the previous formulae).
The Kronecker pairing on singular (co)chains

,
C m (X ) Cm (X ) Z2

is defined, using the various above definitions, by the equivalent formulae

a, = 
(a )(mod 2) using Definition 3.1.1a and b
=  a() using Definition 3.1.1a and 3.1.2b (3.1.1)
= Sm (X ) a()() using Definition 3.1.2a and b.

As in Lemma 2.2.4, we check that the map k : C m (X ) Cm (X ) , given by

k(a) = a, , is an isomorphism.
Let m, i N with 0 i m. Define the i-th face inclusion i : m1 m by

i (x0 , . . . , xm1 ) = (x0 , . . . , xi1 , 0, xi+1 , . . . , xm1 ).

3.1 Singular (Co)homology 61

The boundary operator : Cm (X ) Cm1 (X ) is the Z2 -linear map defined,

for Sm (X ) by


m
() = i . (3.1.2)
i=0

Lemma 3.1.4 = 0.

Proof By linearity, it suffices to prove that () = 0 for Sm (X ). One has

m 
() = ( i ) = i  j , (3.1.3)
i=0 (i, j)A

where A = {0, . . . , m} {0, . . . , m 1}. The set B = {(i, j) A | i j} is

in bijection with A B, via the map (i, j) ( j + 1, i). But if (i, j) B, then
i  j =  j+1 i , which implies that = 0. 

The coboundary operator : C m (X ) C m+1 (X ) is defined by the equation

a, = a, . (3.1.4)

With these definition, ((C (X ), ), (C (X ), ),  , ) is a Kronecker pair. We

define the vector spaces of singular cyclesZ (X ), singular boundaries B (X ), sin-
gular cocyclesZ (X ), singular coboundaries B (X ), singular homologyH (X ) and
singular cohomologyH (X ) as in Sect. 2.3. By Proposition 2.3.5, the pairing on
(co)chain descends to a pairing

,
H m (X ) Hm (X ) Z2


so that the map k : H m Hm , given by k(a) = a, , is an isomorphism (Kronecker
duality in singular (co)homology). The Kronecker pairing extends to a bilinear map
,
H (X ) H (X ) Z2 by setting a, = 0 if a H p (X ) and Hq (X ) with
p = q.

Example 3.1.5 If X is the empty space, then Sm (X ) = for all m and thus H () =
H () = 0. Let X = pt be a point. Then, Sm ( pt) contains one element for each
m N, namely the constant singular simplex m pt. Then, Cm ( pt) = Z2 for
all m N and the chain complex looks like

0 0 0

C2k+1 ( pt)
C2k ( pt)
C2k1 ( pt)

C1 ( pt)
C0 ( pt) 0.
62 3 Singular and Cellular (Co)homologies

Therefore,

0 if > 0
H ( pt) H ( pt) (3.1.5)
Z2 if = 0.

Example 3.1.6 Let K be a simplicial complex. Choose a simplicial order for

K . To an m-simplex = {v0 , . . . , vm } Sm (K ), with v0 vm , we associate
the singular m-simplex R () : m |K | defined by


m
R ()(t0 , . . . , tm ) = ti vi . (3.1.6)
i=0

The linear combination in (3.1.6) makes sense since {v0 , . . . , vm } is a simplex of K .

This defines a map R : Sm (K ) Sm (|K |) which extends to a linear map

R, : C (K ) C (|K |).

This map will be used several times in this chapter. The formula R, = R,
is obvious, so R, is a chain map from (C (K ), ) to (C (|K |), ). We shall
prove, in Theorem 3.6.3, that R, induces an isomorphism between the simplicial
(co)homology) of K and the singular (co)homology) of |K |.
Example 3.1.7 As the affine simplex 0 is a point, one can identify a singular
0-simplex of X with its image, a point of X . This gives a bijection S0 (X ) X
and a bijection between subsets of X and singular 0-cochains. For B X and
x X , one has B, x = B (x), where B stands for the characteristic function for
B. The 1-cochain B is the connecting cochain for B: if S1 (X ), then

(B), = B, = B, (1, 0) + B, (0, 1) . (3.1.7)

In other words (B), = 1 if and only if the (non-oriented) path connects a

point in B to a point in X B. Observe that (B) = (X B).
Following Example 3.1.7, the unit cochain 1 C 0 (X ) is defined by 1 = S0 (X )
X . By Eq. (3.1.7) 1, = 0 for all S1 (X ). This proves that (1) = 0 by
Lemma 2.2.4. Hence, 1 is a cocycle, whose cohomology class is again denoted by
1 H 0 (X ).
Proposition 3.1.8 Let X be a non-empty path-connected space. Then,
(i) H 0 (X ) = Z2 , generated by 1 which is the only non-vanishing singular 0-
cocycle.
(ii) H0 (X ) = Z2 . Any 0-chain is a cycle, which represents the non-zero element
of H0 (X ) if and only if  is odd.
Proof The proof is analogous to that of Proposition 2.4.1. If X is non-empty the unit
cochain does not vanish and, as C 1 (X ) = 0, 1 = 0 in H 0 (X ).
3.1 Singular (Co)homology 63

Let a C 0 (X ) with a = 0, 1. Then there exists x, y X = S(X ) with a(x) =

a(y). Since X is path-connected, there exists S1 (X ) with (1, 0) = x and
(0, 1) = y. As in Eq. (3.1.7), this proves that (a), = 0 so a is not a cocycle.
This proves (i).
Now, H0 (X ) = Z2 since H 0 (X ) H0 (X ) . Any C0 (X ) is a cycle since
C1 (X ) = 0. It represents the non-zero homology class if and only if 1, = 1,
that is if and only if  is odd. 

The reduced (singular) cohomology H (X ) and homology H (X ) of a topological

space X are the graded Z2 -vector spaces defined by
 
H (X ) = coker
 H p : H ( pt) H (X )
 (3.1.8)
H (X ) = ker H p : H (X ) H ( pt)

where p : X pt denotes the constant map to a point. In particular, H ( pt) =

0 = H ( pt). One checks that the Kronecker pairing induces a bilinear map
 , : H m (X ) H m (X ) Z2 such that the correspondence a a, gives

an isomorphism k : H m (X ) H m (X ) .
The full strength of Definition (3.1.8) appears in other (co)homology theories, such
as equivariant cohomology (see p. 266). For the singular cohomology, as H ( pt) =
Z2 1, one gets

H 0 (X )/Z2 1 if m = 0
H m (X ) =
H m (X ) if m = 0

and
  1, 
ker H0 (X ) Z2 if m = 0
Hm (X ) =
Hm (X ) if m = 0.

Thus, by Proposition 3.1.8, H 0 (X ) = 0 = H 0 (X ) if X is path-connected (see also

Corollary 3.1.12).
Let f : Y X be a continuous map between topological spaces. It induces
a map S f : S(Y ) S(X ) defined by S f () = f . The linear map C f :
C (X ) C (Y ) is, using Definition 3.1.2, defined by C f (a) = a S( f ). As for
C f : C (Y ) C (X ), it is the linear map extending S f , using Definition 3.1.3.
One checks that the couple (C f, C f ) is a morphism of Kronecker pair. It thus
defines linear maps of degree zero H f : H (X ) H (Y ) and H f : H (Y )
H (X ). The functorial properties are easy to prove: H and H are functors from
the category Top of topological spaces to the category GrV of graded vector spaces
(see Proposition 3.1.22 for a more general statement). Also, for any map f : Y X ,
the diagram
64 3 Singular and Cellular (Co)homologies

f
Y A / X
AA }}
A }}
p AA
(3.1.9)
~}} p
pt

is obviously commutative. This implies that the reduced cohomology H and homol-
ogy H are also functors from Top to GrV. The notations H f , H f , H f and H f
are sometimes shortened in f and f .
As in Lemma 2.5.4, we prove the following
Lemma 3.1.9 Let f : Y X be a continuous map. Then H 0 f (1) = 1.
Lemma 3.1.9 implies the following result.

Lemma 3.1.10 Let (X, Y ) be a topological pair with X path-connected. Denote by

i : Y X the inclusion. Then there are exact sequences

0 H 0 (X ) H 0 (Y ) H 0 (Y ) 0
H i

and
H i
0 H 0 (Y ) H0 (Y ) H0 (X ) 0.

We now prove some general results useful to compute the (co)homology of a

space. Let X be a topological space which is a disjoint union:


X= X j.
jJ

By this we mean that the above equality holds as sets and that each X j is open
(and therefore closed) in X . Denote the inclusion by i j : X j X . The equality

Sm (X ) = jJ i j (Sm (X j )) implies the following proposition.

Proposition 3.1.11 The family of inclusions i j : X j X for j J gives rise to

isomorphisms
(H i j ) 
H (X ) /
jJ H (X j )

and

 H i j
jJ H (X j )
/ H (X ).

Corollary 3.1.12 Let X be a topological space which is locally path-connected.

Then, the family of inclusions i Y : Y X for Y 0 (X ) gives rise to isomorphisms
3.1 Singular (Co)homology 65

(H i Y ) 
H (X ) / Y 0 (X ) H (Y )

and

 H i Y
Y 0 (K ) H (Y ) / H (X ) .

Proof As X is locally path-connected, each Y 0 (X ) is open in X and X is

topologically the disjoint union of its path-connected components. Corollary 3.1.12
then follows from Proposition 3.1.11. 

Corollary 3.1.13 Let X be a topological space which is locally path-connected.

Then,

H 0 (X ) = 0 H 0 (X ) = 0 X is path-connected.

Also,

H 0 (X ) = H 0 (X ) Z2 and H0 (X ) = H 0 (X ) Z2

if X is not empty.

In the same spirit of reducing the computations of H (X ) to those of smaller

subspaces, another consequence of the definition of the singular (co)homology is the
following proposition.

Proposition 3.1.14 Let X be a topological space. Let K be the set of compact sub-
spaces of X , partially ordered by inclusion. Then, the natural homomorphisms

J : lim

H (K )
H (X )
K K

and

J : H (X )
lim

H (K )
K K

are isomorphisms.

Here, lim denotes the direct limit (also called inductive limit or colimit) and lim

denotes the inverse limit (also called projective limit or just limit) in GrV.

Proof Let A Hr (X ), represented

by Z r (X ). Then, is a finite set of r -
simplexes of X and K = (r ) is a compact subspace of X . One can see
Z r (K ), so J is onto. Now, let K be a compact subspace of X and A Hr (K )
mapped to 0 under Hr (K ) Hr (X ). Represent A by Z r (K ) and let
66 3 Singular and Cellular (Co)homologies

Cr +1 (X ) with = (). As before, there exists a compact subset L of X containing

K with Cr +1 (L), so A is mapped to 0 under Hr (K ) Hr (L). This proves that
J is injective. Finally, the bijectivity of J is deduced from that of J by Kronecker
duality. 
Remark 3.1.15 In Proposition 3.1.14 the morphism J is an isomorphism for the
homology with any coefficients. The morphism J is always surjective but, in general
not injective (except for coefficients in a field, like Z2 ). Its kernel is expressible using
the derived functor lim1 (see e.g. [82, Theorem 3F.8]). The same considerations hold

true for the following corollary.
Corollary 3.1.16 Let X be a topological space and let A be a family of subspaces
of X , partially ordered by the inclusion. Suppose that each compact subspace of X
is contained in some A A. Then, the homomorphisms

j : lim

H (A)
H (X )
AA

and

j : H (X )
lim

H (A)
AA

are isomorphisms.
Proof The hypothesis that each compact K X is contained in some A A implies
a factorization of the homomorphism J of Proposition 3.1.14:

lim H (K )

J
/ H (X )
K K ;
JJJ
j vvv
vv
JJJ
JJ vv
% vv
lim

H (A)
AA

The same hypothesis implies that is onto, whence j is an isomorphism. The

assertion for j comes from Kronecker duality. 

3.1.2 Relative Singular (Co)homology

A (topological) pair is a couple (X, Y ) where X is a topological space and Y is a

subspace of X . The inclusion i : Y  X is a continuous map. Let a C m (X ).
If, using Definition 3.1.1a, we consider a as a subset of Sm (X ), then C i(a) =
a Sm (Y ). If we see a as a map a : Sm (X ) Z2 , then C i(a) is the restriction of
a to Sm (Y ). We see that C i : C (X ) C (Y ) is surjective. Define
3.1 Singular (Co)homology 67

 C i 
C m (X, Y ) = ker C m (X ) C m (Y )

and C (X, Y ) = mN C m (X, Y ). As C i is a morphism of cochain complexes, the

coboundary : C (X ) C (X ) preserves C (X, Y ) and gives rise to a coboundary
: C (X, Y ) C (X, Y ) so that (C (X, Y ), ) is a cochain complex. The cocycles
Z (X, Y ) and the coboundaries B (X, Y ) are defined as usual, giving rise to the
definition

H m (X, Y ) = Z m (X, Y )/B m (X, Y ).

The graded Z2 -vector space H (X, Y ) = mN H m (X, Y ) is the relative (singular)

cohomology of the pair (X, Y ). Observe that H (X, ) = H (X ). We denote by j
the inclusion j : C (X, Y )  C (X ), which is a morphism of cochain complexes,
and use the same notation j for the induced linear map j : H (X, Y ) H (X )
on cohomology. We also use the notation i for both C i and H i. We get thus a
short exact sequence of cochain complexes

j i
0 C (X, Y )
C (X )
C (Y ) 0. (3.1.10)

If a C m (Y ), any cochain a C m (X ) with i (a)

= a is called a extension
of a as a singular cochain in X . For instance, the 0-extension of a is defined by
a = a Sm (Y ) Sm (X ).
With chains, the inclusion Y  X induces an inclusion i : C (Y )  C (X ) of
chain complexes. We define Cm (X, Y ) as the quotient vector space
 
Cm (X, Y ) = coker i : Cm (Y )  Cm (X ) .

As i is a morphism of chain complexes, C (X, Y ) = mN Cm (X, Y ) inherits

a boundary operator = X,Y : C (X, Y ) C1 (X, Y ). The projection j :
C (X ) C (X, Y ) is a morphism of chain complexes and one obtains a short
exact sequence of chain complexes

i j
0 C (Y )
C (X )
C (X, Y ) 0. (3.1.11)

The cycles and boundaries Z (X, Y ) and B (X, Y ) are defined as usual, giving rise
to the definition

Hm (X, Y ) = Z m (X, Y )/Bm (X, Y ).

The graded Z2 -vector space H (X, Y ) = mN Hm (X, Y ) is the relative (singular)

homology of the pair (X, Y ). Observe that H (X, ) = H (X ). The notations i and
j are also used for the induced maps in homology.
68 3 Singular and Cellular (Co)homologies

As in Sects. 2.6 and 2.7 of Chap. 2, one gets a pairing  , : H m (X, Y )

Hm (X, Y ) Z2 which makes (H m (X, Y ), Hm (X, Y ),  , ) a Kronecker pair. Also,
the singular (co)homology connecting homomorphisms

: H (Y ) H +1 (X, Y ). and : H (X, Y ) H1 (Y )

are defined and satisfy  (a), = a, () . The proof of the following lemma
is the same as that of Lemma 2.7.1.

Lemma 3.1.17 Let a Z m (Y ) and let a C m (X ) be any extension of a as a

singular m-cochain of X . Then, X (a)
is a singular (m + 1)-cocycle of (X, Y )
representing (a).

Remark 3.1.18 A class in A Hm (X, Y ) is represented by a relative singular cycle,

i.e. a singular chain Cm (X ) such that () is a singular chain (cycle) of Y . The
homology class of () in Hn1 (Y ) is (A). This is the Kronecker dual statement
of Lemma 3.1.17.

As for the simplicial (co)homology (see Sect. 2.7), the results of Sect. 2.6 give
the following (singular) (co)homology exact sequences, or just the (co)homology
sequence, of the pair (X, Y ).

Proposition 3.1.19 ((Co)homology exact sequences of a pair) Let (X, Y ) be a topo-

logical pair. Then, the sequences

j i j
H m (X, Y )
H m (X )
H m (Y )
H m+1 (X, Y )

and
i j i
Hm (Y )
Hm (X )
Hm (X, Y )
Hm1 (Y )

are exact.

These exact sequences are also available for reduced (co)homology. For this,
the reduced (co)homology of a pair is defined as follows: when Y = , then
H (X, Y ) = H (X, Y ) and H (X, Y ) = H (X, Y ); otherwise H (X, ) = H (X )
and H (X, ) = H (X ).

Proposition 3.1.20 (Reduced (co)homology sequences of a pair) The exact

sequences of Proposition 3.1.19 hold with reduced (co)homology.

Proof An argument is only required around m = 0. For the homology exact

sequence, consider the commutative diagram:
3.1 Singular (Co)homology 69

i j
/ H (X, Y ) / H (Y ) / H (X ) / H (X, Y ) / 0
1 0 0 0
 
= =
   
j
/ H1 (X, Y ) / H0 (Y ) i
/ H0 (X ) / H0 (X, Y ) / 0

1, 1,
 
= / Z2
Z2

The commutativity of the bottom square is due to Lemma 3.1.9. As i = 0,

1, () = 0 for all H1 (X, Y ) and therefore : H 1 (X, Y ) H 1 (Y ) exists.
Since the sequence of the second line is exact, an easy diagram-chase shows that the
sequence of the first line is exact as well.
The reduced cohomology sequence can be established in an analogous way or
deduced from the homology one by Kronecker duality, using Lemma 2.3.8. 
Remark 3.1.21 Let (X, Y ) be a topological pair with Y path-connected and non-
empty. By Proposition 3.1.20 and its proof, we get the isomorphisms

j : H 0 (X )
H0 (X, Y ) and H 0 (X ).
j : H 0 (X, Y ) (3.1.12)

Also, if Y = {x}, we get the isomorphisms

j : H (X )
H (X, x) and H (X ).
j : H (X, x) (3.1.13)

A direct proof of (3.1.13), say for cohomology, is given by the diagram.

H ( pt)
JJJ
p JJJ
JJ
j

%
H (X, x) / H (X )
i / H (x) (3.1.14)
LLL
LLL
LL 
%
H (X )

where the line and the column are exact and p : X pt is the constant map onto
a point. We see that the choice of x X produces a supplementary vector subspace
to p (H ( pt)) in H (X ).
We now study the naturality of the relative (co)homology and of the exact
sequences. Let (X, Y ) and (X  , Y  ) be topological pairs. A map f of (topological)
pairs from (X, Y ) to (X  , Y  ) is a continuous map f : X X  such f (Y ) Y  .
With these maps, topological pairs constitute a category Top2 . The correspondence
X (X, ) makes Top a full subcategory of Top2 .
70 3 Singular and Cellular (Co)homologies

Let f : (X, Y ) (X  , Y  ) be a map of topological pairs. The morphism

C f : C (X  ) C (X ) then restricts to a morphism of cochain complexes C f :
C (X  , Y  ) C (X, Y ) and the morphism C f : C (X ) C (X  ) descends to a

morphism of chain complexes C f : C (X, Y ) C (X  , Y  ). As in Sect. 2.7, we

prove that (C f, C f ) is a morphism of Kronecker pair. One then gets degree zero
linear maps H f : H (X  , Y  ) H (X, Y ) and H f : H (X, Y ) H (X  , Y  )
satisfying H a, = a, H for all a H m (X  , Y  ), Hm (X, Y ) and all
m N. Functorial properties are easy, so we get the following
Proposition 3.1.22 The relative singular cohomology H ( , ) is a contravariant
functor from the category Top2 to the category GrV of graded Z2 -vector spaces. The
relative singular homology H ( , ) is a covariant functor between these categories.
The same holds true for the reduced singular (co)homology.
As for Proposition 2.7.6, we can prove the following
Proposition 3.1.23 The (co)homology sequences are natural with respect to maps
of topological pairs.
Here is a special form of the cohomology sequence of a pair.
Proposition 3.1.24 Let A and B be topological spaces. Then the cohomology

sequence of the pair (AB, A) cuts into short exact sequences and there is a com-
mutative diagram

0 / H (B) / H (AB)
/ H (A) / 0

id id . (3.1.15)
  
0 / H (AB,
A) / H (AB)
/ H (A) / 0

Proof If i A : A AB and i B : B AB
denote the inclusions, Proposi-
tion 3.1.11 provides a commutative diagram

(i A ,i B )

H (AB) / H (A) H (B)
LLL
oo
LLL
ooooo .
i A LL& wooo proj1
H (A)

This proves that i a is surjective, which cuts the cohomology sequence of (AB, A),
giving the bottom line of (3.1.15). Also, ker i A is the image of H (B) under the
monomorphism j : H (B) H AB) given by j (u) = (i A , i B )1 (0, u), which
we placed in the top line of (3.1.15). 
As in simplicial (co)homology, the exact sequences of a pair generalize to that of
a triple. A (topological) triple is a triplet (X, Y, Z ) where X is a topological spaces
3.1 Singular (Co)homology 71

and Y , Z are subspaces of X with Z Y . A map f of triples, from (X, Y, Z ) to

(X  , Y  , Z  ) is a continuous map f : X X  such that f (Y ) Y  and f (Z ) Z  .
A triple T = (X, Y, Z ) gives rise to pair inclusions

i j
(Y, Z )
(X, Z )
(X, Y )

and to a commutative diagram

j X,Y i X,Y
0 / C (X, Y ) / C (X ) / C (Y ) / 0
(3.1.16)
C j = id i Y,Z
  
j X,Z i X,Z
0 / C (X, Z ) / C (X ) / C (Z ) / 0

where the horizontal lines are exact sequences of cochain complexes as in (2.7.9),
we get a short exact sequence of cochain complexes

C j C i
0 C (X, Y ) C (X, Z ) C (Y, Z ) 0. (3.1.17)

The same arguments with the chain complexes gives a short exact sequence

C i C j
0 C (Y, Z ) C (X, Z ) C (X, Y ) 0. (3.1.18)

As in Sect. 2.7 of Chap. 2, short exact sequences (3.1.17) and (3.1.18) produces
connecting homomorphisms T : H (Y, Z ) H +1 (X, Y ) and T : H (X, Y )
C1 (Y, Z ). They satisfy T (a), = a, T () as well as following proposition.

Proposition 3.1.25 ((Co)homology sequences of a triple) Let T = (X, Y, Z ) be a

triple. Then,
(a) the sequences

H j H i T H j
H m (X, Y ) H m (X, Z ) H m (Y, Z ) H m+1 (X, Y )

and
H i H j T H i
Hm (Y, Z ) Hm (X, Z ) Hm (X, Y ) Hm1 (Y, Z )

are exact.
(b) the exact sequences of Point (a) are natural for maps of triples.

Remark 3.1.26 As H () = 0, we get a canonical GrV-isomorphisms H (X, )
H (X ), etc. Thus, the (co)homology sequences for the triple (X, Y, ) give back those
of the pair (X, Y )
72 3 Singular and Cellular (Co)homologies

H j H i H j
H m (X, Y ) H m (X ) H m (Y )
H m+1 (X, Y )
(3.1.19)
and
H i H j H i
Hm (Y ) Hm (X ) H m (X, Y )
Hm1 (Y ) (3.1.20)

where i : Y X and j : (X, ) (X, Y ) denote the inclusions. This gives a more
precise description of the morphisms j and j of Proposition 3.1.19.

We now draw a few consequences of Proposition 3.1.19. A topological pair (X, Y )

is of finite (co)homology type if its singular homology (or, equivalently, cohomology)
is of finite type. In this case, the Poincar series of (X, Y ) is that of H (X, Y ):
 
Pt (X, Y ) = dim Hi (X, Y ) t i = dim H i (X, Y ) t i N[[t]].
iN iN

Corollary 3.1.27 Let (X, Y, Z ) be a topological triple. Suppose that two of the pairs
(X, Y ), (Y, Z ) and (X, Z ) are of finite cohomology type. Then, the third pair is of
finite cohomology type and there is Q t N[[t]] such that the equality

Pt (X, Y ) + Pt (Y, Z ) = Pt (X, Z ) + (1 + t) Q t , (3.1.21)

holds in N[[t]].

Proof This follows from the cohomology sequence of T = (X, Y, Z ) and elementary
linear algebra. If Tk : H k (Y, Z ) H k+1 (X, Y ) denotes the connecting homomor-
phism, one checks that (3.1.21) holds true for

Qt = t k codim Tk . 

Corollary 3.1.27 implies straightforwardly the following result.

Corollary 3.1.28 Let (X, Y, Z ) be a topological triple. Suppose that dim H (Y, Z )
< and that dim H (X, Y ) < . Then dim H (X, Z ) < and

dim H (X, Z ) dim H (X, Y ) + dim H (Y, Z ).

Corollary 3.1.16 has the following generalization with relative (co)homology.

Proposition 3.1.29 Let (X, Y ) be a topological pair. Let A be family of subspaces

of X , partially ordered by inclusion. Suppose that each compact subspace of X is
contained in some A A. Then, the natural homomorphisms

J : lim

H (A, A Y )
H (X, Y )
AA
3.1 Singular (Co)homology 73

and

J : H (X, Y )
lim

H (A, A Y )
AA

are isomorphisms.
Proof By Kronecker duality, only the bijectivity of J must be proven. Let Hr (Y ) =
Hr (AY ), Hr (X ) = lim Hr (A), and Hr (X, Y ) = lim Hr (A, AY ).
lim AA AA AA
For each A A, one has the homology sequence of the pair (A, AY ). By naturality
of these exact sequences under inclusions, one gets the diagram:

/ Hr (X ) / Hr (X, Y ) / Hr 1 (Y ) / Hr 1 (X )
Hr (Y )
J
    
/ Hr (X ) / Hr (X, Y ) / Hr 1 (Y ) / Hr 1 (X )
Hr (Y )

The top horizontal line is exact because the direct limit of exact sequences is exact.
The bijectivity of the vertical arrows comes from Corollary 3.1.16. By the five-lemma,
one deduces that J is an isomorphism. 

3.1.3 The Homotopy Property

Let f, g : (X, Y ) (X  , Y  ) be two maps between topological pairs. Let I = [0, 1].
A homotopy between f and g is a map of pairs F : (X I, Y I ) (X  , Y  ) such
that F(x, 0) = f (x) and F(x, 1) = g(x). If such a homotopy exists, we say that f
and g are homotopic.
Proposition 3.1.30 (Homotopy property) Let f, g : (X, Y ) (X  , Y  ) be two maps
between topological pairs which are homotopic. Then H f = H g and H f = H g.
Proof Note that H f = H g implies H f = H g by Kronecker duality, using
Diagram (2.3.4). We shall construct a Z2 -linear map D : C (X ) C+1 (X  ) such
that

C f + C g = D + D , (3.1.22)

i.e. D is a chain homotopy from C f to C g. The map D will satisfy D(C (Y ))

C+1 (Y  ) and will so induce a linear map D : C (X, Y ) C+1 (X  , Y  ) satisfy-
ing (3.1.22). As in the proof of Proposition 2.5.9, this will prove that H f = H g.
That H f = H g is then deduced by Kronecker duality, using Diagram (2.3.4). Let
F : (X I, Y I ) (X  , Y  ) be a homotopy from f to g.
By linearity, it is enough to define D on singular simplexes. Let : m X be
a singular m-simplex of X . Consider the convex-cell complex P = m I . One has
74 3 Singular and Cellular (Co)homologies

V (P) = V (m ) {0, 1}. Using the natural total order on V (m ), we can define an
affine order on P by deciding that the elements of V (m ){1} are greater than those

of V (m ) {0}. Lemma 2.1.10 thus provides a triangulation h : |L (P)| P,
with V (L (P)) = V (P). Set L = L (P) and h = h . The order becomes a
simplicial order on L, giving rise to a chain map R, : C (L) C (P) from the
simplicial chains of L to the singular chains of P (see Example 3.1.6). Consider
Sm+1 (L) as an (m + 1)-simplicial cochain of L and define D() to be the image of
Sm+1 (L) under the composed map

R, C h C (id) C F
Sm+1 (L) Sm+1 (|L|) Sm+1 (P) Sm+1 (X I ) Sm+1 (X  ).
(3.1.23)
Observe that, if Sm (L) such that h(| |) hits the interior of P, then is the face
of exactly two (m + 1)-simplexes of L. Therefore, (Sm+1 (L)) = Sm (L(Bd P)).
But
Bd P = m {0} m {1} Bd m I.

As all the maps in (3.1.23) are chain maps, this permits us to prove that

D() = C f () + C g() + D () (3.1.24)

As (3.1.24) holds true for all S(X ), it implies (3.1.22). 

Remark 3.1.31 In the proof of Proposition 3.1.30, the chain homotopy D is not
unique. Some authors (e.g. [43, 155, 179]) just give an existence proof, based on an
easy case of the acyclic carriers technique (like in our proof of Proposition 2.5.9).
We used above an explicit triangulation of m I . The same triangulation occurs
the proof of [82, p. 112], presented differently for the sake of signs control. The idea
of such triangulations of m I will be used again in the proof of the small simplex
theorem 3.1.34.

A map of pairs f : (X, Y ) (X  , Y  ) is a homotopy equivalence if there exists a

map of pairs g : (X  , Y  ) (X, Y ) such that g f is homotopic to id(X,Y ) and f g
is homotopic to id(X  ,Y  ) . The pairs (X, Y ) and (X  , Y  ) are then called homotopy
equivalent. Two spaces X and X  are homotopy equivalent if the pairs (X, ) and
(X  , ) are homotopy equivalent. Two homotopy equivalent spaces (or pairs) are also
said to have the same homotopy type.
By functoriality (Proposition 3.1.22), Proposition 3.1.30 implies that (co)homo-
logy is an invariant of homotopy type:

Corollary 3.1.32 (Homotopy invariance of (co)homology) Let f : (X, Y )

(X  , Y  ) be a homotopy equivalence. Then H f : H (X, Y ) H (X  , Y  ) and
H f : H (X  , Y  ) H (X, Y ) are isomorphisms.

A (non-empty) topological space X is contractible if there exists a homotopy from

id X to a constant map. For instance, the cone C X over a space X
3.1 Singular (Co)homology 75
   
CX = X I X {1} , (3.1.25)

with the quotient topology, is contractible. A homotopy from idCX to a constant map
is given by F((x, ), t) = [x, t + (1 t) ].

Corollary 3.1.33 The (co)homology of a contractible space is isomorphic to that of

a point:

0 if > 0
H (X ) H (X )
Z2 if = 0.

Proof Let x0 X such that there exists a homotopy from id X to the constant map onto
x0 . Then, the inclusion {x0 } X is a homotopy equivalence and Corollary 3.1.33
follows from Corollary 3.1.32. 

For a direct proof of Corollary 3.1.33, see Exercise 3.2.

3.1.4 Excision

Let X be a topological space. Let B be a family of subspaces of X . A map f : L X

is called B-small if f (L) is contained in an element of B. Let SmB (X ) be the set of
singular m-simplexes of X which are B-small. The vector spaces of (co)chains CBm (X )
and CmB (X ) are defined as in Sect. 3.1, using B-small m-simplexes. We get, in the
same way, a pairing  , : CBm (X ) CmB (X ) Z2 identifying CBm (X ) to CmB (X ) .
The boundary of a B-small simplex is a B-small chain, so (CB (X ), ) is a subcomplex
of chains of (C (X ), ), the inclusion being denoted by i B : CBm (X ) C m (X ).
Define : CBm (X ) CBm+1 (X ) by (a), = a, () and i B C m (X ) CBm (X )
by i B (a), = a, i B () . Then, ((CB (X ), ), (CB (X ), ),  , ) is a Kronecker
pair.
The (co)homologies obtained by these definitions are denoted by HB (X ) and
HB (X ). One uses the notations i B : HmB (X ) Hm (X ) and i B : H m (X ) HBm (X )
for the induced linear maps. The following result is very useful.

Proposition 3.1.34 (Small simplexes theorem) Let X be a topological space with a

family B of subspaces of X , whose interiors cover X . Then i B : HB (X ) H (X )
and i B : H (X ) HB (X ) are isomorphisms.

The proof of Proposition 3.1.34 uses iterations of the subdivision operator, a chain
map sd : C (X, Y ) C (X, Y ) which replaces chains by chains with smaller
simplexes. Intuitively, sd replaces a singular simplex : m X by the sum of
restricted to the barycentric subdivision of m .
More precisely, consider the standard simplex m as the geometric realization
of the full complex Fm over the set {0, 1, . . . , m}. The barycentric subdivision Fm
76 3 Singular and Cellular (Co)homologies

is endowed with its natural simplicial order of (2.1.2), p. 11. As explained in

Example 3.1.6, we get a chain map

R = R, : C (Fm ) C (|Fm |) = C (m ).

Let Sm (X ). As a continuous map from m to X , induces C : C (m )

C (X ). Define

sd () = C (Sm (Fm )).

This formula determines a unique linear map sd : C (X ) C (X ) which is clearly

a chain map. If Y is a subspace of X , then sd (C (Y )) C (X ), so we get a chain
map sd : C (X, Y ) C (X, Y ), giving rise to a GrV-morphism

sd : H (X, Y ) H (X, Y ).

By Kronecker duality, we get a cochain map sd : C (X, Y ) C (X, Y )

and a GrV-morphism sd : H (X, Y ) H (X, Y ) satisfying sd (a), =
a, sd () for all a C (X, Y ) and C (X, Y ).
Observe that sd sends CB (X, Y ) into CB (X, Y ) and thus HB (X, Y ) into
B
H (X, Y ) and HB (X, Y ) into HB (X, Y )

Lemma 3.1.35 The subdivision operators sd : H (X, Y ) H (X, Y ) and sd :

H (X, Y ) H (X, Y ) are equal to the identity. The same holds true for sd :
HB (X, Y ) HB (X, Y ) and sd : HB (X, Y ) HB (X, Y ).

Proof We shall construct a Z2 -linear map D : C (X ) C+1 (X ) such that

id + sd = D + D . (3.1.26)

In other words, D is a chain homotopy from id to sd (see p. 34). The map D will
satisfy D(C (Y )) C+1 (Y ) and will so induce a linear map D : C (X, Y )
C+1 (X, Y ) satisfying (3.1.26). As in the proof of Proposition 2.5.9, this will
prove that sd = id. That sd = id is then implied by Kronecker duality, using
Diagram (2.3.4). Also, the map D will satisfy D(CB (X )) C+1 B (X ).
By linearity, it is enough to define D on singular simplexes. The proof is similar
to that of Proposition 3.1.30 (an idea of V. Puppe). Let : m X be a singular
m-simplex of X . Consider the convex-cell complex P = m I , where the upper
face m {1} is replaced by its barycentric subdivision |Fm |. One has V (P) =
V (m ) {0} V (Fm ) {1}. We use the natural total order on V (m ) {0} and the
natural simplicial order on V (Fm ) {1} (see (2.1.2)). Deciding in addition that the
elements of V (Fm ) {1} are greater than those of V (m ) {0} provides an affine

order on P. Lemma 2.1.10 thus constructs a triangulation h : |L (P)|
P,
with V (L (P)) = V (P). Set L = L (P) and h = h . Seeing as a simplicial
order on L gives rise to a chain map R, : C (L) C (P) from the simplicial
3.1 Singular (Co)homology 77

chains of L to the singular chains of P (see Example 3.1.6). Consider Sm+1 (L) as
an (m + 1)-simplicial cochain of L and define D() to be the image of Sm+1 (L)
under the composed map

R, C h C ()

Sm+1 (L) Sm+1 (|L|) Sm+1 (P) Sm+1 (X ), (3.1.27)

where : P X is the map (x, t) = (x). Observe that the inclusion |Fm |
 {1} is already the piecewise affine triangulation of m {1} determined by the
m

simplicial order on Fm . Therefore, the construction of the proof of Lemma 2.1.10
leaves m {1} unchanged. Formula (3.1.26) is then deduced as in the proof of
Proposition 3.1.30. Finally, if is B-small, so is the map . B (X ),
Hence D() Cm+1
which proves the lemma for the B-small (co)homology. 
Proof of Proposition 3.1.34 By Kronecker duality, using Corollary 2.3.11, only
the homology statement must be proved. Let sd k = sd sd (k times). We shall
need the following statement.
Claim let C (X ). Then, there exists k() N such that sd k () CB (X )
for all k k().
Let us show that the claim implies Proposition 3.1.34. Let Hm (X, Y ) rep-
resented by Cm (X ) with () Cm1 (Y ). The claim implies that, for k big
enough, sd k () CmB (X ), and thus () Cm1 (Y ). This implies that sd k () is
in the image of i B . By Lemma 3.1.35, sd k () = , so is in the image of i B , which
proves that i B is surjective. For the injectivity, let HmB (X, Y ) with i B () = 0.
Represent by CmB (X ) with () Cm1 (Y ). The hypothesis i B () = 0 says
that = () + with Cm+1 (X ) and Cm (Y ). The claim tells us that,
for k big enough, sd k () Cm+1B (X ) (and, so, sd k () C B (Y )). This implies that
m
sd ()
k = 0 in H (X, Y ). But sd k ()
B C B (X ) and Lemma 3.1.35 tells us that sd k
m m
coincides with the identity of HmB (X, Y ). Thus, = 0 for all ker i B .
It remains to prove the claim. Let (m, k) be the maximal distance between two
points of a simplex of the k-th barycentric subdivision of m . An elementary argu-
ment of Euclidean geometry shows that
 k
m
(m, k) (m, 0) (3.1.28)
m+1

(of course, (m, 0) = 2). For details, see e.g., [155, Proof of Theorem 15.4] or
[82, p. 120]. By hypothesis, the family B = {int B | B B} is an open covering
of X . Consider the induced open covering 1 B of m . By (3.1.28), (m, k) 0
when k . Using a Lebesgue number for the open covering 1 B, this proves
the claim.
The main application of the small simplexes theorem is the invariance under
excision (see also Sect. 3.1.6).
Proposition 3.1.36 (Excision property) Let (X, Y ) be a topological pair. Let U be
a subspace of X with U int Y . Then, the linear maps induced by inclusions
78 3 Singular and Cellular (Co)homologies

i : H (X, Y )
H (X U, Y U ) and i : H (X U, Y U )
H (X, Y )

are isomorphisms.

Proof By Corollary 2.3.11, i is an isomorphism if and only if i is an isomorphism.

We shall prove that i is an isomorphism.
Let B = {Y, X U }. One has a commutative diagram

0 / C (Y ) / C B (X ) / C B (X )/C (Y ) / 0

= id i B IB
  
0 / C (Y ) / C (X ) / C (X )/C (Y ) / 0

where all arrows are induced by inclusions and the horizontal lines are short exact
sequences of chain complexes. As in Sect. 2.6, this gives a commutative diagram
between the corresponding long homology sequences

... / Hm (Y ) / HmB (X ) / Hm (CB (X )/C (Y )) / Hm1 (Y ) / ...

= id i B IB = id
   
... / Hm (Y ) / Hm (X ) / Hm (X, Y ) / Hm1 (Y ) / ...

As U int Y , the family B = {Y, X U } satisfy the hypotheses of Proposition 3.1.34

and i B is an isomorphism. By the five-lemma, IB is an isomorphism. Therefore, it
suffices to show that H (X U, Y U ) H (CB (X )/C (Y )) is an isomorphism.
But it is easy to see that this is already the case at the chain level:

CB (X )/C (Y ).
C (X U, Y U ) = C (X U )/C (Y U ) 

3.1.5 Well Cofibrant Pairs

Let (Z , Y ) be a topological pair and denote by i : Y Z the inclusion. A (con-

tinuous) map r : Z Y is called a retraction if r i = idY . It is a retraction by
deformation if i r is homotopic to the identity of Z . A retraction by deformation is
thus a homotopy equivalence.
Note that Z retracts by deformation on Y if and only if there is a homotopy
h : Z I Z which, for all (z, t) Z I , satisfies h(z, 0) = z, h(z, 1) Y and
h(y, t) = y when y Y . A topological pair (X, A) is called good if A is closed in
X and if there is a neighbourhood V of A which retracts by deformation onto A. For
instance, (X, ) is a good pair (V = and h(x, t) = x).
3.1 Singular (Co)homology 79

Good pairs were introduced in [82] (with the additional condition that A is non-
empty). Earlier books rather rely on the notion of cofibration, developed in the 1960s
essentially by Puppe and Steenrod (see [185] for references). Both are useful in
different circumstances, so we introduce below the mixed notion of a well cofibrant
pair, especially useful in equivariant homotopy theory (see e.g. Chap. 7). We begin
by cofibrant pairs, starting with the following lemma.

Lemma 3.1.37 For a topological pair (X, A), the following conditions are equiva-
lent.
(1) There is a retraction from X I onto X {0} A I .
(2) Let f : X Z and FA : A I Z be continuous maps such that FA (a, 0) =
f (a). Then, FA extends to a continuous map F : X I Z such that F(x, 0) =
f (x) for all x X .

Proof We give below the easier proof available when A is closed in X (for a proof
without this hypothesis: see [39, (1.19)]). Let r : X I X {0} A I be a
retraction. Given f and FA as in (2), define the map g : X {0} A I Z by
g(x, 0) = f (x) and g(a, t) = FA (a, t). If A is closed, then g is continuous and the
map F = g r satisfies the required condition. Hence, (1) implies (2). Conversely, if
f and FA are the inclusions of X and A I into Z = X {0} A I , the extension
F given by (2) is a continuous retraction from X I onto Z . 

A pair (X, A) with A closed in X which satisfies (1) or (2) of Lemma 3.1.37 is
called cofibrant. According to the literature, the inclusion A  X is a cofibration,
or satisfies the absolute homotopy extension property (AHEP) (see e.g. [44, 73]). See
e.g. [38, Chap. 5] for other characterizations and properties of cofibrant pairs.
As a motivation of our concept of well cofibrant pair, we first give an example.

Example 3.1.38 Mapping cylinder neighbourhoods. Let (X, A) be a topological

pair. A neighbourhood V of A is called a mapping cylinder neighbourhood if there is
a continuous map : V A (where V is the frontier of V ) and a homeomorphism
: M V where

M = [(V I ) A] {(x, 0) (x) | x V }

is the mapping cylinder of . The homeomorphism is required to satisfy (x, 1) =

x and (x, 0) = (x) for all x V . Here are examples of mapping cylinder
neighbourhoods
if X is a smooth manifold and A is a smooth submanifold of codimension 1,
then a closed tubular neighbourhood of A [95, Sect. 4.6] is a mapping cylinder
neighbourhood.
if A is the boundary of a smooth manifold X , then a collar neighbourhood of A
[95, Sect. 4.6] is a mapping cylinder neighbourhood.
a subcomplex of a CW-complex admits a mapping cylinder neighbourhood. The
proof of this will be given in Lemma 3.4.2.
80 3 Singular and Cellular (Co)homologies

Given a mapping cylinder neighbourhood as above, a continuous retraction

F : X I X {0} A I is defined by


((v), t (1 2 )) if x = (v, ) with 1/2.
F(x, t) = ((v, 2 1), 0) if x = (v, ) with 1/2.


(x, 0) if x X int V.

Let u : X I defined by

2 if x = (v, ) with 1/2.
u(x) =
1 otherwise.

Let h : X I X defined by h = p X F, where p X : X I X is the

projection. Then, u(h(x)) u(x) which implies that, for all T < 1, h restricts to a
strong deformation retraction form u 1 ([0, T ]) onto A = u 1 (0). Hence, (X, A) is
a good and cofibrant pair.

A topological pair (X, A) is well cofibrant if there exists continuous maps

u : X I and h : X I X such that
(1) A = u 1 (0) (in particular, A is closed in X ).
(2) h(x, 0) = x for all x X .
(3) h(a, t) = a for all (a, t) A I .
(4) h(x, 1) A for all x X such that u(x) < 1.
(5) u(h(x, t)) u(x) for all (x, t) X I .
We say that (u, h) is a presentation of (X, A) as a well cofibrant pair. Conditions (1
4) define a NDR-pair (neighbourhood deformation retract pair) in the sense of [38,
140, 185] (see also Remarks 3.1.42 (b)).
The pairs (X, ) and (X, X ) are well cofibrant. One takes u(x) = 1 for (X, ),
u(x) = 0 for (X, X ) and h(x, t) = x for both pairs. Another basic example of well
cofibrant pairs is given by the following lemma.

Lemma 3.1.39 Suppose that A X admits a mapping cylinder neighbourhood in

X . Then, (X, A) is well cofibrant.

Proof The pair (u, h) in Example 3.1.38 is a presentation of (X, A) as a well cofibrant
pair. 

Lemma 3.1.40 Let (X, A) and (Y, B) be two well cofibrant pairs. Then, the prod-
uct pair (X Y, A Y X B) is well cofibrant.

The following proof, coming from that of [185, Theorem 6.3], will be convenient
for the equivariant setting (see Lemma 7.2.12).
3.1 Singular (Co)homology 81

Proof Let (u, h) and (v, j) be presentations of (X, A) and (Y, B) as well cofibrant
pairs. Define w : X Y I by w(x, y) = u(x)v(y). Define q : X Y I X Y
by

(x,


y)

if (x, y) A B.
q(x, y, t) = h(x, t), j (y, u(x)
v(y) t) if v(y) u(x) and v(y) > 0.

 v(y) 
h(x, u(x) t), j (y, t) if v(y) u(x) and u(x) > 0.

One checks that (w, q) is a presentation of (X Y, A Y X B) as a well cofibrant

pair. Details for (1)(4) are given in [185, p. 144] and (5) is obvious. 
Lemma 3.1.41 Let (X, A) be a well cofibrant pair. Then, (X, A) is good and cofi-
brant.
Proof Let (u, h) be a presentation of (X, A) as a well cofibrant pair. As noticed
in Example 3.1.38, the condition u(h(x, t)) u(x) implies that, for all T < 1,
h restricts to a strong deformation retraction form u 1 ([0, T ]) onto A. Since A =
u 1 (0), it is closed. Hence, (X, A) is good. To see that (X, A) is cofibrant, let
(Y, B) = (I, {0}) presented as well cofibrant pair by (v, j) where v(y) = y/2 and
j (y, t) = (1 t)y. Let (w, q) be the presentation of

(X Y, A Y X B) = (X I, X {0} A I )

as a well cofibrant pair given in the proof of Lemma 3.1.40. As

w(x, y) = u(x)y/2 < 1,

the formula r (x, y) = q(x, y, 1) defines a retraction

r
XI
X {0} A I. (3.1.29)

By Lemma 3.1.37, (X, A) is cofibrant. 

Remarks 3.1.42 (a) The fact that the retraction r of (3.1.29) is a strong deformation
retraction should not be a surprise. If r = (r1 , r2 ) : X I X {0} A I X I
is any retraction, then the map R : X I X I defined by
 
R(x, t, s) = r1 (x, (1 s)t), st + (1 s)r2 (x, t)

is a homotopy from id X I to r [73, Lemma 16.28].

(b) The proof of Lemma 3.1.41 shows that a NDR-pair is cofibrant. The converse is
also true (see [140, Sect. 6.4]).
If (X, A) is a topological pair, we denote by X/A the quotient space where all
points of A are identified in a single class. The projection : (X, A) (X/A, A/A)
is a map of pairs.
82 3 Singular and Cellular (Co)homologies

Lemma 3.1.43 Let (X, A) be a well cofibrant pair and let B A. Then (X/B, A/B)
is well cofibrant. In particular, the pair (X/A, A/A) is well cofibrant.

Proof Let (u, h) be a presentation of (X, A) as a well cofibrant pair. By (1) and (3),
u and h descend to continuous maps u : X/B I and h : (X/B) I X/B,
giving a presentation (u, of (X/B, A/B).
h) 

Lemma 3.1.44 Let (X, A) be a cofibrant pair such that A is contractible. Then the
quotient map X X/A is a homotopy equivalence.

Proof If A is contractible, there is a continuous map FA : A I A X such that

FA (a, 0) = a and F(A 1) = {a0 }. As (X, A) is cofibrant, there is a continuous
map F : X I X such that F(x, 0) = x and F(a, t) = FA (a, t) for a A.
F|X 1 admits a factorization

F|X 1
X TTTT / X {1} / X
=
TTT II {
TTT I
TTT IIII {{{
TTT I$ $ {
q
TT) ) {{ g
X/A

Using F, g q is homotopic to id X . On the other hand, as F(A I ) A, the map

q F descends to a continuous map F : X/A I X/A which is a homotopy
from id X/A to q g. 

Proposition 3.1.45 Let (X, A) be a well cofibrant pair. Then, the homomorphisms

: H (X/A, A/A)
H (X, A) and : H (X, A)
H (X/A, A/A)

are isomorphisms.

Proof By Corollary 2.3.11, is an isomorphism if and only if is an isomorphism.

We shall prove that is an isomorphism (the proof is the same for both). There is
nothing to prove if A = , so we assume that A is not empty.
By Lemma 3.1.41, (X,  A)  is cofibrant. Let r be a retraction from X I to X
{0} A I . Let C A = A I A {1} be the cone over A and let X = X A C A.
As A is closed, r extends (by the identity on C A I ) to a continuous retraction from
X I onto X {0} C A I . Hence, the pair ( X , C A) is cofibrant. But

/
H ( X , C A) excision/ H ( X [A {1}], C A [A {1}]) H (X,
homotopy
A).

On the other hand, X/A = X /C A. Set X = X /C A and C = C A/C A { pt}. The

quotient map q : ( X , C A) ( X , C)
provides a morphism of exact sequences
3.1 Singular (Co)homology 83

H k1 ( X ) / H k1 (C)
/ H k ( X , C)
/ H k ( X ) / H k (C)

q q q q q .
    
H k1 ( X ) / H k1 (C A) / H k ( X , C A) / H k ( X ) / H k (C A)

As C and C A are contractible, q : H (C)

H (C A) is an isomorphism and so
is q : H ( X ) H ( X ) by Lemma 3.1.44. By the five lemma, q : H k ( X , C)

H k ( X , C A) is an isomorphism, which proves Proposition 3.1.45. 

Remark 3.1.46 The proof of Proposition 3.1.45 uses only that the pair (X, A) is
cofibrant. Another proof exists using that (X, A) is a good pair (see [82, Proposi-
tion 2.22] or Proposition 7.2.15). It is interesting to note that these relatively short
proofs both use almost all the axioms of a cohomology theory (see Sect. 3.9): func-
toriality, homotopy, excision and functorial exactness.

Corollary 3.1.47 Let (X, A) be a well cofibrant pair with A non-empty. Then,

: H (X/A) H (X/A)
H (X, A) and : H (X, A)

are isomorphisms.

Proof If A = , then A/A is a point. Therefore, by (3.1.13), H (X/A)

H (X/A). The results then follows form
H (X/A, A/A) and H (X/A, A/A)
Proposition 3.1.45. 

Corollary 3.1.48 Let (X, A) be a well cofibrant pair. Denote by i : A X the

inclusion and by j : X X/A the quotient map. Then, there is a functorial exact
sequence in reduced cohomology

H j
H k1 (X ) H k1 (A)
H k (X/A) H k (X ) H k (A)
H i H i

The corresponding sequence exists in reduced homology.

Proof The result is obvious if A is empty. Otherwise, this comes from the exact
sequence of Proposition 3.1.20 together with the isomorphism H (X/A)

H (X, A) of Corollary 3.1.47.
H (X, A) 

One application of well cofibrant pairs is the suspension isomorphism. Let X be

a topological space. The suspension  X of X is the quotient space
 
 X = CX X {0}

where CX is the cone over X (see (3.1.25)). The pairs (CX, X 0) and ( X, X 21 )
are well cofibrant by Lemma 3.1.39, since the subspaces admits mapping cylinder
84 3 Singular and Cellular (Co)homologies

neighbourhoods. The (cohomology) suspension homomorphism is the degree-1 linear

map  : H m (X ) H m+1 ( X ) given by the composition

j
 : H m (X )
H m+1 (CX, X 0) H m+1 ( X, [X 0])
H m+1 ( X ),
(3.1.30)
where the middle isomorphism comes from Proposition 3.1.45.

Proposition 3.1.49 For any topological space X , the suspension homomorphism

 : H m (X ) H m+1 ( X )

is an isomorphism for all m.

Because of Proposition 3.1.49, the homomorphism  is called the suspension

isomorphism (in cohomology). Observe that Formula (3.1.30) can also be used to
define  : H m (X ) H m+1 ( X ). By Proposition 3.1.49, this unreduced suspen-
sion homomorphism is an isomorphism if m 1.

Proof If X is empty, so is  X and Proposition 3.1.49 is trivial. Suppose then that

X = . As [X 0] is a point, its reduced cohomology vanish and j is an iso-
morphism in Formula (3.1.30), by the reduced cohomology sequence of the pair
( X, [X 0]). As CX is contractible, its reduced cohomology also vanish and
is an isomorphism in Formula (3.1.30), by the reduced cohomology sequence of the
pair (CX, X 0). Finally, note that  X is path-connected, so H m ( X ) = 0 for
m 0, so Proposition 3.1.49 is also true for m 0. 

Analogously, we define  : H m+1 ( X ) H m ( X ) by the composition

j
 : H m+1 ( X ) H m+1 ( X, [X 0]) H m+1 (CX, X 0) H m (X )
(3.1.31)
which satisfies  (a), = a,  () for all a H m (X ) and all H m+1 (X ).
By Proposition 3.1.49 (or directly), we deduce that  is an isomorphism, called the
suspension isomorphism (in homology).
Let (X j , x j ) (
j J ) be a family of pointed spaces, i.e. xi X i . Their bouquet
(or wedge) X = jJ X j is defined as the quotient space

 

X= Xj = Xj {x j }.
jJ jJ
jJ

By naming x X the equivalence class x = jJ {x j }, the couple (X, x) is a

pointed space. For each j J , one has a pointed inclusion i j : (X j , x j ) (X, x).
The bouquet plays the role of a sum in the category of pointed spaces and pointed
maps: if f j : (X j , x j ) (Y, y) are continuous pointed maps, then there is a unique
continuous pointed map f : (X, x) (Y, y) such that f i j = f j .
3.1 Singular (Co)homology 85

A well pointed space is a pointed space (X, x) such that (X, {x}) is a well cofibrant
pair. Observe that this definition is stronger than that in other textbooks.

Lemma 3.1.50 If (X j , x j ) ( j J ) are well pointed spaces, then their wedge (X, x)
is a well pointed space.

Proof Let (u j , h j ) be a presentation of (X j , x j ) as a well cofibrant pair. Then

( jJ u j , jJ h j ) is a presentation of ( jJ X j , jJ {x j }) as a well cofibrant

pair. By Lemma 3.1.43, the quotient pair (X, {x}) is well cofibrant, so (X, x) is a
well pointed space. 

Proposition 3.1.51 Let (X j , x j ), with j J , be

 a family of well pointed spaces.
Then, the family of inclusions i j : X j X = jJ X j , for j J , gives rise to
isomorphisms on reduced (co)homology

( H i j ) 
H (X ) / H (X j )
jJ

and

 H i j
/ H (X ).
jJ H (X j )


Proof It is enough to establish that H i j is an isomorphism. The cohomology
statement can be proved analogously or by Kronecker duality, using Diagram (2.3.4).

Write, as above, x = jJ {xi } X . The map of pairs (X j , {xi }) (X, x) give

rise, for each m N, to a commutative diagram between exact sequences


jJ Hm+1 ({x j })
/  jJ Hm+1 (X j ) /  jJ Hm+1 (X j , {x j })
/

H i j
  





Hm+1 ( jJ {x j }) / H m ( jJ X j ) / Hm+1 ( jJ X j , jJ {x j }) /

/  jJ Hm ({x j }) /  jJ Hm (X j )

 

/ Hm (

jJ {x j })
/ Hm (

jJ X j )

The isomorphisms
 for the vertical arrows are due to Proposition 3.1.11. By
the five-
lemma, H i j is an isomorphism. As (X j , x j ) is well pointed, the pair ( jJ X j ,

jJ {x j }) is well cofibrant. By Proposition 3.1.45, the quotient map q : ( jJ X j ,

86 3 Singular and Cellular (Co)homologies

jJ {x j }) (X, x) induces an isomorphism on homology. One has the commu-

tative diagram

 H i j
/ H (X )
jJ H (X j )

  

/ H (

H i j H q
(X , {x }) / H (X, {x}) ,
jJ X j , jJ {x j })
jJ H j j


where the vertical arrows are isomorphisms by Remark 3.1.21. Therefore, H i j
is an isomorphism. 

The (co)homology of X = jJ X j may somehow be also controlled using the
projection j : X X j defined by

z if z X j
j (z) = (3.1.32)
x otherwise,

where x = jJ {xi } X . As j i j = id X j , Proposition 3.1.51 implies the

following

Proposition 3.1.52 Let (X j , x j ), with j J , be a family of well pointed spaces.

Then, the composition

 H j ( H i j ) 
/ H (X ) / H (X j )
jJ H (X j ) jJ

is the inclusion of the direct sum into the product. Also, the composition

 H i j ( H j ) 
/ H (X ) / H (X j )
jJ H (X j ) jJ

is the inclusion of the direct sum into the product.

In particular, if J is finite, then

 H j
H (X j ) / H (X ) and

( H j ) 
H (X ) / H (X j )
jJ

are isomorphisms.
3.1 Singular (Co)homology 87

3.1.6 Mayer-Vietoris Sequences

Let X be a topological space. Let B = {X 1 , X 2 } be a collection of two subspaces of

X . Write X 0 = X 1 X 2 and

X0
i1
/ X1

i2 j1
 j2

X2 / X

for the inclusions. We call (X, X 1 , X 2 , X 0 ) a Mayer-Vietoris data. The sequence of

cochain complexes

(C j1 ,C j2 ) C i 1 +C i 2
0 C (X ) C (X 1 ) C (X 2 ) CB (X 0 ) 0

is then exact, as well as the sequence of chain complexes

(C i 1 ,C i 2 ) C j1 +C j2
0 C (X 0 ) C (X 1 ) C (X 2 ) CB (X ) 0

By Sect. 2.6, these short exact sequences give rise to connecting homomorphisms

M V : H (X 0 ) HB+1 (X ) and M V : HB (X ) H1 (X 0 )

involved in long (co)homology exact sequences. If the interiors of X 1 and X 2 cover

X , the theorem of small simplexes (3.1.34) implies that HB (X ) H (X ) and
HB (X ) H (X ). Therefore, we obtain the following proposition.

Proposition 3.1.53 (Mayer-Vietoris sequences I) Let (X, X 1 , X 2 , X 0 ) be a Mayer-

Vietoris data. Suppose that X = int X 1 int X 2 . Then, the long sequences

(H j1 ,H j2 ) H i 1 +H i 2 M V
H m (X ) H m (X 1 ) H m (X 2 ) H m (X 0 ) H m+1 (X )

and
(H i 1 ,H i 2 ) H j1 +H j2 M V
Hm (X 0 ) Hm (X 1 ) Hm (X 2 ) H m (X ) Hm1 (X 0 )

are exact.

These Mayer-Vietoris sequences are natural for maps f : X X  such that

f (X i ) X i .
The hypotheses of Proposition 3.1.53 may not by directly satisfied in usual situ-
ations. Here is a variant which is more useful in practice.
88 3 Singular and Cellular (Co)homologies

Proposition 3.1.54 (Mayer-Vietoris sequences II). Let (X, X 1 , X 2 , X 0 ) be a Mayer-

Vietoris data, with X i closed in X . Suppose that X = X 1 X 2 and that (X i , X 0 ) is
a good pair for i = 1, 2. Then, the long sequences

(H j1 ,H j2 ) H i 1 +H i 2 M V
H m (X ) H m (X 1 ) H m (X 2 ) H m (X 0 ) H m+1 (X )

and
(H i 1 ,H i 2 ) H j1 +H j2 M V
Hm (X 0 ) Hm (X 1 ) Hm (X 2 ) H m (X ) Hm1 (X 0 )

are exact.

Proof Choose a neighbourhood Ui of X 0 in X i admitting a retraction by deformation

onto X 0 , called it : Ui Ui , t I . Let X 1 = X 1 U2 , X 2 = X 2 U1 and
X 0 = X 1 X 2 = X 0 U1 U2 . We claim that X = int X 1 int X 2 . Indeed,
as U2 is a neighbourhood of X 0 in X 2 , there exists an open set V2 of X such that
X 0 V2 X 2 U2 . As X 2 is closed, X 1 X 2 = X X 2 is open in X . Therefore

X 1 (X 1 X 2 ) (V2 X 2 ) = (X 1 X 2 ) V2 int X 1 .

In the same way, X 2 int X 2 . Hence, X = int X 1 int X 2 .

As X = int X 1 int X 2 , the Mayer-Vietoris sequences of Proposition 3.1.53 hold
true with (X 1 , X 2 , X 0 ). But X 1 retracts by deformation onto X 1 , using the retraction
1t : X 1 X 1 given by

2t (x) if x U2
1t =
x if x X 1 .

In the same way, X 2 retracts by deformation onto X 2 and X 0 retracts by deformation

onto X 0 . This proves Proposition 3.1.54. 

For Mayer-Vietoris sequences with other hypotheses, see Exercise 3.11, from
which Proposition 3.1.54 may also be deduced.

3.2 Spheres, Disks, Degree

So far, we have not encountered any space whose (co)homology is not zero in positive
dimensions. The unit sphere S n in Rn+1 will be the first example. The shortest way
to describe the (co)homology of such simple spaces is by giving their Poincar
polynomials. The definitions are the same as for simplicial complexes. A topological
pair (X, Y ) is of finite (co)homology type if its singular homology (or, equivalently,
cohomology) is of finite type. In this case, the Poincar series of (X, Y ) (or of X if
Y is empty) is that of H (X, Y ):
3.2 Spheres, Disks, Degree 89
 
Pt (X, Y ) = dim Hi (X, Y ) t i = dim H i (X, Y ) t i N[[t]].
iN iN

When the series is a polynomial, we speak of the Poincar polynomial of (X, Y ).

Proposition 3.2.1 The Poincar polynomial of the sphere S n is

Pt (S n ) = 1 + t n .

Proof The sphere S 0 consists of two points, so the result for n = 0 follows
from (3.1.5) and Corollary 3.1.12. We can then propagate the result by the suspen-

sion isomorphism  : H (S n )
H (S n+1 ) (see Proposition 3.1.49), since S n+1
is homeomorphic to  S n . The homology statement uses the homology suspension

isomorphism  : H (S n+1 )
H (S n ). 

As a consequence of Proposition 3.2.1, the sphere S n is not contractible, though

it is path-connected if n > 0. Also S n and S p are not homotopy equivalent if n = p.
A useful corollary of Proposition 3.2.1 is the following

Corollary 3.2.2 Pt (D n , S n1 ) = t n .

Proof This follows from Proposition 3.2.1 and the (co)homology exact sequence of
the pair (D n , S n1 ). 

It will be useful to have explicit cycles for the generators of Hn (D n , S n1 ) and

Hn (S n ). Let  n = n int n be the topological boundary of the standard simplex
 . The identity map i n : n n is a relative cycle of (n , 
n n ), representing a
class [i n ] Hn ( , 
n ). The boundary (i n ) belongs to Z m1 (
n n ) and represents
).
([i n ]) in Hn1 ( n

Proposition 3.2.3 For all n N, the following two statements hold true:
n ) = Z2 .
An : [i n ] is the non-zero element of Hn (n , 

Bn : [(i n+1 )] is the non-zero element of Hn ( n+1 ) = Z2 .

Proof Statements An and Bn are proven together, by induction on n, as follows:

(a) A0 and B0 are true.
(b) An implies Bn .
(c) Bn implies An+1 .
As the affine simplex 0 is a point and  0 is empty, Statement A(0) follows from
the discussion in Example 3.1.5. To prove B0 , observe that  1 consists of two points
p and q. Identifying a singular 0-simplex with a point, one has (i 1 ) = p + q, which
represents a non-vanishing element of H0 ( 1 ). But 1, p + q = 0, which shows

that [(i 1 )] = 0 in H0 ( ).
1
90 3 Singular and Cellular (Co)homologies

Let us prove (b). Consider the inclusion  : n   n+1 given by (t0 , . . . , tn ) =

(t0 , . . . , tn , 0). Let n = adh ( n+1 ( )). Consider the homomorphisms:
n

j H 
H n (
n+1 ) / Hn (
n+1 , n ) o n ).
Hn (n , 

The arrow j is bijective, as in (3.1.13), since m is contractible; the arrow H  is

bijective by excision and homotopy. As in H ( n+1 , n ) we neglect the singular
chains in n , one has

j ([(i n+1 )]) = H  ([i n ])

which proves (b).

To prove (c), we use that : H n+1 (n+1 ,  H n (
n+1 ) n ) is an isomorphism,
since  n+1 is contractible, and that ([i n+1 ]) = [(i n+1 )]. 

For the sphere S 1 , Proposition 3.2.3 has the following corollary.

Corollary 3.2.4 Let : 1 S 1 given by (t, 1t) = e2it . Then C1 (S 1 ) is a

singular 1-cycle of S 1 and its homology class is the non-zero element of H1 (S 1 ) = Z2 .

Proof Since (1, 0) = (0, 1), the 1-cochain is a cycle. The map factors as

/ S1
1 F
FF p xx<
FF s xx
FF x
# xx
1
1 /

where s is a homeomorphism. Under the composed homomorphism

H p
/ H1 (1 / / H1 (1 ) H s
/ H1 (S 1 ) ,
1)
H1 (1 ,  1 , [
1 ])

the class [i 1 ] goes to []. By Proposition 3.2.3, [i 1 ] is a generator of H1 (S 1 ), which

proves Corollary 3.2.4. 

Let f : S n S n be a continuous map. The linear map H n f : H n (S n ) H n (S n )

is a map between Z2 and itself. The degree deg( f ) Z2 of f by

0 if H n f = 0
deg( f ) =
1 otherwise.

One can define the same degree using H n f . For instance, the degree of a home-
omorphism is 1 and the degree of a constant map is 0. Let f, g : S n S n . By
3.2 Spheres, Disks, Degree 91

Proposition 3.1.30, deg( f ) = deg(g) if f, g : S n S n are homotopic. Also, using

that Hn (g f ) = Hn g Hn f one gets

deg(g f ) = deg(g) deg( f ) (3.2.1)

These simple remarks have the following surprisingly strong consequences.

(For a refinement of Proposition 3.2.5 below using the integral degree (see [155,
Theorems 21.4 and 21.5]).

Proposition 3.2.5 Let f : S n S n be a continuous map with deg f = 0. Then,

(a) f admits a fixed point.
(b) there exists x S n with f (x) = x.

Proof Suppose that there is no fixed point. Then f is homotopic to the antipodal map
a(x) = x: a homotopy is obtained by following the arc of great circle from f (x)
to x not containing x. Therefore deg f = deg a = 1 since a is a homeomorphism.
If f (x) = x for all x, then deg f = 1 because f is homotopic to the identity
(following the arc of great circle from f (x) to x not containing x). 

We now give three recipes to compute the degree of a map from S n to itself. A point
u S n is a topological regular value for f : S n S n if there is a neighbourhood
U of u such that U is evenly covered by f . By this, we mean that f 1 (U ) is a
disjoint union of U j , indexed by a set J , such that, for each j J , the restriction of
f to U j is a homeomorphism from U j to U . In particular, f 1 (u) is a discrete closed
subset of S n indexed by J , so J is finite since S n is compact. For instance, a point
u which is not in the range of f is a topological regular value of f (with J empty).
For a topological regular value u of f , we define the local degree d( f, u) N of f
at u by
d( f, u) =  f 1 (u).

Proposition 3.2.6 Let f : S n S n be a continuous map. For any topological

regular value u of f , one has

deg( f ) = d( f, u) mod 2.

Example 3.2.7 The map S 1 S 1 given by z z k has degree the residue class of
k mod 2.

Proof of Proposition 3.2.6 When n = 0, each of the two points of S 0 is a regular

value of f and the equality of Proposition 3.2.6 is easy to check by examination of
the various cases. We then suppose that n > 0.
If u is a topological regular value, there is a neighbourhood B of u which is
evenly covered by f and which is homeomorphic to a closed disk D n . Its preimage
B = f 1 (B) is a finite disjoint union of n-disks B j , indexed by j J . Define
J = J {0}
and set B0 = B.
92 3 Singular and Cellular (Co)homologies

For j J , define V j = S n B j and set V = S n B. Consider the quotient

n n
spaces S j = S /V j ( j J ), which are homeomorphic to S n . Thus, S n /V

jJ S j is homeomorphic to a bouquet of J copies of S . Denote the quotient
n n

maps by  j : S n S nj and  : S n S n /V jJ S nj .
If u j U j is the point such that f (u j ) = u (u 0 = u), then S n {u j } is a
neighbourhood of V j which retracts by deformation onto V j . Therefore, (S n , V j ) is
a good pair and, as S n {u j } is homeomorphic to Rn , the space V j is contractible.
Also, V j admits a mapping cylinder neighbourhood in S n , so the pair (S n , V j ) is
well cofibrant by Lemma 3.1.39. By the reduced homology sequence of the pairs
(S n , V j ) and (S nj , [V j ]) and Proposition 3.1.45, we get three isomorphisms in the
commutative diagram

Hn  j
H n (S n ) / H n (S n )
j

 
H n (S n , V j ) / H n (S n , [V j ]) ,
j

which shows that Hn  j is an isomorphism.

Let 0 = Hn (S n ), and, for j J , let 0 = j Hn (S nj ). The map f descends
to a continuous map f : S n /V S n . Let us consider the commutative diagram:
0

Hn  
H n (S n ) / H n (S nj )
jJ

H n f Hn f (3.2.2)
 
H n 0

Hn (S n ) / H n (S n )
0

The restriction of f to S nj is a homeomorphism, so Hn f( j ) = 0 . Let k :


jJ S j Sk be the projection onto the kth component (see Eq. (3.1.32)). Then
n n

 j = j . By Proposition 3.1.52, this implies that H n () = ( j ). Then,

H n f H n () = d( f, u) 0 . On the other hand, H n 0 H n f () = deg( f ) 0 .
As Diagram (3.2.2) is commutative, this proves Proposition 3.2.6.
The second recipe is the following lemma.
Lemma 3.2.8 Let f : S n S n be a continuous map, with n > 0. Let B1 , . . . , Bk
be disjoint n

embedded closed n-disks of S with boundary Bi . Letn V be the closure
of S Bi . Suppose that f sends V onto a single point v S and thus induces
n

continuous maps f i : S n Bi / B i S n . Then


k
deg f = deg f i .
i=1
3.2 Spheres, Disks, Degree 93

Proof Let Sin = Bi / B i , homeomorphic to S n . The map f factors in the following

way:

f
S n MM / n
MMM qqq8 S
MMpM q
MM& qqq
qqq fi
k
S n /V i=1 Sn

k
Obviously, H p([S n ]) = i=1 [Si ].
n Hence,


k

k

deg f [S n ] = H f i ([Sin ]) = deg f i [S n ].
i=1 i=1 
The third recipe concerns the self-maps of S 1 . Recall the elementary way to prove
that [S 1 , S 1 ] Z. Let f : S 1 S 1 be a (continuous) map. As t exp(2it)
is a local homeomorphism R S 1 , there exists a map g : I R such that
f (exp(2it)) = exp(2ig(t)). The integer

DEG ( f ) = g(1) g(0) Z

depends only on the homotopy class of f . This defines a bijection

DEG : [S 1 , S 1 ]
Z. (3.2.3)

For instance, for the map f (z) = z n , one can choose g(t) = nt. Thus, DEG ( f ) = n
if and only if f is homotopic to z z n .
Proposition 3.2.9 For a map f : S 1 S 1 ,

deg( f ) = DEG ( f ) mod 2.

Proof If DEG ( f ) = n, then f is homotopic to z z n . This map satisfies deg( f ) =

n mod 2 by Proposition 3.2.6. 
Remarks 3.2.10 (a) Our degree is the reduction mod 2 of the integral degree obtained
using integral homology (see e.g. [82, Sect. 2.2]). Proposition 3.2.6 would also
hold for the integral degree, provided one takes into account the orientations in the
definition of the local degree.
(b) A continuous map f : S n S n may not have any topological regular value. For
example, S. Ferry constructed a map f : S 3 S 3 with (integral) degree 2 so that
the preimage of every point is connected [62].
(c) Suppose that f = |g|, where g : K L is a simplicial map, with |K | and |L|
homeomorphic to S n . Let Sn (L) and u be a point in the interior of | |. Then u
is a regular value and d( f, u) = d(g, ) (see Eq. (2.5.5)).
94 3 Singular and Cellular (Co)homologies

(d) Let f : S n S n be a smooth map. Any smooth regular value is a topological

regular value. Then, deg( f ) coincides with the degree mod 2 of f as presented in
e.g. [152, Sect. 4].

3.3 Classical Applications of the mod 2 (Co)homology

At our stage of development of homology, textbooks usually present a couple of

classical applications in topology. Several of them only require Z2 -homology, other
need the integral homology. We discuss this matter in this section.
Retractions and Brouwers fixed point Theorem. Let X be a topological space
with a subspace Y . A continuous retraction of X onto Y is a continuous map r : X
Y extending the identity of Y . In other words, r i = idY , where i : Y X denotes
the inclusion. Therefore, Hr H i = id and H i H r = id, which implies the
following lemma.

Lemma 3.3.1 If there exists a continuous retraction from X onto Y , then H i :

H (Y ) H (X ) is injective and H i : H (X ) H (Y ) is surjective. The same
holds true for the reduced (co)homology.

As H (D n ) = 0 while H n1 (S n1 ) = Z2 , Lemma 3.3.1 has the following corol-

lary.

Proposition 3.3.2 There is no continuous retraction of the n-disk D n onto its bound-
ary S n1 .

The most well known corollary of Proposition 3.3.2 is the fixed point theorem
proved by Luitzen Egbertus Jan Brouwer around 1911 (see [40, Chap. 3]).

Corollary 3.3.3 A continuous map from the disk D n to itself has at least one fixed
point.

Proof Suppose that f (x) = x for all x D n . Then a retraction r : D n

S n1 is constructed using the following picture, contradicting Proposition 3.3.2.
r(x)

x
f (x)

Brouwers theorem says that, for a map f : D n D n , the equation f (x) = x

admits a solution under the only hypothesis that f is continuous. Given the possible
3.3 Classical Applications of the mod 2 (Co)homology 95

wildness of a continuous map, this is a very deep theorem. It is impressive that such
a result is due to the fact that Hn (D n ) = 0 and Hn (S n1 ) = Z2 .
Invariance of dimension. An n-dimensional topological manifold is a topological
space such that each point has an open neighbourhood homeomorphic to Rn . The
following result is known as the topological invariance of the dimension and goes
back to the work of Brouwer in 1911 (see [40, Chap. II]).

Theorem 3.3.4 Suppose that a non-empty m-dimensional topological manifold is

homeomorphic to an n-dimensional topological manifold. Then m = n.

Proof Let M be a non-empty m-dimensional topological manifold and N be an

n-dimensional topological manifold. Let h : M N be a homeomorphism. Let
x M. Then h restricts to a homeomorphism from M {x} onto N {h(x)}.
Hence, H h : H (M, M {x}) H (N , N {h(x)}) is an isomorphism. But this
contradicts the fact that
 
Z2 if k = 0, m Z2 if k = 0, n
Hk (M, M {x}) = and Hk (N , N {h(x)}) =
0 otherwise 0 otherwise.
(3.3.1)

Indeed, it enough to prove (3.3.1) in the case of M. As x has a neighbourhood in M

homeomorphic to Rm , it has a neighbourhood B homeomorphic to closed m-ball.
By the excision of M B and homotopy, one has

H (M, M {x}) H (B, B {x}) H (B, Bd B)

and (3.3.1) follows from Corollary 3.2.2. 

Balls and spheres in spheres. The following results concerns the complements
of k-balls or k-spheres in S n .

Proposition 3.3.5 Let h : D k S n be an embedding. Then H (S n h(D k )) = 0.

Proof We follow the classical proof (see e.g. [82, Proposition 2b.1]), which goes by
induction on k. For k = 0, D 0 is a point and S n h(D 0 ) is then contractible. For the
induction step, suppose that H i (S n h(D k )) contains a non-zero element 0 . We use
the homeomorphism D k D k1 I0 with I0 = [0, 1]. Then S n h(D k ) = A B
with A = S n (D k1 [0, 1/2]) and B = S n (D k1 [1/2, 1]). Since, by
induction hypothesis, H (A B) = H (S n h(D k1 {1/2}) = 0, the Mayer-
Vietoris sequence implies that for I1 = [0, 1/2] or I1 = [1/2, 1], the homomorphism
Hi (S n h(D k I0 )) Hi (S n h(D k I1 )) sends 0 to 0 = 1 Hi (S n
(D k1 I1 )). Iterating this process produces a nested sequence I j of closed intervals
converging to a point p I and a non-zero element { j } lim H (X j ) where
X j = S n (D k1 I j ). Set X = S n (D k1 { p}). As each compact subspace of
X is contained in some X j , Corollary 3.1.16 implies that lim H (X j ) is isomorphic

to H (X ), contradicting the induction hypothesis. 
96 3 Singular and Cellular (Co)homologies

Proposition 3.3.6 Let h : S k S n be an embedding with k < n. Then H (S n

h(S k )) H (S nk1 ).

Proof The proof is by induction on k. The sphere S 0 consisting of two points,

S n h(S 0 ) is homotopy equivalent to S n1 . For the induction step, write S k as
the union of two hemispheres D . The Mayer-Vietoris sequence for S n h(D )
together with Proposition 3.3.5 gives the isomorphism H (S n h(S k )) H 1
(S n h(S k1 )). 

The case k = n 1 in Proposition 3.3.6 gives the following corollary.

Corollary 3.3.7 (Generalized Jordan Theorem) Let h : S n1 S n be an embed-

ding. Then S n h(S n1 ) has two path-connected components.

Remarks 3.3.8 (a) Topological arguments show that, in Corollary 3.3.7, h(S n1 ) is
the common frontier of each of the components of its complement (see e.g. [155, The-
orem 6.3]). For a discussion about the possible homotopy types of these components,
see e.g. [155, Sect. 36] or [82, Sect. 2B].
(b) A well known consequence of the generalized Jordan theorem is the invariance
of domain : if U is an open set in Rn , then its image h(U ) under an embedding
h : U Rn is an open set in Rn . This can be deduced from Corollary 3.3.7 by a
purely topological argument (see, e.g. [155, Theorem 36.5] or [82, Theorem 2B.3]).

Unavailable applications. Some applications cannot be obtained using Z2 -

homology. The most well known are the following.
(a) The antipodal map in S 2n is not homotopic to the identity and its consequence,
the non-existence of non-zero vector fields on even-dimensional spheres (see,
e.g. [82, Theorem 2.28] or [155, Corollary 21.6]). This requires the (co)homology
with coefficients in Z or in a field of characteristic = 2.
(b) The determination of [S 1 , S 1 ] and the fundamental theorem of algebra (using
H 1 (S 1 ; Z)) (see, e.g., [155, Exercise 2,Sect. 21] or [82, Theorem 1.8]).

3.4 CW-Complexes

CW-complexes were introduced and developed by J.H.C. Whitehead in the years

19401950 [40, p.221]. The spaces having the homotopy type of a CW-complex
(CW-space) are closed under several natural construction (see [147]). They are the
spaces for which many functors of algebraic topology, like singular (co)homology,
are reasonably efficient.
Let Y be a topological space and let (Z , A) be a topological pair. Let : A Y
be a continuous map. Consider the space

Z Y = Z Y {z = (z) | z A},
3.4 CW-Complexes 97

endowed with the quotient topology. The space Y is naturally embedded into Z Y .
We say that Z Y is obtained from Y by attachment (or adjunction) of Z , using
the attaching map . When (Z , A) is homeomorphic to ( D n ,  S n1 ), where
 is a set (considered as a discrete space), we say that Z is obtained from Y by
attachment of n-cells, indexed by . For , the image of {} int D n in X is
the open cell indexed by .
A CW-structure on the space X is a filtration

= X 1 X 0 X 1 X = Xn , (3.4.1)
nN

such that, for each n, the space X n is homeomorphic to a space obtained from X n1
by attachment of n-cells, indexed by a set n = n (X ). A space endowed with a
CW-structure is a CW-complex. We see n as the set of n-cells of X . The space X n is
called the n-skeleton of X . The topology of X is supposed to be the weak topology:
a subspace A X is open (or closed) if and only if A X n is open (or closed) for
all k N.
If X is a CW-complex, a subspace Y X is a subcomplex of X if Y n = Y X n is
obtained from Y n1 = Y X n1 by attaching n-cells, indexed by n (Y ) n (X )
and using the same attaching maps. For instance, the skeleta of X are subcomplexes
of X . A topological pair (X, Y ) formed by a CW-complex X and a subcomplex Y is
called a CW-pair.
Let X be a CW-complex. With the above definition, the following properties hold
true:
(1) X is a Hausdorff space.
(2) for each n and each k , there exists a continuous map : (D n , S n1 )
(X n , X n1 ) (X, X n1 ) such that its restriction to int D n is an embedding
from int D n into X . Indeed, such a map, called a characteristic map for the
n-cell , may be obtained by choosing a homeomorphism between (X n , X n1 )
and (  D X
n n1 , X n1 ).
(3) a map f : X Z to the topological space Z is continuous if and only if its
restriction to each skeleton is continuous. Also, f is continuous if and only if
f is continuous for any characteristic map and any cell .
(4) each subcomplex of X is a closed subset of X .
(5) X 0 is a discrete space.
(6) A compact subset of a CW-complex meets only finitely many cells. In conse-
quence, a CW-complex is compact if and only if it is finite, i.e. it contains a finite
number of cells.
These properties are easy to prove (see, e.g. [82, pp. 519523]).

Proposition 3.4.1 A CW-pair (X, A) is well cofibrant.

The literature contains many proofs that a CW-pair is good (see e.g.
[82, Proposition A.5] or [64, Proposition 1.3.1], or cofibrant (see e.g. [73, Proposition
98 3 Singular and Cellular (Co)homologies

14.13] or [38, Proposition 8.3.9]). The proof of Proposition 3.4.1 uses the following
lemma.

Lemma 3.4.2 Let Z be a space obtained from a space Y by attaching a collection of

n-cells. Then Y admits in Z a mapping cylinder neighbourhood (see Example 3.1.38).

Proof Let :  D n Y be the attaching map. Let C n = {x D n | |x| 1/2}.

Then, Z contains V = ( C n ) Y as a closed neighbourhood of Y . The reader
will check that V is homeomorphic to the mapping cylinder of . 

Proof of Proposition 3.4.1 Let X n = X n A. By Lemmas 3.4.2 and 3.1.39, the pair
( X , X n1 ) is well cofibrant for all n. Let (v n , g n ) be a presentation of ( X n , X n1 )
n

as a well cofibrant pair. As X 0 is the disjoint union of A with a discrete set, we may
assume that v 0 (x) = 1 if x / A. Let W n be the closure of (v n )1 ([0, 1)). For n 1,
by replacing v by min{2v , 1} if necessary, we may assume that g n restricts to a
n n

strong deformation retraction of W n onto X n1 .

We now define a map u : X I by constructing, inductively on n N, its
restriction to X n , denoted by u n . We set u 0 = v 0 and

min{1, v n (x) + u n1 (g n (x, 1))} if x W n
u n (x) =
1 if x X n int W n .

We check that u n is continuous. If x X n , then u k (x) = u n (x) for

k n, therefore
u is well defined and continuous. The space V = u 1 ([0, 1)) = n V n is a closed
neighbourhood of A in X , where V n = V X n W n .
Define h n : V n I V n by


x if t 1/2n+1
h (x, t) = g n (x, 2n+1 t 1)
n
if 1/2n+1 t 1/2n

n
g (x, 1) if t 1/2n .

Define h nt : V n V n by h nt (x) = h(x, t). If x V n , define h(x, t) =

h 1t h nt (x). Note that h kt (x) = x for k > n so, if x V n V m , then
h 1t h nt (x) = h 1t h m
t (x). Therefore, h : V I V is well defined and
continuous.
The pair (u, h) satisfies all the conditions for being a presentation of (X, A) as
a well cofibrant pair, except that h is only defined on V I instead of X I . To
fix that, choose a continuous map : I I such that ([0, 1/2]) = {1} and

vanish on aneighbourhood  of 1. Let h : X I X and u : X I defined by
h(x, t) = h x, (u(x))t and u(x) = u(x). One checks that (u, is a presentation
h)
of (X, A) as a well cofibrant pair.
Here are classical examples of CW-complexes.

Example 3.4.3 The sphere S n has an obvious CW-structure with one 0-cell and one
n-cell (attached trivially).
3.4 CW-Complexes 99

Example 3.4.4 Observe that the sphere

S n = {x = (x0 , x1 , . . . , xn ) Rn+1 | |x|2 = 1}

n+1
is obtained from S n1 by adjunction of two (n + 1)-cells D , attached by the iden-
tity map of S . Indeed, the embeddings D S given by y = (y1 , . . . , yn )
n1 n n

( 1 |y|2 , y1 , . . . , yn ) extend the inclusion S n1  S n and provide a home-
omorphism between S n1 D n and S n . Starting from S 0 = {1}, we thus get
n
a CW-structure on S with two cells in each dimension and whose k-skeleton is
S k . Taking the inductive limit S of those S n gives a CW-complex known as
the infinite dimensional sphere. This is a contractible space (see e.g. [82, Exam-
ple 1.B.3p. 88]).

Example 3.4.5 The CW-structure on S n of Example 3.4.4 is invariant under the

antipodal map. It then descends to a CW-structure on the projective space RP n =
S n /{x x}, having one cell in each dimension. Its k-th skeleton is RP k and
the (k + 1)-cell is attached to RP k by the projection map S k RP k . This is
called the standard CW-structure on RP n . Taking the inductive limit RP of these
CW-complexes gives a CW-complex known as the infinite dimensional (real) projec-
tive space. Analogous CW-decompositions for complex and quaternionic projective
spaces are given in Sect. 6.1.

Example 3.4.6 If X and Y

are CW-complexes, a CW-structure on
X Y may be
defined, with (X Y )n = p+q=n X p X q and n (X Y ) = p+q=n  p (X )
q (Y ) (see [64, Theorem 2.2.2]). The weak topology may have more open sets
than the product topology so the identity i : (X Y )CW (X Y )prod is only
a continuous bijection. If X or Y is finite, or if both are countable, then i is a
homeomorphism (see [64, p. 60]). These consideration are not important for us
since the two topologies have the same compact sets. Therefore, they have the same
singular simplexes, whence i induces an isomorphism on singular (co)homology.

We now establish a few lemmas useful for the cellular (co)homology. Let X be a
CW-complex. Fix an integer n and choose, for each n , a characteristic maps
: (D n , S n1 ) (X n , X n1 ). These maps produce a global characteristic map

n : (n D n , n S n1 ) (X n , X n1 ).

Lemma 3.4.7 Let X be a CW-complex and let n N. Let n be a global charac-

teristic map for the n-cells. Then

(i) H n : H (n D n , n S n1 ) Hk (X n , X n1 ) is an isomorphism.

(ii) H n : H (X n , X n1 )
H k (n D n , n S n1 ) is an isomorphism.

Proof By Kronecker duality, using Corollary 2.3.11, only statement (i) must be
proved. The proof for n = 0 is easy and left to the reader, so we assume that n 1.
100 3 Singular and Cellular (Co)homologies

By Lemma 3.4.2, (X n , X n1 ) is a well cofibrant pair. As n 1, the space X n1

is not empty (unless X = , a trivial case). The continuous map

: n D n n S n1 X n / X n1

induced by n is a homeomorphism, both spaces being homeomorphic to a bouquet

of copies of S n indexed by n . In the commutative diagram

H (n D n , n S n1 ) / H (n D n n S n1 )

H n H ,
 
/ H (X n / X n1 )
H (X n , X n1 )

the horizontal maps are isomorphisms by Corollary 3.1.47. Therefore, H n is an

isomorphism. 

Corollary 3.4.8 Let X be a CW-complex and let n N. Then

(i) 
Z2 if k = n.
Hk (X n , X n1 ) n

0 if k = n.

(ii) 
Z2 if k = n.
H k (X n , X n1 ) n

0 if k = n.

Proof Again, the easy case n = 0 is left to the reader. If n > 0, we use that, as
noticed in the proof of Lemma 3.4.7, the map

X n X n1 Sn
n

is a homeomorphism. Corollary 3.4.8 then follows from Proposition 3.1.51. 

Lemma 3.4.9 Let X be a CW-complex and let n N. Then

(i) the homomorphism Hk (X n ) Hk (X ) induced by the inclusion is an
isomorphism for k < n and is surjective for k = n.
(ii) the homomorphism H k (X ) H k (X n ) induced by the inclusion is an
isomorphism for k < n and is injective for k = n.

Proof By Kronecker duality, using Corollary 2.3.11, only statement (i) must be
proved. The homomorphisms induced by inclusions form a sequence
3.4 CW-Complexes 101

Hk (X k )  Hk (X k+1 )
Hk (X k+2 )
Hk (X ) . (3.4.2)

The bijectivity or surjectivity of Hk (X r ) Hk (X r +1 ) is deduced from the homology

sequence of the pair (X r +1 , X r ) and Corollary 3.4.8. By Proposition 3.1.29, H (X )
is the direct limit of H (K ), for all compact sets K of X . Using that each compact
set of X is contained in some skeleton, one checks that H (X ) is the direct limit of
H (X k ). By (3.4.2), this proves (i). 
Lemma 3.4.10 Let X be a CW-complex and let n N. Then H k (X n ) = Hk (X n ) =
0 if k > n.
Proof The proof is by induction on n. The lemma is true for n = 0 since X 0 is
a discrete set. The induction step uses the exact sequence of the pair (X n , X n1 )
together with Corollary 3.4.8. 
Let (X, Y ) be a CW-pair. Let M = {(r, s) N N | r s} endowed with the
lexicographic order. The pairs (X r , Y s ) ((r, s) M), together with the inclusion
 
(X r , Y s )  (X r , Y s ) when (r, s) (r  , s  ), forms a direct system. The inclusions
jr,s : (X r , Y s )  (X, Y ) induce a GrV-morphism

J : lim

H (X r , Y s )
H (X, Y )
(r,s)M

and a GrA-morphism

J : H (X, Y )
lim

H (X r , Y s ).
K K

To get a more general result, which will be useful, we can take the product with an
arbitrary topological space Z .
Proposition 3.4.11 Let (X, Y ) be a CW-pair and M be as above. Let Z be a topo-
logical space. Then, the GrV-morphism

J : lim

H (X r Z , Y s Z )
H (X Z , Y Z )
(r,s)M

and the GrA-morphism

J : H (X Z , Y Z )
lim

H (X r Z , Y s Z ).
K K

are isomorphisms.
Proof By Kronecker duality, only the homology statement needs a proof. Let K
be a compact subspace of X Z . By Property (6) of p. 97, K is contained in
X r Z for some integer r . Hence, if Y is empty, Proposition 3.4.11 follows from
Corollary 3.1.16. When Y = , we use the long exact sequences in homology and
the five lemma, as in the proof of Proposition 3.1.29. 
102 3 Singular and Cellular (Co)homologies

3.5 Cellular (Co)homology

Let X be a CW-complex. For m N, the m-cellular (co)chain vector spaces C m (X )

and C m (X ) are defined as

C m (X ) = Hm (X m , X m1 ) and C m (X ) = H m (X m , X m1 ).

The cellular boundary operator : Cm (X ) C m1 (X ) is defined by the composed

homomorphism

: Hm (X m , X m1 )
Hm1 (X m1 ) Hm1 (X m1 , X m2 ).

The expression for contains the sequence Hm1 (X m1 ) Hm1 (X m1 ,

Hm2 (X m2 ) and then = 0.
X m2 )
The cellular co-boundary operator : C m (X ) C m+1 (X ) is defined by the
composed homomorphism

: H m (X m , X m1 ) H m (X m )
H m+1 (X m+1 , X m ).

with again = 0. (Co)cycles Z m , Z m and (co)boundaries B m , B m are defined as

usual, which leads to the definition

H m (X ) = Z m (X ) B m (X ) and H m (X ) = Z m (X ) B m (X ).

The graded Z2 -vector space H (X ) is the cellular homology of the CW-complex X

and the graded Z2 -vector space H (X ) is its cellular cohomology. The Kronecker
pairing

 ,
H m (X m , X m1 ) Hm (X m , X m1 ) Z2

gives a pairing

 ,
C m (X ) C m (X ) Z2

(C (X ), ),
which makes ((C (X ), ),  , ) a Kronecker pair.
In the language of former sections, the cellular (co)chains admit the usual equiv-
alent definitions:

Definition 3.5.1 (subset definitions)

(a) A cellular m-cochain is a subset of m .
(b) A cellular m-chain is a finite subset of m .
3.5 Cellular (Co)homology 103

Definition 3.5.2 (colouring definitions)

(a) A cellular m-cochain is a function a : m Z2 .
(b) A cellular m-chain is a function : m Z2 with finite support.
Definition 3.5.2b is equivalent to

Definition 3.5.3 C m (X ) is the Z2 -vector space with basis m :


C m (X ) = Z2 .
m (X )

The Kronecker pairing on (co)chains admits the usual equivalent formula

a, = 
(a ) (mod 2)
(3.5.1)
= a().

We now give a formula for the cellular boundary operator : C m (X )

Cm1 (X ). By Definition 3.1.3, it is enough to define () for m . Choose
an attaching map : S m1 X m1 for the m-cell . When m = 1, the formula

for () is easy:

0 if  (S 0 ) = 1
() = (3.5.2)
(S 0 ) otherwise (using the subset definition).

Let us now suppose that m > 1. For m1 , define , : S m1 S m1 as the

composed map:


S m1 X m1  X m1 / X m2 S m1 S m1 ,
m1

where is the projection onto the -th component. Using the colouring definition

of cellular chains, we must give, for each m1 , the value ()() Z2 .

Lemma 3.5.4 For m > 1, the cellular boundary operator : C m (X ) C m1 (X )

is the unique linear map satisfying

()() = deg(, ). (3.5.3)

for each m .

Proof The attaching map : S m1 X m1 extends to a characteristic map

: D m X m . Consider the commutative diagram:
104 3 Singular and Cellular (Co)homologies

H
Hm (D m , S m1 ) / Hm (X m , X m1 )
SSSS
SSS S
SSSS
  )
H
Hm1 (S m1 ) / Hm1 (X m1 ) / Hm1 (X m1 / X m2 )

H

Hm1 (S m1 )

Let be the generator of Hm (D m , S m1 ) = Z2 and let be that of Hm1 (S m1 ).

Using Lemma 3.4.7 and its proof, one sees that
(a) C m (X ) corresponds to H () Hm (X m , X m1 ).
(b) if Hm1 (X m1 / X m2 ), then H () = () (we use the colouring
definition and see as a function from m1 to Z2 ).
As () = , one has

()() = H H ()
= H H ()
= deg(, ) ,

which proves the lemma. 

Formulae (3.5.2) and (3.5.3) for the cellular boundary operator take a special form
when X is a regular CW-complex, i.e. when each cell admits a characteristic map
which is an embedding onto a subcomplex of X . A cell of this subcomplex is
called a face of .

Lemma 3.5.5 Let X be a regular CW-complex. Let m and m1 . Then


()() = 1 if is a face of
0 otherwise.

Proof When m = 1, this follows from (3.5.2), where the case  (S 0 ) = 1 does
not happen since X is regular. When m > 1, we use Lemma 3.5.4 and compute the
degree of , using Proposition 3.2.6: since is an embedding, any topological
regular value of , has exactly one element in its preimage. 

We now prove the main result of this section.

Theorem 3.5.6 Let X be a CW-complex. Then, the cellular and the singular
(co)homology of X are isomorphic:

H (X ) H (X ) and H (X ) H (X ).
3.5 Cellular (Co)homology 105

Proof We consider the commutative diagram:

Hm+1 (X m+1 , X m )
0 / Hm (X m+1 )
RRR
RRRm+1 oo7 7
RRR ooooo
RR( ooo 
m+1 Hm (X m ) / / Hm (X )
lv
jm
lllll
ll
 lv ll
Hm (X m ,X m1 ) 0 (3.5.4)
RRR
RRRm
RRR
RR( 
m Hm1 (X m1 )
mv
jm1 mmm
m mmm
 vmmm
Hm (X m1 ,X m2 )

The properties of arrows (surjective, injective, bijective) come from Lemmas 3.4.9,
3.4.10 and Corollary 3.4.8. From Diagram (3.5.4), we get

jm
Hm (X )

Hm (X m+1 ) Hm (X m )/Im m+1 / ker m /Im m+1 = H m (X )

As the isomorphism H (X ) H (X ) does not come from a morphism of chain com-

plex, we cannot invoke Kronecker duality to deduce the isomorphism in cohomology.
Instead, we consider the Kronecker dual of Diagram (3.5.4)

H m+1 (X m+1 , X m )
O iRRR
RRRm+1
RRR
RR
m+1 H m (X m ) o o H m (X )
Jm llll
l6 6 O gPPP
PPP
ll l PPP
lll Pg 
H m (X m , X m1 ) 0 H (X m+1 )
m (3.5.5)
O hRRR
RRRm
RRR
RR
m H m1 (X m1 )
Jm1 llll66
l l
ll
lll
H m (X m1 , X m2 )
106 3 Singular and Cellular (Co)homologies

which gives

H m (X ) ker m+1 Jm1 (ker m+1 )/Im m
H m (X m+1 )
= ker m+1 /Im m = H m (X ). 

Here are some applications of the isomorphism between cellular and singular
(co)homology.

Corollary 3.5.7 Let X be a CW-complex with no m-dimensional cell. Then Hm (X ) =

H m (X ) = 0.

Proof If m (X ) = , then C m (X ) = C m (X ) = 0, which implies H m (X ) =

H m (X ) = 0, and then Hm (X ) = H m (X ) = 0 by Theorem 3.5.6. 

Corollary 3.5.8 Let X be a CW-complex with k cells of dimension m. Then

dim Hm (X ) = dim H m (X ) k.

Proof One has dim H m (X ) dim Z m (X ) dim C m (X ) = k. Therefore,

dim Hm (X ) k by Theorem 3.5.6. The result on cohomology is proven in the
same way or deduced by Kronecker duality. 

A CW-complex is finite if it has a finite number of cells.

Corollary 3.5.9 Let X be a compact CW-complex. Then

dim H (X ) = dim H (X ) < .

Proof By the weak topology, a compact CW-complex is finite (see Remark (6) p. 97).
Hence, C (X ) is a finite dimensional vector space, and so is Z (X ) and H (X ).
Corollary 3.5.9 then follows from Theorem 3.5.6 and Kronecker duality. 

Let X be a finite CW-complex. Its Euler characteristic (X ) is defined as


(X ) = (1)m m (X ) Z.
mN

Proposition 3.5.10 Let X be a finite CW-complex. Then

 
(X ) = (1)m dim Hm (X ) = (1)m dim H m (X ) .
mN m

Proof If we use the cellular (co)homology, the proof of Proposition 3.5.10 is the
same as that of Proposition 2.4.8. The result then follows from Theorem 3.5.6. 

A CW-complex X (or a CW-structure on X ) is called perfect if the cellular bound-

ary vanishes. For instance, if X does not have cells in consecutive dimensions, then
it is perfect. Also, the standard CW-structure on RP n (n ) is perfect (see e.g.
3.5 Cellular (Co)homology 107

Proposition 6.1.1). If X is a perfect CW-complex, C (X ) = H (X ) and the iden-

tification between the singular and cellular homologies, out of Diagram (3.5.4), is
particularly simple:

Hm (X ) H m (X m , X m1 )
Hm (X m ) H m (X )
C m (X ). (3.5.6)

The natural functoriality of cellular (co)homology is for cellular maps. If X and

Y are CW-complexes, a continuous map f : Y X is cellular if f (Y m ) X m
for all m N. We thus get GrV-morphisms C f and C f making the following
diagrams commute
C f C f
C m (Y ) / C m (X ) C m (Y ) o C m (X )
O O
= = = =
 
C f
Hm (Y m , Y m1 )
H f
/ Hm (X m , X m1 ) H m (Y m , Y m1 ) o H m (X m , X m1 )

They satisfy C f (a), = a, C f () for all a C (X ) and C (X ). It is

useful to have a formula for C f , using that
 
C m (Y ) = Z2 and C m (X ) = Z2 .
m (Y ) m (X )

For m (Y ) and m (X ), consider the map f , : S m S m defined by the

composition
j  f 
f , : S m
S m Y m /Y m1
X m / X m1 S m S m ,
m (Y ) m (X )

where j is the inclusion of the -component and the projection onto the -
component.
Lemma 3.5.11 For m 1, C f : Cm (Y ) Cm (X ) is the unique linear map such
that

C f () = deg( f , ) .
m (X )

m1 ) (X m , X m1 ) induces a map f :

 The map mf : (Y , Y
Proof m Sm
m (Y )
m (X ) S making the following diagram commute


C m (Y ) / Hm (Y m , Y m1 ) / Hm ( m (Y ) Sm )

C f H f H f
   
/ Hm (Y m , Y m1 ) / Hm (
C m (X ) m (X ) Sm )
108 3 Singular and Cellular (Co)homologies

As in the proof of Lemma 3.5.4, one checks that, under the top horizontal isomor-
phisms, C m (Y ) corresponds to H j ([S m1 ]). Also, if Hm1 (X m1 / X m2 ),
then H () = () [S m1 ] (seeing as a function from m (X ) to Z2 by the
colouring definition). Hence,

C f ()() = H H f H j ([S m1 ]) = H f , ([S m1 ])

= deg( f , ) [S m1 ],

which proves the lemma. 

3.5.12 Homology-cell complexes. The results of this section and the previous one are
also valid for complexes where cells are replaced by homology cells. A well cofibrant
is a homology n-cell if H (B) = 0 and H ( B)
pair (B, B) H (S n1 ). This GrV-
isomorphism is abstract, i.e. not assumed to be given by any continuous map. It
follows that H (B, B) H (D n , S n1 ). We also say that B is a homology n-cell
with boundary B. If  is indexing a family of homology n-cell {(B(), B())}


and if : B() Y is a family of continuous maps, we say that the quotient space
 
X = Y  B()

is obtained from Y by attaching homology n-cells (they may be different for various
s). We identify  with the set of homology n-cells.
A homology-cell complex is defined as in p. 97 with attachments of n-cells
replaced by attachments of a set n (X ) of homology n-cells. The cellular
(co)homology H (X ) and H (X ) are defined accordingly and Theorem 3.5.6 holds
true, with the same proof. Homology-cell structures are used in the proof of Poincar
duality (see Sect. 5.2).

3.6 Isomorphisms Between Simplicial and Singular

(Co)homology

Let K be a simplicial complex. In this section, we prove three theorems showing

that the simplicial (co)homology of K and the singular (co)homology of |K | are
isomorphic.

Theorem 3.6.1 Let K be a simplicial complex. Then

H (K ) H (|K |) and H (K ) H (|K |)

Proof The geometric realization |K | of K is naturally endowed with a structure of

a regular CW-complex, with |K |m = |K m |, m (|K |) = Sm (K ), with a canonical
characteristic map for the m-cell Sm (K ) given by the inclusion of ||
into |K |.
Thus, C m (|K |) = Cm (K ) and, using Lemma 3.5.5, the diagram
3.6 Isomorphisms Between Simplicial and Singular (Co)homology 109

= / Cm (K )
C m (|K |)

 
= / Cm1 (K )
C m1 (|K |)

is commutative. Therefore, H (|K |) = H (K ) and, by Theorem 3.5.6, the singular

homology H (|K |) and the simplicial homology H (K ) are isomorphic. The equality
H (|K |) = H (K ) is deduced from H (|K |) = H (K ) by Kronecker duality and,
using by Theorem 3.5.6 again, the singular cohomology H (|K |) and the simplicial
cohomology H (K ) are also isomorphic. 

We now go to the second isomorphism theorem, which uses the ordered simplicial
(co)homology of Sect. 2.10. To an ordered m-simplex (v0 , . . . , vm ) Sm (K ), we
associate the singular m-simplex R(v0 , . . . , vm ) : m |K | defined by


m
R(v0 , . . . , vm )(t0 , . . . , tm ) = ti vi . (3.6.1)
i=0

The linear combination in (3.6.1) makes sense since {v0 , . . . , vm } is a simplex of K .

This defines a map R : Sm (K ) Sm (|K |) which extends to a linear map

R : C (K ) C (|K |) .

The formula R = R is obvious, so R is a morphism of chain complexes

to (C (|K |), ). Define the linear map R : C (|K |) C (K ) by
from (C (K ), )
R (a), = a, R () . By Lemma 2.3.6, (R , R ) is a morphism of Kronecker

pair. We also denote by R and R the induced linear maps on (co)homology:

R : H (K ) H (|K |) and R : H (|K |) H (K ) .

If f : L K be a simplicial map, the formulae

R C f = C | f | R and C f R = R C | f |

are easy to check. They induce the formulae

R H f = H | f | R and H f R = H C | f | (3.6.2)

on (co)homology. In particular, if f is the inclusion of a subcomplex L of K , the

above considerations permit us to construct degree zero linear maps

R : H (K , L) H (|K |, |L|) and R : H (|K |, |L|) H (K , L)

110 3 Singular and Cellular (Co)homologies

so that (R , R ) is a morphism of Kronecker pair. Finally, if f : (K , L) (K  , L  )

is a simplicial map of simplicial pairs, then Formulae (3.6.2) hold true in relative
(co)homology.

Theorem 3.6.2 Let (K , L) be a simplicial pair. Then the linear maps

R : H (K , L) H (K , L)
H (|K |, |L|) and R : H (|K |, |L|)

are isomorphisms. They are functorial for simplicial maps of simplicial pairs

Proof The functoriality has already been established. By Kronecker duality, it is

enough to prove that R is an isomorphism. The proof goes through a couple of
particular cases.
Case 1: (K , L) = (F A, F A), where F A is the full complex on the finite set of m +1
elements A = {v0 , . . . , vm } (see p. 24), which is isomorphic to an m-simplex. Then
(|F A|, |F A|) (D m , S m1 ). By Corollaries 2.4.7 and 3.2.2,

H k (F A, F A) = Hk (F A, F A) = Hk (|F A|, |F A|) = 0

if k = m and

H m (F A, F A) Hm (F A, F A) Hm (|F A|, |F A|) Z2 .

Thus, it is enough to prove that R : H m (F A, F A) Hm (|F A|, |F A|) is

not trivial. The vector space H m (F A, F A) is generated by the ordered simplex
= (v0 , . . . , vm ). Let r = R() : (m ,  m ) |F A|, |F A|). One has [r ] =
Hr ([i m ]) where i m is the identity map of (m ,  m ). But [i m ] = 0 in Hm (m , 
m)
by Proposition 3.2.3 and r is a homeomorphism of pairs. Thus R () = 0 in
Hm (|F A|, |F A|).
Case 2: (K , L) = (K m , K m1 ) with m 1. The non-vanishing homology groups
are

H m (K m , K m1 ) Hm (K m , K m1 ) Hm (|K m |, |K m1 |) Z2
Sm (K )

For each Sm (K ) choose an ordered simplex = (v0 , . . . , vm ) with {v0 , . . . ,

vm } = . Then H m (K m , K m1 ) Hm (K m , K m1 ) is the Z2 -vector space with
basis { | Sm (K )}. Denote by the subcomplexes of K generated by and by
the subcomplex of the proper faces of .
The map r : (||, (|K m |, |K m1 |)
||)
is a characteristic map for the m-cells of |K | corresponding to . The union r m of
the r is then a global characteristic map for the m-cells of |K |. Let us consider the
commutative diagram
3.6 Isomorphisms Between Simplicial and Singular (Co)homology 111

 
H m (,
)
R
/ Hm (||,
||)
Sm (K ) Sm (K )

H r m
 
H m (K m , K m1 )
R
/ Hm (|K m |, |K m1 |)

The bijectivity of the left vertical arrow was seen above. That of the right vertical
arrow is Lemma 3.4.7. The bijectivity of the top horizontal arrow comes from Case 1.
Hence, R : H m (K m , K m1 ) Hm (|K m |, |K m1 |) is an isomorphism.
Case 3: (K , L) = (K m , ). This is proven by induction on m, the case m = 0
being obvious. By the naturality of R , one has the commutative diagram of exact
sequences:

H +1 (K m , K m1 )

/ H (K m1 ) / H (K m ) / H (K m , K m1 )

/ H 1 (K m1 )

R R R R R
    
H+1 (K m , K m1 )

/ H (K m1 ) / H (K m ) / H (K m , K m1 )

/ H1 (K m1 )

(one has to check that the diagrams with are commutative). The bijectivity of
the vertical arrows come by induction hypothesis and by case 2. By the five-lemma,
R : H (K m ) H (|K m |) is an isomorphism.
General case. We first prove that R : H m (K ) Hm (|K |) is an isomorphism for
all m. By the naturality of R , one has the commutative diagram:

H m (K m+1 )
R
/ Hm (|K m+1 |)

 
H m (K )
R
/ Hm (|K |) .

The bijectivity of the left vertical arrow is obvious. That of the right vertical arrow is
Lemma 3.4.9. The bijectivity of the top horizontal arrow was established in Case 3.
Therefore, the bottom horizontal arrow is bijective. Finally, the general case (K , L)
is deduced from the absolute cases using, as in Case 3, the homology sequences of
the pair (K , L) and the five-lemma. 
Four our third isomorphism theorem, choose a simplicial order on K . Define
a map R : S(K ) S(|K |) by R () = R() where, if = {v0 , . . . , vm }, then
= (v0 , . . . , vm ) with v0 vm . As above, we check that R induces linear
maps R, : H (K , L) H (|K |, |L|) and R : H (|K |, |L|) H (K , L) of

degree zero.
Theorem 3.6.3 Let (K , L) be a simplicial pair. For any simplicial order on K ,
the linear maps
112 3 Singular and Cellular (Co)homologies

R, : H (K , L)
H (|K |, |L|) and R : H (|K |, |L|)
H (K , L)

are isomorphisms. Moreover, these isomorphisms do not depend on the simplicial

order .

Proof By Kronecker duality, only the homology statement requires a proof. By our
definitions, one has the commutative diagram

R,
H (K , L) / H (|K |, |L|)
MMM p7
MMHM R ppp
M pp
MM& ppp
H (K , L)

The bijectivity of the arrows come from Corollary 2.10.13 and Theorem 3.6.3. There-
fore, R, is an isomorphism. As H is independent of by Corollary 2.10.13,
so is R, . 

3.7 CW-Approximations

It is sometimes useful to know that any space has the (co)homology of a CW-complex
or of a simplicial complex (see e.g. p. 132, 161 and 326 in this book). We give below
classical functorial results about that. Relationships with similar constructions in the
literature are discussed in Remarks 3.7.5 at the end of the section.
We shall need a standard notion of category theory: natural transformations. Let
a and b be two (covariant) functors from a category C to a category C . A natural
transformation associates to each object X in C a morphism X : a(X ) b(X ) in
C such that the diagram

a( f )
a(X ) / a(Y )

X Y (3.7.1)
 b( f )

b(X ) / b(Y )

is commutative for every morphism f : X Y in C.

We first consider the category of CW-spaces and cellular maps. It is denoted by
CW and, as usual, by CW2 for pairs of CW-complexes. We denote by j be the
inclusion morphism from CW2 to Top2 .

Theorem 3.7.1 There is a covariant functor cw : (X, Y ) (X CW , Y CW ) from

Top2 CW2 and a natural transformation = (X,Y ) : (X CW , Y CW ) (X, Y )
from j cw to the identity functor of Top2 , such that H and H are isomorphisms.
3.7 CW-Approximations 113

The construction in the proof below is sometimes called in the literature the thick
geometric realization of the singular complex of X .
Proof We start with some preliminaries. If m is the standard m-simplex and I
{0, 1, . . . , m}, we set

mI = {(t0 , . . . , tm ) m | ti = 0 if i
/ I}

which is a simplex of dimension I 1. This gives rise to an obvious inclusion map

 I : I 1 m .
Let X be a topological space. The space X CW is defined as the quotient space

 
X CW = Sm (X ) m (3.7.2)
m0

where is the equivalence relation (,  I (u)) (  I , u) for all Sm (X ),

I {0, . . . , m} and u I 1 . Then X C W is a CW-complex whose k-skeleton is

 
(X CW )k = Sm (X ) m .
0mk

In particular, (X CW )0 is just the space X endowed with the discrete topology. The
k-cells are indexed by Sk (X ). The characteristic map for the k-cell corresponding to
Sk (X ) is the restriction to k of the quotient map from the disjoint union
in (3.7.2) onto (X CW )k .
A continuous f : X 1 X 2 determines a cellular map f CW : X 1CW X 2CW
induced by f CW (, u) = ( f , u). Note that, if Y is a subspace of X , then Y CW
is a subcomplex of X CW . We thus check that cw is a covariant functor (X, Y )
(X C W , Y C W ) from Top2 to CW2 .
For Sm (X ), one has a continuous map : {} m X defined by
(, u) = (u). The disjoint union of those descends to a continuous map
: X CW X , or : (X CW , Y CW ) (X, Y ). One has

f CW (, u) = ( f , u) = f (u) = f (, u)

which amounts, using (3.7.1), to being a natural transformation from j cw to the

identity functor of Top2 .
It remains to prove that H is a GrV-isomorphism (that H is a GrA-
isomorphism will follow by Kronecker duality). We start with the absolute case
Y = . We shall construct a diagram

H
H (X CW ) / H (X )
fMMM ss
MMM ss
M
MM sss (3.7.3)
sy
H (X CW )
114 3 Singular and Cellular (Co)homologies

=
such that H = id and and are isomorphisms. The bijection S(X )
{cells of X CW } extends to a linear map : C (X ) C (X CW ) which satisfies

= and thus induces the isomorphism : H (X ) H (X CW ). For ,
one associates to the k-cell of X CW indexed by the map

[Link]
k
{} k X CW

which is an element of Sk (X CW ). This extends to a linear map : C k (X CW )

Ck (X CW ). Again, we check that = . We thus get the linear map :
H (X CW ) H (X CW ). The equation H = id is straightforward.
It remains to prove that is an isomorphism. To simplify the notation, write
X = X CW . Note that induces linear maps k : H ( X k ) H ( X k ) and
k+1,k : H ( X k+1 , X k ) H ( X k+1 , X k ). Obviously, H ( X ) = limk H ( X k ).
By Corollary 3.1.16, one also has that H ( X ) = limk H ( X k ). Therefore, it suf-
fices to show that k is an isomorphism for all k. This is done by induction on
k. It is obviously true for k = 0, since X 0 is a discrete space. For the induction
step, suppose that k : H i ( X k ) Hi ( X k ) is an isomorphism for all i N. Then,
k+1 : H i ( X k+1 ) Hi ( X k+1 ) is an isomorphism for all i, except perhaps for
i = k, k + 1 where we must consider the commutative diagram

0 / H k+1 ( X k+1 ) / H k+1 ( X k+1 , X k ) / H k ( X k ) / H k ( X k+1 ) / 0

k+1 k+1,k k k+1

   
0 / Hk+1 ( X k+1 ) / Hk+1 ( X k+1 , X k ) / Hk ( X k ) / Hk ( X k+1 ) / 0

(3.7.4)

where the horizontal lines are the cellular and singular homology exact sequences
of the pair ( X k+1 , X k ). By the five lemma, it thus suffices to prove that k+1,k is an
isomorphism. One has the commutative diagram


H k+1 ( X k+1 , X k ) o H k+1 ({} (k+1 , Bdk+1 ))
Sk+1 (X )

k+1,k (3.7.5)

 

Hk+1 ( X k+1 , X k ) o Hk+1 ({} (k+1 , Bdk+1 ))
Sk+1 (X )

where sends the (k + 1)-cell {} (k+1 (generator of H k+1 ({} (k+1 ,

Bdk+1 )) = Z2 ) to the tautological singular simplex k+1 {} k+1 . The
3.7 CW-Approximations 115

latter is the generator of Hk+1 ({}(k+1 , Bdk+1 )) = Z2 (see Proposition 3.2.3).

Hence, k+1,k is an isomorphism.
We have proven that H : H (X CW ) H (X ) is a GrV-isomorphism for all
topological space X . Using the homology sequences and the five lemma, this implies
that H : H (X CW , Y CW ) H (X, Y ) is a GrV-isomorphism for all topological
pairs (X, Y ). 
A slightly more sophisticated construction for the functor of Theorem 3.7.2 gives
the following result. Let RCW be the category of regular CW-complexes and cellular
maps and let j be the inclusion morphism from RCW2 to Top2 .
Theorem 3.7.2 There is a covariant functor rcw : (X, Y ) (X RCW , Y RCW ) from
Top2 RCW2 and a natural transformation = (X,Y ) : (X RCW , Y RCW )
(X, Y ) from j rcw to the identity functor of Top2 , such that H and H are
isomorphisms.
The proof of Theorem 3.7.2 requires some preliminaries.
Let FN be the full simplicial complex with vertex set the integers N. If X is a
topological space, the set of N-singular simplexes of X is defined by

NS(X ) = {(s, ) | s S(FN) and : |s | X is a continuous map}.

where s is the simplicial complex formed by s and all its faces (see p. 6). Let
NSn (X ) be the subset of NS(X ) formed by those pairs (s, ), where s is of dimension
n and let NCn (X ) be the Z2 -vector space with basis NSn (X ). Using the facets of s ,
we define a boundary operator : NCn (X ) NCn1 (X ) making NC (X ) a chain
complex. The homology of this chain complex is the N-singular homology of X ,
denoted by NH (X ). The relative homology NH (X, Y ) is defined as in Sect. 3.1.2.
The order on N provides a simplicial order on FN. Thus, if s Sn (FN), there

is a canonical homeomorphism h s : |s | n (see (3.1.6)). We define maps :
Sn (X ) NSn (X ) and : NSn (X ) Sn (X ) by:
() = (s0 , h s0 (n) ), where s0 (n) = {0, 1, . . . , n} and
(s, ) = h 1
s .
The linear extensions C : Cn (X ) NCn (X ) and C : NCn (X ) Cn (X )
commute with the boundary operators and are thus morphisms of chain complexes.
The constructions extend to pairs and we get GrV-morphisms H : Hn (X, Y )
NHn (X, Y ) and H : NHn (X, Y ) Hn (X, Y ).

Lemma 3.7.3 H : Hn (X, Y ) NHn (X, Y ) and H : NHn (X, Y )

Hn (X, Y ) are isomorphisms, inverse of each other.

Proof Clearly, = id on S(X ), thus H H = id on H (X, Y ). To see that

H H = id on NH (X, Y ), we first restrict ourselves to the absolute case Y = .
We shall prove that C C and the identity of NC (X ) admit a common acyclic
carrier A with respect to the basis NS(X ). The condition that H H = id on
NH (X ) then follows from Proposition 2.9.1.
116 3 Singular and Cellular (Co)homologies

For (s, ) NSn (X ) and k N, define Bk (s, ) by

Bk (s, ) = {(t, | p|) | t Sk (FN) and p : t s is a simplicial map }.

Let Ak (s, ) be the Z2 -vector space with basis Bk (s, ). Using the restriction of p to
the facets of t, one defines a boundary operator : Ak (s, ) Ak1 (s, ) making
A (s, ) a subchain complex of NC (X ).
For (s, ) NS(X ), one has (s ) A (s, ) ( p = ids and

(s, ) = (s0 , h 1
s h s0 (n) ) A (s, ),

since h 1
s h s0 (n) = | p| for p : s s0 the unique simplicial isomorphism preserving
the order. The conditions for the correspondence (s, ) A (s, ) being an acyclic
carrier (see Sect. 2.9) are easy to check once we know that H0 (A (s)) = Z2 and
Hm (A (s)) = 0 for m > 0 which we prove below.
Consider the simplicial complex K (s) with vertex set V (K (s)) = N V (s) and
whose k-simplexes are the sets {(n 0 , s0 ), . . . , (n k , sk )} with n 0 < < n k . We check
that the correspondence sending (t, | p|) to the graph of p : V (t) V (s) induces
an isomorphism of chain complex between A (s, ) and the simplicial chain of K (s).
We have thus to prove that H (K (s)) H ( pt). But K (s) is the union of K n (s),
where K n (s) is the union of all simplexes of K (s) with vertices in {0, . . . , n} V (S).
The inclusion V (K n (s))  V (K n+1 (s)) together with the map k (n + 1, k)

provides a bijection V (K n (s)) V (s)
V (K n+1 (s)) and a simplicial isomorphism

K n (s) s 0
K n+1 (s),

where s 0 is the 0-skeleton of s . Hence, the inclusion K n (s)  K n+1 (s) factors
through the cones on K n (s) contained in the join K n (s) s 0 . Therefore, H (K (s))
lim H (K n (s)) H ( pt).

We have thus proved Lemma 3.7.3 in the case Y = . Using the homol-
ogy sequences, this proves that H : Hn (X, Y ) NHn (X, Y ) and H :
NHn (X, Y ) Hn (X, Y ) are both isomorphisms. But we have already noted that
H H = id on H (X, Y ). Therefore, H H = id on NH (X, Y ). 
We are now ready for the proof of Theorem 3.7.2.
Proof of Theorem 3.7.2 If t s are simplexes of FN, we denote by i t,s : t s
the simplicial map given by the inclusion. The space X RCW is defined as the quotient
space


 
X RCW = {(s, ))} |s | (3.7.6)
(s, )NS (X )

where is the equivalence relation ((s, ), |i t,s |(u)) (t, |i t,s |), u) for all (s, )
NS(X ), all subsimplex t of s and all u |t|. As in the proof of Theorem 3.7.1, X RCW
is a naturally a CW-complex. The characteristic map for the k-cell corresponding to
3.7 CW-Approximations 117

(s, ) NSk (X ) is the restriction to {(s, ))} |s | of the quotient map in (3.7.6). In
particular, (X RCW )0 is the set N X endowed with the discrete topology. Because
of the role of N in the indexing of the cells, one checks that X RCW is a regular CW-
complex. For (s, ) NSm (X ), one has a continuous map (s, ) : {(s, )}|s | X
defined by (s, ) ((s, ), u) = (u). The disjoint union of those evaluation maps
descends to a continuous map : X RCW X , if Y is a subspace of X ,
then Y RCW is a subcomplex of X RCW . The functoriality of the correspondence
(X, Y ) (X RCW , Y RCW ), as well as that is a natural transformation from j rcw
to the identity functor of Top2 , are established as in the proof of Theorem 3.7.1.
We now prove that NH : NH (X RCW ) NH (X ) is a GrV-isomorphism,
following the pattern of the proof of Theorem 3.7.1. Similarly to (3.7.3), we construct
the diagram

N H
NH (X RCW ) / NH (X )
gOOO qq
OOO
O qqqqq (3.7.7)
N OO xqq N
H (X RCW )

such that NH N N = id and N and N are isomorphisms. As in (3.7.3), the

=
bijection NS(X )
{cells of X RCW } gives the isomorphism N. The gr v-morphism
comes from associating to the k-cell of X RCW indexed by (s, ) the map
[Link]
|s |
{(s, )} |s | X RCW

which is an element of NSk (X RCW ). The equation NH N N = id is straight-

forward.
The proof that N is an isomorphism is quite similar to that (for ) in the
proof of Theorem 3.7.1. Indeed, using Lemma 3.7.3, NH ( ) is a homology the-
ory and thus diagrams like in (3.7.4) and (3.7.5) do exist; the isomorphism H :
NH (k1 , Bdk+1 ) f lH (k1 , Bdk+1 ) is also explicit enough and permits
to proceed as in the proof of Theorem 3.7.1. Details are left to the reader.
The GrV-isomorphism H of Lemma 3.7.3 is natural; one has thus a commu-
tative diagram

N H
NH (X RCW ) / NH (X )


 
H
H (X RCW ) / H (X )

which shows that H is an isomorphism. The relative case is obtained as at the

end of the proof of Theorem 3.7.1 and that H is a GrA-isomorphism comes from
Kronecker duality. 
118 3 Singular and Cellular (Co)homologies

Theorem 3.7.4 There is a covariant functor symp : (X, Y ) (K X , K Y ) from

Top2 Simp2 and a natural transformation = (X,Y ) : (|K X |, |K Y |) (X, Y )
from | symp| to the identity functor of Top2 such that H and H are isomor-
phisms.
Proof Let X be a topological space. By the proof of Theorem 3.7.2, the regular CW-
complex X RCW comes equipped with characteristic embeddings (s, ) : {(s, ))}
|s | X RCW ((s, )) NS(X )), satisfying the following condition: if t is a face of
s with simplicial inclusion i t,s : t s , then (t, |it,s |) = (s, ) |i t,s |. The only
missing thing to make X RCW a simplicial complex is that several simplexes may have
the same boundary. But this can be avoided by taking the barycentric subdivision of
each cell, with the characteristic embedding (s, ) : {(s, ))} |(s ) | X RCW . We
thus get a functorial triangulation of X RCW . 
Remarks 3.7.5 The following facts about the above constructions should be noted.
(a) The proof of Theorem 3.7.1 goes back to Giever [67]; for a more recent treat-
ment, see [73, p. 146]. Theorem 3.7.2 may be obtained from Theorem 3.7.1 by
subdivision techniques in semi-simplicial complexes (see [73, Theorem 16.41]).
Our proof of Theorem 3.7.2 is different.
(b) The construction X CW in the proof of Theorem 3.7.1 is sometimes called in
the literature the thick geometric realization of the singular complex of X . A
quotient X CW of X CW was introduced by Milnor [146], in which the degenerate
simplexes are collapsed. Thus, X CW has one k-cell for each non-degenerate
singular k-simplex of X . Under mild conditions, the Milnor functor behaves
well with products (see [146, Sect. 2]).
(c) The maps of Theorems 3.7.1, 3.7.2 and 3.7.4 are actually weak
homotopy equivalences (see e.g. [73, Corollary 16.43]). Such maps are called
CW-approximations [82] or resolutions [73]. In particular, if X is itself a
CW-complex, these maps are homotopy equivalences by the Whitehead the-
orem [82, Theorem 4.5] (but they are not homeomorphisms). Somehow simpler
(but not functorial) proofs that a spaces has the weak homotopy type of a CW-
complex may be found in e.g. [82, Proposition 4.13] or [73, Proposition 16.4].
(d) By its construction in the proof of Theorem 3.7.2, X RCW is a regular -set is
the sense of [82, pp. 53334]. In this appendix of [82], the reader may find
enlightening considerations related to our constructions in this section.

3.8 Eilenberg-MacLane Spaces

The Eilenberg-MacLane spaces are used to make the cohomology H () a

representable functor. With Z2 as coefficients, they admit an ad hoc presentation
given below, which only uses the material developed in this book. The equivalence
with the usual definition using the homotopy groups is proven at the end of the
section.
3.8 Eilenberg-MacLane Spaces 119

A CW-complex K is an Eilenberg-MacLane space in degree m if

(i) H m (K) = Z2 ; we denote by the generator of H m (K).
(ii) for any CW complex X , the correspondence f H f () gives a bijection

: [X, K]
H m (X ),

where [X, K] denotes the set of homotopy classes of continuous maps from X
to K.
If f : X K is a map, the class H f () is said to be represented by f . Property
(ii) says that the functor H () would be representable by K in the sense of category
theory [134].
The notation K (Z2 , m) is usual for a CW-complex which is an Eilenberg-
MacLane space in degree m. We shall also use the notation Km . The unambiguity of
these notations is guaranteed by the following existence and uniqueness result.

Proposition 3.8.1 (a) For any integer m, there exists an Eilenberg-MacLane space
in degree m.
(b) Let Km and Km  be two Eilenberg-MacLane spaces in degree m. Then, there
 K whose homotopy class is unique.
exists a homotopy equivalence g : Km m

Example 3.8.2 By Corollary 3.1.12, we see that the point is an Eilenberg-MacLane

space in degree 0.

Proof We start by the uniqueness statement (b). Let K and K be two Eilenberg-
MacLane spaces in degree m. Then, there is a bijection

Z2 = H m (K ) [K , K]

under which the constant maps corresponds to 0. Let g : K K be a continuous

map representing the non-vanishing class (unique up to homotopy). In the same way,
let h : K K represent the non-vanishing class of Z2 = H m (K) [K, K ].
Then, g h represent the non-vanishing class of Z2 = H m (K) [K, K] and h g
represent the non-vanishing class of Z2 = H m (K ) [K , K ]. As idK and idK do
the same, we deduce that g h is homotopic to idK and h g is homotopic to idK .
Therefore, h and g are homotopy equivalences.
We now construct an Eilenberg-MacLane space K in degree m 1 (K0 = pt,
as noticed in Example 3.8.2). Its m-skeleton Km is the sphere S m , with one 0-cell
{v} and one m-cell called . Then, for each map : S m Km of degree 0, an
(m + 1)-cell is attached to to Km via , thus getting Km+1 . Finally, for k m + 2,
Kk is constructed by induction by attaching to Kk1 a k-cell for each continuous
map f : S k1 Kk1 .
As the (m + 1)-cells of K are attached to Km by maps of degree 0, the cellular
boundary : C n+1 (K) C n (K) vanishes by Lemma 3.5.4. Therefore, Hm (K) =
Z2 by Theorem 3.5.6 and H m (K) = Z2 by Kronecker duality. The singleton {},
120 3 Singular and Cellular (Co)homologies

seen as a cellular m-cycle of K, is called Z m (K). Seen as an m-cocycle, we

denote it by Z m (K) (it represents H m (K)).
Let us prove the surjectivity of : [X, K] H m (X ). Let a H m (X ), rep-
resented by a cellular cocycle a C m (X ) C m (X m ). We shall construct a map
f : X K such that H f () = a. Let j : D m K be a characteristic map for the
unique m-cell of K. The map f sends X m1 to the point {v} = K0 . Its restriction to
an m-cell e of X is equal to j if e a and the constant map onto v otherwise. This
gives a map f : X m Km which, by construction and Lemma 3.5.11, satisfies

C f () = a, (3.8.1)

for all C m (X ). Hence

C f (), = , C f () = , a, = a,

for all C m (X ). By Lemma 2.3.3, we deduce that C f () = a.

To extend f to X m+1 , let m+1 (X ) with attaching map : S m X m . As
a is a cocycle, one has a, = (a), = 0. Using Lemmas 3.5.4 and 3.5.11
together with Eq. (3.8.1), we get that f = f : S m Km S m is a map of
degree 0. By construction of K, an (m + 1)-cell e is attached to Km via f , so f
may be extended to , using a characteristic map for e extending f . This produces a
cellular map f m+1 : X m+1 Km+1 . Finally, suppose, by induction on k m + 1,
that f m+1 extends to f k : X k Kk . Let k+1 (X ) with attaching map
: S k X k . Set g = f k . By construction of K, there exists eg k+1 (K)
with attaching map g . Thus f k may be extended to the cell , using a characteristic
map for e extending g . Doing this for each k+1 (X ) produces the desired
extension f k+1 : X k+1 Kk+1 . The surjectivity of : [X, K] H m (X ) is thus
established.
For the injectivity of , let f 0 , f 1 : X K such that H f 0 () = H f 1 ().
Since any map between CW-complexes is homotopic to a cellular map (see e.g.
[64, Theorem 2.4.11]), we may assume that f 0 and f 1 are cellular. We must construct
a homotopy F : X I K between f 0 and f 1 , which will be done cell by cell.
As f 0 (X m1 ) = f 1 (X m1 ) = {v}, the maps f 0 and f 1 descend to cellular maps
from X/ X m1 to K. Hence, we can assume that X m1 = X 0 is a single point
w, with f 0 (w) = f 1 (w) = v. The homotopy F is defined to be constant on w:
F(w, t) = v.
As X m1 is a point, the homology class H f 0 () = H f 1 () is represented
by a single cellular cocycle a C m (X ) ( B m (X ) = 0). Let m (X )
with characteristic map : S m X . By Lemma 3.5.11, one has, for j =
0, 1:

a, = C f j (), = , C f j () = , deg( f j ) = deg( f j ).

(3.8.2)
Let  m be the boundary of D m I , homeomorphic to S m . A map F :  m Km
is defined by
3.8 Eilenberg-MacLane Spaces 121


f 0 ( (x)) if t = 0
F (x, t) = f 1 ( (x)) if t = 1 (3.8.3)


v if x S m1 .

Using (3.8.2) together with Lemma 3.2.8 (with B1 = D m {0} and B2 = D m {1}),
we deduce that deg F = 0. Then, there is an (m + 1)-cell of K is attached to
Km with F . This implies that F extends to F : D m I Km+1 which is
a homotopy from f 0 to f 1 over X m union the cell . Doing this for each
m (X ) produces a homotopy F m : X m I Km+1 between f 0 and f 1 . We
can thus assume, by induction on k m, that a homotopy F k : X k I Kk+1
between f 0 and f 1 has been constructed. We must extend it to F k+1 : X k+1 I
Kk+2 , which can be done individually over each cell k+1 (X ). We define
F :  k+1 Kk+1 as in (3.8.3). As k + 1 > m, a (k + 2)-cell of K is attached
to Kk+1 with F , which permits us, as above, to extend the homotopy F k over the
cell .
The proof of Proposition 3.8.1 is now complete. 

The above construction of an Eilenberg-MacLane space uses a lot of cells so we

may expect that the (co)homology of Kn is complicated. It was computed by Serre
[175, Sect. 2], whose result will be given in Theorem 8.5.5. In degree 1 however, we
have the following simple example of an Eilenberg-MacLane space.

Proposition 3.8.3 The projective space RP is an Eilenberg-MacLane space in

degree 1 ( RP K (Z2 , 1)).

Proof We use the standard CW-structure on K = RP of Example 3.4.5, with

one cell in each dimension and so that Kk = RP k . Let pk : S 1 S 1 given by
pk (z) = z k . The following properties hold true:
(i) the 2-cell of K is attached to K1 S 1 by the map p2 which, by Proposition 3.2.6
is of degree 0.
(ii) each map g : S 1 K1 of degree 0 is null-homotopic (i.e. homotopic to a
constant map) in K2 . Indeed, it is classical that any map from S 1 to S 1 S 1
is homotopic to pk for some integer k (see e.g. [136, Theorem 5.1] or [82,
Theorem 1.7]). By Proposition 3.2.6, deg pk = 0 if and only if k = 2r . Point (i)
implies that g = p2 is null homotopic and so is p2r = p2 pr .
(iii) for k 2, each map g : S k Kk is null-homotopic into Kk+1 . Indeed, the
lifting property of covering spaces tells us that g admits a lifting

Sk
g z=
z
z p
z g

Sk / RP k

and the (k + 1)-cell of K is attached via the covering map p.

122 3 Singular and Cellular (Co)homologies

By Point (i), H 1 (K) = Z2 . Points (ii) and (iii) imply that the argument of the proof of
Proposition 3.8.1 may be used to prove that : [X, RP 1 ] H 1 (X ) is a bijection.
Hence, K = RP is an Eilenberg-MacLane space in degree 1. 

Corollary 3.8.4 Let f : RP n RP k be a continuous map, with n < k .

Then f is either homotopic to a constant map or to the inclusion RP n  RP k .

Proof The lemma is true for k = by Proposition 3.8.3. Therefore, there is a homo-
topy from the composition of f with the inclusion RP k  RP to either a constant
map or the inclusion. Making this homotopy cellular (see [207, Corollary 4.7,p. 78])
produce a homotopy whose range is in RP n+1 . 

We finish this section with the relationship between our definition of Eilenberg-
MacLane spaces and the usual one involving the homotopy groups. Recall that the
i-th homotopy group i (X, x) of a pointed space (X, x) is defined by i (X, x) =
[S i , X ] , for some fixed base point in S n . Below, the base points are omitted from
the notation.

Proposition 3.8.5 A CW-complex X is an Eilenberg-MacLane space Km if and only

if i (X ) = 0 if i = m and m (X ) = Z2 .

Proof We first prove that Km satisfies the conditions. By Propositions 3.8.1 and 3.8.3,
the space K1 is homotopy equivalent to RP 1 . The statement then follows using
the 2-fold covering S RP and the fact that S is contractible (see [82,
Example 1.B.3p. 88]). By Proposition 3.8.1 and its proof, the space Km admits a
CW-structure whose (m 1)-skeleton is a point. Thus, when m > 1, Km is simply
connected and [S i , Km ] [S i , Km ] (see [82, proposition 4A.2]). The cohomol-
ogy of S i , computed in Proposition 3.8.1, implies that the set [S i , Km ] H m (S i )
contains one element if i = m and two elements if i = m.
Conversely, if X is a CW-complex satisfying i (X ) = 0 if i = m and m (X ) =
Z2 , we must prove that X is homotopy equivalent to Km . This requires techniques
not developed in this book. When m = 1, there exists a map f : X r o K1
inducing an isomorphism on the fundamental group (see (4.3.1)). The map f then
induces an isomorphism on all the homotopy groups, what is called a weak homotopy
equivalence. By the Whitehead theorem [82, Theorem 4.5], a weak homotopy equiv-
alence between connected CW complexes is a homotopy equivalence. When m > 1,
let : S m X representing the non-zero element of m (X ). By the Hurewicz
theorem [82, Theorem 4.32], the integral homology Hm (X ; Z) = Z2 and, from the
universal coefficient theorem [82, Theorem 3B.5], it follows that Hm (X ) = Z2 and
H : Hm (S m ) Hm (X ) is an isomorphism. By Kronecker duality, H m (X ) = Z2
and H : H m (X ) H m (S m ) is an isomorphism. Let g : X Km representing
the non-zero element of H m (X ). As H : H m (X ) H m (S m ) is an isomorphism,
the map g induces an isomorphism from m (X ) to m (Km ) (we have proved above
that m (Km ) = Z2 ). Hence, g is a weak homotopy equivalence and therefore a
homotopy equivalence by the Whitehead theorem. 
3.9 Generalized Cohomology Theories 123

3.9 Generalized Cohomology Theories

The axiomatic viewpoint for (co)homology was initiated by Eilenberg and Steenrod
in the late 1940s [51, 52] and had a great impact on the general understanding of
the theory. We give below a version in the spirit of [82, Sect. 2.3 and Chap. 3]. Our
application will be the Knneth theorem 4.6.7.
A cohomology theory is a contravariant functor h from the category Top2 of
topological pairs to the category GrV of graded Z2 -vector spaces, together with a
natural connecting homomorphism : h (A) h +1 (X, A) (the notation h (A)
stands for h (A, )). In addition, the following axioms must be satisfied.
(1) Homotopy axiom: if f, g : (X, A) (X  , A ) are homotopic, then h f = h g.
(2) Exactness axiom: for each topological pair (X, A) there is a long exact sequence

h m (X, A) h m (X ) h m (A)
h m+1 (X, A)

where the unlabeled arrows are induced by inclusions. This exact sequence is
functorial, i.e. if f : (X  , A ) (X, A) is a map of pair, there is a commutative
diagram

/ h (X ) / h (A) / h +1 (X, A) / h +1 (X ) /

h f h f h f h f
   

/ h (X  ) / h (A ) / h +1 (X  , A ) / h +1 (X  ) /

(3) Excision axiom: let (X, A) be a topological pair, with U be a subspace of X

satisfying U int A. Then, the GrV-morphism induced by inclusions i :
h (X, A) h (X U, A U ) is an isomorphism.

(4) Disjoint union axiom: for a disjoint union (X, A) = jJ (X j , A j ) the homo-
morphism 
h (X, A) h (X j , A j )
jJ

induced by the family of inclusions (X j , A j )  (X, A) is an isomorphism.

Example 3.9.1 The singular cohomology H is a generalized cohomology theory.

Axioms (1)(3) are fulfilled, as seen in Sects. [Link].4. The disjoint union axiom
corresponds to Proposition 3.1.11 for a pair (X, ); it may be extended to arbitrary
topological pairs, using the exactness axiom and the five lemma.
K -theory and cobordism are other examples of generalized cohomology theories.

Let h and k be two cohomology theories. A natural transformation from h

to k is a natural transformation of functors commuting with the connecting homo-
morphisms. In particular, for each topological pair (X, A), one has a commutative
diagram of exact sequences:
124 3 Singular and Cellular (Co)homologies

/ h (X ) / h (A) / h +1 (X, A) / h +1 (X ) /

   

/ k (X ) / k (A) / k +1 (X, A) / k +1 (X ) /
(3.9.1)
The aim of this section is to prove the following theorem.
Proposition 3.9.2 Let h and k be two cohomology theories and let be a natural

transformation from h to k . Suppose that : h (pt)
k (pt) is an isomorphism.

Then : h (X, A)
k (X, A) is an isomorphism for all CW-pairs (X, A) where
X is finite dimensional.
The hypothesis that X is finite dimensional is not necessary in Proposition 3.9.2
(see [82, Proposition 3.19]), but it simplifies the proof considerably. Proposition 3.9.2
is enough for the applications in this book (see Sect. 4.6).
Proof We essentially recopy the proof of [82, Proposition 3.19]. By Diagram (3.9.1)
and the five-lemma, it suffices to show that is an isomorphism when A = .
The proof goes by induction on the dimension of X . When X is 0-dimensional, the
result holds by hypothesis and by the axiom for disjoint unions. Diagram (3.9.1)
for (X, A) = (X m , X m1 ) and the five-lemma reduce the induction step to showing
that is an isomorphism for the pair (X m , X m1 ). Let m : m (D m , S m1 )
(X, X m1 ) be a global characteristic maps for all the m cells of X. Like in the proof
of Lemma 3.4.7, the axioms (essentially excision) imply that h m and k m are
isomorphisms so, by naturality, it suffices to show that is an isomorphism for
m (D m , S m1 ). The axiom for disjoint unions gives a further reduction to the
case of the pair (D m , S m1 ). Finally, this case follows by applying the five-lemma to
Diagram (3.9.1), since D m is contractible and hence is covered by the 0-dimensional
case, and S n1 is (n 1)-dimensional. 

3.10 Exercises for Chapter 3

3.1. Give the list of the (maximal) simplexes of the triangulation of m I used in
the proof of Proposition 3.1.30. Draw them for m = 1, 2. Same question for the
triangulation used in the proof of Lemma 3.1.35.
3.2. Let X be a topological space.
(a) Show that X is contractible if and only if there is a correspondence f f
associating to a continuous map f : A X a continuous extension f :
CA X , where CA is the cone over A. This correspondence is natural in the
following sense: if g : B A is a continuous map, then  f g = f Cg.
(b) Show that if X is contractible then, for any CW-pair (A, B), any continuous
map f : B X admits a continuous extension g : A X .
3.10 Exercises for Chapter 3 125

(c) Using (a), find a direct proof of that the (co)homology of a contractible
space is isomorphic to that of a point (Corollary 3.1.33). [Hint: use that
n+1 Cn .]
3.3. Let X be a 2-sphere or a 2-torus. Let A be a non-empty subset of X containing
n points. Compute H (X A) and H (X, A).
3.4. Find topological pairs (X, Y ) and (X  , Y  ) such that H (X, Y )  H (X  , Y  )
while X is homeomorphic to X  and Y is homeomorphic to Y  .
3.5. Let X be a topological space. Let A be a subspace of X which is open and
closed. Show that (X, A) is well cofibrant.
3.6. Show that there is no continuous retraction from the Mbius band onto its
boundary.
3.7. Show that the Klein bottle K is made out of two copies of the Mbius band glued
along their common boundaries. Compute H (K ), using the Mayer-Vietoris
exact sequence for this decomposition.
3.8. Let f : S n S n be a continuous map such that no antipodal pair of points
goes to an antipodal pair of points. Show that the degree of f is 0.
3.9. Let (X, Y, Z ) be a topological triple. Draw a commutative diagram linking the
cohomology sequences of the pairs (X, Y ), (X, Z ), (Y, Z ) and that of the triple
(X, Y, Z ).
3.10. Let (X, X 1 , X 2 , X 0 ) be a Mayer-Vietoris data with X = X 1 X 2 . Suppose
that X is a CW-complex and that X i are subcomplexes. Find a short proof
of the existence of the Mayer-Vietoris for the cellular (co)homology. [Hint:
analogous to the simplicial case.]
3.11. Let (X, X 1 , X 2 , X 0 ) be a Mayer-Vietoris data with X = X 1 X 2 . Suppose
that the homomorphism H (X 1 , X 0 ) H (X , X 2 ) induced by the inclusion
is an isomorphism. Deduce the Mayer-Vietoris (co)homology sequences for
(X, X 1 , X 2 , X 0 ).
3.12. Using a tubular neighbourhood and the Mayer-Vietoris sequence, compute the
homology of the complement of a (smooth) knot in S 3 .
3.13. Let X be a countable CW-complex. Show that H (X ) is countable. Is it true
for H (X )?
3.14. For n N1 , consider the circle Cn := {z C | |z 1/n| = 1/n}. The
Hawaiian earring is the subspace B of C consisting of the union of Cn for
n 1.

(a) Show that H1 (B) surjects onto n=1 Z2 .
(b) Show that [B, RP ] is countable.
(c) Deduce from (a) and (b) that B does not have the homotopy type of a
CW-complex.

3.15. Let Rq , q N1 be a sequence of Z2 -vector spaces. Find a path-connected

space X such that Hq (X ) Rq .
3.16. Let X be a 2-dimensional CW-complex with a single 0-cell, m 1-cells and n
2-cells. Show that m = n if and only if b1 (X ) = b2 (X ).
3.17. Find perfect CW-decompositions for the 2-torus and the Klein bottle.
126 3 Singular and Cellular (Co)homologies

3.18. Let P = A|R be a presentation of a group G with a set A of generators

and a set B of relators. The presentation complex X P is the 2-dimensional
complex obtained from a bouquet C A of circles indexed by A by attaching, for
each relator r R, a 2-cell according to r 1 (C A ). Hence, 1 (X P ) G.
Compute H (X P ) in the following cases.
P1 = a, b, c | abc1 b1 , bca1 c1 and P2 = x, y | xyxy1 x 1 y 1 (two
presentations of the trefoil knot group).
P3 = a, b, c | a 5 , b3 , (ab)2 (a presentation of the alternate group A5 ).

3.19. Let K = K (Z2 , n) be an Eilenberg-MacLane space. Let f : K K be

a continuous map. Show that f is either homotopic to the identity or to a
constant map.
3.20. Let Kn = K (Z2 , n) be an Eilenberg-MacLane space. Let X be a non-
j
contractible CW-complex. Suppose that there are continuous maps X
r
Kn X such that r j is homotopic to the identity. Show that X and Kn
have the same homotopy type.
Chapter 4
Products

So far, the reader may not have been impressed by the essential differences between
homology and cohomology: the latter is dual to the former via the Kronecker pairing,
so they are even isomorphic for spaces of finite homology type. However, cohomol-
ogy is a definitely more powerful invariant than homology, thanks to its cup product,
making H () a graded Z2 -algebra. Thus, the homotopy types of two spaces with
isomorphic homology may sometimes be distinguished by the algebra-structure of
their cohomology. Simple examples are provided by RP 2 versus S 1 S 2 , or by the
2-torus versus the Klein bottle.
In this chapter, we present the cup product for simplicial and singular cohomology,
out of which the cap and cross products are derived, with already many applications
(more will come in other chapters).
Cohomology and its cup product occurred in 1935 (40 years after homology) in
the independent works of Kolmogoroff and Alexander, soon revisited and improved

by Chech and by Whitney [29, 209]. These people were all present in the international
topology conference held in Moscow, September 1935. Vivid recollections of this
memorable meeting were later written by Hopf and by Whitney [102, 211]. For
surveys of the interesting history of cohomology and products, see [40, Chap. IV]
and [137].

4.1 The Cup Product

4.1.1 The Cup Product in Simplicial Cohomology

Let K be a simplicial complex. Choose a simplicial order on K . Let a C p (K )

and b C q (K ). Using Point (c) of Lemma 2.3.3, we define a cochain a  b
C p+q (K ) by the formula

a  b,  = a, {v0 , . . . , v p } b, {v p , . . . , v p+q } ,

Springer International Publishing Switzerland 2014 127

J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_4
128 4 Products

required to be valid for all = {v0 , . . . , v p+q } S p+q (K ), with v0 < v1 < <
v p+q . This defines a map

C p (K ) C q (K ) C p+q (K ) .

We can see  as a composition law on C (K ):


C (K ) C (K ) C (K ) .

Lemma 4.1.1 (C (K ), +,  ) is a (non-commutative) graded Z2 -algebra.

Proof The associativity and distributivity properties are obvious. The neutral element
for  is the unit cochain 1 C 0 (K ). 

Lemma 4.1.2 (a  b) = a  b + a  b.

Proof Set a C p (K ), b C q (K ) and = {v0 , . . . , v p+q+1 } S p+q+1 (K ) with

v0 < v1 < < v p+q+1 . One has

a  b,  = a, {v0 , . . . , v p+1 } b, {v p+1 , . . . , v p+q+1 }

= a, {v0 , . . . , v p+1 } b, {v p+1 , . . . , v p+q+1 }

p+1
= a, {v0 , . . . , vi , . . . , v p+1 } b, {v p+1 , . . . , v p+q+1 } . (4.1.1)
i=0

In the same way,


p+q+1
a  b,  = a, {v0 , . . . , v p } b, {v p , . . . , vi , . . . , v p+q+1 } . (4.1.2)
i= p

The last term in the sum of (4.1.1) is equal to the first term in the sum of (4.1.2).
Hence, these terms cancel when adding up the two sums and the remaining terms
are those of a  b, () = (a  b), . 

Lemma 4.1.2 implies that Z (K )  Z (K ) Z (K ), B (K )  Z (K )

B (K )and Z (K )  B (K ) B (K ). Therefore,  induces a map H p (K )

H p+q (K ), seen as a composition law on H (K ):
H q (K )

H (K ) H (K )
H (K )

called the cup product on simplicial cohomology. The notation  and the name cup
product (the latter due to the former) were first used by Whitney [209]. It follows
from Lemma 4.1.1 that (H (K ), +, ) is a graded Z2 -algebra. Dropping the index
is justified by the following proposition.
4.1 The Cup Product 129

Proposition 4.1.3 The cup product on H (K ) does not depend on the simplicial
order .

Proof The procedure to define the cup product may be done with the ordered
cochains. For a C p (K ) and b C q (K ), we define a  b C p+q (K ) by
the formula

a  b,  = a, (v0 , . . . , v p ) b, (v p , . . . , v p+q ) ,

required to be valid for all (v0 , . . . , v p+q ) S p+q (K ). This defines a graded Z2 -
algebra structure on C (K ). The formula (a  b) = a  b + a  b is proven
as for Lemma 4.1.2, whence a graded algebra structure on H (K ). These definitions
imply that the isomorphism

H : ( H (K ), +, )
(H (K ), +,  )

of Sect. 2.10 is an isomorphism of graded algebras. As, by Corollary 2.10.10, H

is independent of the simplicial order , so is the cup product on H (K ). 

Corollary 4.1.4 (Commutativity of the cup product) The cup product in simplicial
cohomology is commutative, i.e. a  b = b  a for all a, b H (K ).

b Z (K ) representing a, b. Let be a simplicial order on K . In

Proof Let a,

Z (K ), one has

a  b = b  a ,

where is the opposite order of . By Proposition 4.1.3, this proves

Corollary 4.1.4. 

The commutativity of the cup product is an important feature of the mod2 coho-
mology. In other coefficients, holds true only up to signs.
Let GrA be the category whose objects are commutative graded Z2 -algebras and
whose morphisms are algebra maps. Corollary 4.1.4 says that H (K ) is an object of
GrA. There is an obvious forgetful functor from GrA to GrV.

Proposition 4.1.5 (Functoriality of the cup product) Let f : L K be a simplicial

map. Then H f : H (K ) H (L) is multiplicative: H f (a  b) = H f (a) 
H f (b) for all a, b H (K ).

Proof The proof of Proposition 4.1.3 shows that H (K ) is an object of GrA. Using
Corollary 2.10.8, it also shows that the isomorphism H : H (K ) H (K )
is a GrA-isomorphism. Let a C p (K ) and b C q (K ). Then, for all =
(v0 , . . . , v p+q ) S p+q (K ), one has

C f (a  b),  = a  b, ( f (v0 ), . . . , f (v p+q ))

= a, ( f (v0 ), . . . , f (v p )) b, ( f (v p ), . . . , f (v p+q ))
130 4 Products

= C f (a), (v0 , . . . , v p ) C f (b), (v p , . . . , v p+q )

= C f (a)  C f (b),  .

By Lemma 2.3.3, this implies that C f (a  b) = C f (a)  C f (b). We deduce

that H f : H (K ) H (L) is multiplicative. Using Proposition 2.10.11, this
implies that H f is multiplicative. 

Corollary 4.1.6 The simplicial cohomology is a contravariant functor from Simp

to GrA.

The cup product may also be defined in relative simplicial cohomology. Let L 1
and L 2 be two subcomplexes of K . For any simplicial order on K , one has

C (K , L 1 )  C (K , L 2 ) C (K , L 1 L 2 ) .

Hence, we get a map


H (K , L 1 ) H (K , L 2 )
H (K , L 1 L 2 )

which is bilinear and commutative. In particular, we get relative cup products

 
H (K , L) H (K )
H (K , L) and H (K ) H (K , L)
H (K , L)

which are related as described by the following two lemmas.

Lemma 4.1.7 Let (K , L) be a simplicial pair. Denote by j : (K , ) (K , L) the

inclusion. Let a H p (K , L) and b H q (K , L). Then, the equality

H j (a)  b = a  b = a  H j (b)

holds in H p+q (K , L).

Proof Denote also by a Z p (K , L) and b Z q (K , L) cocycles represent-

ing the cohomology classes a and b. Choose a simplicial order on K and let
= {v0 , . . . , v p+q } S p+q (K ) S p+q (L) with v0 < < v p+q . Let
1 = {v0 , . . . , v p } and 2 = {v p , . . . , v p+q }. One has

C j (a)  b,  = C j (a), 1 b, 2  = a, C j (1 )b, 2  (4.1.3)

and

a  b,  = a, 1 b, 2  . (4.1.4)

If 1 S p (L), then C j (1 ) = 0 and the right hand sides of (4.1.3) and (4.1.4)
both vanish. If 1 / S p (L), then C j (1 ) = 1 and the right hand sides
of (4.1.3) and (4.1.4) are equal. As C p+q (K , L) is the vector space with basis
4.1 The Cup Product 131

S p+q (K ) S p+q (L), this proves that H j (a)  b = a  b. The other equation
is proven similarly. 

The proof of the following lemma, quite similar to that of Lemma 4.1.7, is left to
the reader (Exercise 4.1).

Lemma 4.1.8 Let (K , L) be a simplicial pair. Denote by j : (K , ) (K , L) the

inclusion. Let a H p (K ) and b H q (K , L). Then, the equality

H j (a  b) = a  H j (b)

holds in H p+q (K ).

There is also a relationship between the relative cup product and the connecting
homomorphism of a simplicial pair.

Lemma 4.1.9 Let (K , L) be a simplicial pair. Denote by i : L K the inclusion.

Let a H p (K ) and b H q (L). Then, the equality

(b  H i(a)) = b  a

holds true in H p+q+1 (K , L).

Proof Denote also by a Z p (L) and b Z q (L) the cocycles representing the
cohomology classes a and b. Let b C q (K ) be an extension of the cochain b. The
cochain b  a C p+q (K ) is then an extension of b  C i(a). By Lemma 2.7.1,
K (b  a) Z p+q+1 (K , L) represents (b  H i(a)), where K : C (K )
C +1 (K ) is the coboundary homomorphism for K . As a is a cocycle, one has K (b 
a) = K (b)  a. By Lemma 2.7.1 again, K (b)  a represents the cohomology
class b  a. This proves the lemma. 

4.1.2 The Cup Product in Singular Cohomology

Let X be a topological space and let : m X be an element of Sm (X ). For

0 p, q m, we define p S p (X ) and q Sq (X ) by
p
(t0 , . . . , t p ) = (t0 , . . . , t p , 0 . . . , 0) and
q (t0 , . . . , tq ) = (0, . . . , 0, t0 , . . . , tq ) .

The singular simplexes p and q are called the front and back faces of . Let
a C p (X ) and b C q (X ). Using Point (c) of Lemma 2.3.3, we define a cochain
a  b C p+q (X ) by the formula

a  b,  = a, p b, q  , (4.1.5)

132 4 Products

required to be valid for all S p+q (X ). This defines a bilinear map


C p (X ) C q (X )
C p+q (X ) . (4.1.6)

The formula of Lemma 4.1.2 holds true, with the same proof. Hence, we get a cup

product in singular cohomology: H p (X ) H q (X )
H p+q (X ), giving rise to a
composition law

H (X ) H (X )
H (X ) .

Proposition 4.1.10 (H (X ), +, ) is a commutative graded Z2 -algebras.

Proof The associativity and distributivities are easily deduced from the definitions,
like for the cup product in simplicial cohomology. If X is empty, then H (X ) = 0
and there is nothing to prove. Otherwise, the neutral element for  is the class of
the unit cochain 1 H 0 (X ). Proving the commutativity directly is rather difficult.
We shall use that the singular cohomology of X is that of a simplicial complex
(see Theorem 3.7.4), together with Proposition 4.1.11, whose proof is straight-
forward. 

Proposition 4.1.11 Let K be a simplicial complex. For any simplicial order on

K , the isomorphism

R : H (|K |)
H (K )

of Theorem 3.6.3 is an isomorphism of graded algebras.

Proposition 4.1.12 The singular cohomology is a contravariant functor from TOP

to GrA.

Proof By Proposition 4.1.10, we already know that H (X ) is an object of GrA. We

also know, by Proposition 3.1.19, that H () is a contravariant functor from TOP to
GrV. It remains to prove the multiplicativity of H f : H (X ) H (Y ) for a
continuous map f : Y X . If S p+q (X ), then f p = p( f ) and f q =
( f )q . Thus, the proof that C f (a  b) = C f (a)  C f (b) is the same as for
Proposition 4.1.5. 

To get relative cup products as in simplicial cohomology, some hypothesis related

to the techniques of small simplexes (Sect. 3.1.4) is required. Let Y1 and Y2 be
subspaces of a topological space X . Let Y = Y1 Y2 and B = {Y1 , Y2 }. We say that
(Y1 , Y2 ) is an excisive couple if H (Y ) HB (Y ) is an isomorphism.

Lemma 4.1.13 A couple (Y1 , Y2 ) of subspaces of X is excisive if and only if the

inclusion (Y1 , Y1 Y2 )  (Y1 Y2 , Y2 ) induces an isomorphism in (co)homology.
4.1 The Cup Product 133

Proof Let Y = Y1 Y2 and B = {Y1 , Y2 } as above. There is a morphism

0 / C (Y, Y2 ) / C (Y ) / C (Y2 ) / 0

  
0 / C (Y1 , Y1 Y2 ) / C (Y ) / C (Y2 ) / 0
B

of short exact sequences of singular cochain complex. It induces a morphism of the

associated long exact sequences on cohomology which, by the five-lemma, implies
the result. 

Lemma 4.1.14 Let (Y1 , Y2 ) be an excisive couple of a topological space X . Then,

(4.1.6) defines a relative cup product

H (X, Y1 ) H (X, Y2 )
H (X, Y1 Y2 ) (4.1.7)

which is bilinear. The analogues of Lemmas [Link].9 hold true.

Proof Let Y = Y1 Y2 and B = {Y1 , Y2 }. Equation (4.1.5) gives a bilinear map


C (X, Y B )
C (X, Y1 ) C (X, Y2 )

where C (X, Y B ) = ker(C (X ) CB (Y )). There is a commutative diagram

H k (X ) / H k (Y ) / H k+1 (X, Y ) / H k+1 (X ) / H k+1 (Y )

    
k
H (X ) / H k (Y ) / H k+1 (X, Y B ) / H k+1 (X ) / H k+1 (Y )
B B

where the lines are exact. By the five-lemma, if H (Y ) HB (Y ) is an isomor-

phism, so is H (X, Y ) H (X, Y B ), which gives (4.1.7). The properties of the
relative cup product listed at the end of Lemma 4.1.14 are proved as in the simplicial
case. 

Remark 4.1.15 The couple (Y1 , Y2 ) is excisive in X if and only if it is excisive in

Y1 Y2 . Thus, by Proposition 3.1.34, (Y1 , Y2 ) is excisive when Y1 and Y2 are both
open. Also, (Y1 , Y2 ) is excisive when one of the subspaces Yi is contained in the
other, for instance if one is empty or if Y1 = Y2 . In some situations, the hypothesis
can be fulfilled by enlarging Yi to Yi without changing the homotopy type, and then
(4.1.7) makes sense. As in Proposition 3.1.54 and its proof, this is the case if X is
a CW-complex and Yi are subcomplexes. Note that, if (Y1 , Y2 ) is excisive, then the
Mayer-Vietoris sequence for (Y1 Y2 , Y1 , Y2 , Y1 Y2 ) holds true, by Lemma 4.1.13
and Exercise 3.11.
134 4 Products

4.2 Examples

4.2.1 Disjoint Unions

Let X be a topological space which is a disjoint union:



X= Xj .
jJ

By Proposition 3.1.11, the family of inclusions i j : X j X induce an isomorphism

in GrV
(H i j ) 
H (X ) / jJ H (X j ) .

By Proposition 4.1.12, H i j is a homomorphism of algebras for each j J . Hence,

the above map (H i j ) is an isomorphism of graded algebras.

4.2.2 Bouquets

Let (X j , x j ), with j J , be a family of well pointed spaces which are path-

connected.
 By Proposition 3.1.54, the family of inclusions i j : X j X =
jJ X j , for j J , gives rise to isomorphisms on reduced cohomology

( H i j ) 
H (X ) / H (X j ) .
jJ

The reduced and unreduced cohomologies share the same positive parts: H >0 ( ) =
H >0 ( ). As each space X j is path-connected, so is the their bouquet X . Thus
H >0 (X ) = H (X ) and we get a GrV-morphism

(H i j ) 
H >0 (X ) / H >0 (X j ) .
jJ

Being induced by continuous maps, (H i j ) is multiplicative. As X is path-connected,

this produces the GrA-isomorphism


H (X )
Z2 1 H >0 (X j ) . (4.2.1)
j|
4.2 Examples 135

When J is finite, one can also use the projections j : X X j defined in (3.1.32).
By Proposition 3.1.52, they produce a GrV-isomorphism

 H j
>0 / H >0 (X )
jJ H (X j ) (J finite) .


Being induced by continuous maps, H j is multiplicative. As X is path-
connected, this produces the GrA-isomorphism

Z2 1 H >0 (X j )
H (X ) (J finite) . (4.2.2)
jJ

4.2.3 Connected Sum(s) of Closed Topological Manifolds

A closed n-dimensional topological manifold is a compact space such that each point
has an open neighbourhood homeomorphic to Rn .
Let M1 and M2 be two closed n-dimensional topological manifolds. We suppose
that M1 and M2 are connected. Let B j M j be two embedded compact n-balls
with boundary S j . We suppose that each ball B j is nicely embedded in a bigger ball;
this implies that (M j , B j ) and (M j , S j ) are good pairs. Given a homeomorphism

h : B1
B2 , form the closed topological manifold

M = M1 h M2 = (M1 int B1 ) h (M2 int B2 ) .

The manifold M is called a connected sum of M1 and M2 . Though connected

topological manifolds are homogeneous for nicely embedded balls (see e.g.
[95, Theorem 6.7]), the homeomorphism type of M may depend on h: for exam-
ple, if h is obtained from h by precomposition with a homeomorphism of B1 which
reverses the orientation, then M1 h M2 does not have, in general, the same homotopy
type as M1 h M2 (for instance, if M1 = M2 = CP 2 , the two cases are distinguished
by their integral intersection form). In most applications in the literature, the con-
nected sum is defined for oriented manifolds and one requires that h reverses the
orientation; this makes the oriented homeomorphism type of M1 M2 well defined.
However, by Proposition 4.2.1 below, the mod 2-cohomology algebra of M1 h M2
does not depend on h, up to algebra isomorphism.
If each M j admit a triangulation |K j | M j , then K j is a connected n-
dimensional pseudomanifold (see Corollary 5.2.7). The connected sum may be
done in the world of pseudomanifolds, using n-simplexes for the balls B j . By
Proposition 2.4.4, Hn (M j ) = Z2 , generated by the fundamental class [Mi ]. The
statement Hn (M j ) = Z2 also holds for closed connected topological manifolds (see
e.g. [82, Theorem3.26]). We denote by [M j ] the generator of H n (Mi ) = Z2 .
136 4 Products

Proposition 4.2.1 Under the above hypotheses, the cohomology ring H (M1 h M2 )
is isomorphic to the quotient of Z2 1 H >0 (M1 ) H >0 (M2 ) by the ideal generated
by [M1 ] + [M2 ] :

H (M1 h M2 ) Z2 1 H >0 (M1 ) H >0 (M2 ) ([M1 ] + [M2 ] ) .

In particular, under this isomorphism, the classes [M1 ] and [M2 ] both correspond
to the fundamental class [M] of M.

Proof Form the space M = M1 h M2 and let B M be the common image of

B1 and B2 in M, with boundary S. As (M j , B j ) are good pairs, Proposition 3.1.54
M1 , M2 , B). As B has the cohomology
provides a Mayer-Vietoris sequence for ( M,
of a point, one gets a multiplicative GrV-isomorphism:

: H >0 ( M) H >0 (M1 ) H >0 (M2 ) .

As M j is connected, so is M and extend to a GrA-isomorphism

: H ( M) Z2 1 H >0 (M1 ) H >0 (M2 ) .

Let M = M1 h M2 M. The pair ( M,

M) is obviously a good pair, whence,
M) H (B, S). The non-zero
by excision and homotopy, the isomorphism H ( M,
n
part of H (B, S) is H (B, S) = Z2 . Therefore, the homomorphism : H k ( M)
k
H (M) induced by the inclusion is an isomorphism, except possibly for k = n 1
M) looks like
or n. In these degrees, the cohomology sequence of ( M,


H n1 ( M) / H n1 (M) / H n ( M,
M) / H n ( M)
/ H n (M) / 0

  
Z2 Z2 Z2 Z2

H k (M) is an isomorphism for k n 1 and the

Therefore, : H k ( M)
GrA homomorphism : H ( M) H (M) is onto. The kernel of : H n ( M)

H n (M) is of dimension 1 and, by symmetry (M1 h M2 = M2 h 1 M1 ), it must be
generated by [M1 ] + [M2 ] . 

Remark 4.2.2 If we work simplicially with pseudomanifolds, the fact that ker( :
H n (M)) contains [M1 ] + [M2 ] may be seen directly. Indeed the n-
H n ( M)
cocycle consisting of the n-simplex B j represents [M j ] by Proposition 2.4.4. Hence,
and is in ker .
the n-cocycle {B} represents [M1 ] + [M2 ] in H n ( M)
4.2 Examples 137

4.2.4 Cohomology Algebras of Surfaces

We start with the triangulation M of RP 2 drawn in Fig. 2.2, p. 26. We use the
simplicial order given by the numeration 0, . . . , 5 of the vertices. The computation
of H (M) is given in (2.4.8) and the generator of H 1 (M) = Z2 is given by the
cocycle a given in (2.4.9):
 
a = = {1, 2}, {2, 3}, {3, 4}, {4, 5}, {5, 1} S1 (RP 2 ) .

We see that, in C 2 (M),

 
a  a = {1, 2, 3}, {2, 3, 4}, {3, 4, 5} .

As a  a contains an odd number of 2-simplexes, Proposition 2.4.4 implies that

a  a is the generator [M] of H 2 (M). Therefore, one gets a GrA-isomorphism

H (RP 2 ) Z2 [a] (a 3 )

from H (RP 2 ) to the quotient of the polynomial ring Z2 [a] by the ideal generated
by a 3 . Using (4.2.1), this shows that RP 2 and S 1 S 2 do not have the same homotopy
type though they have the same Betti numbers.
Our next example is the torus T 2 . We use the triangulation given in Fig. 2.3 on
p. 27 which shows two 1-cocycles a, b C 1 (T 2 ) whose cohomology classes, again
denoted by a and b, form a basis of H 1 (T 2 ) Z2 Z2 . One checks that the following
equations hold in C 2 (T 2 ):
 
a  a = {4, 5, 6}, {5, 6, 7}, {6, 7, 8}, {7, 8, 9}
 
b  b = {2, 3, 6}, {3, 6, 8}
 
a  b = {6, 7, 8}
 
b  a = {7, 8, 9} .

In H 2 (T 2 ) = Z2 , generated by [T 2 ] , Proposition 2.4.4 implies that

aa=bb=0 and a  b = b  a = [T 2 ] .

Observe that a  b = b  a in C 2 (T 2 ), the equality only holding true in cohomol-

ogy. We get a GrA-isomorphism

H (T 2 ) Z2 [a, b] (a 2 , b2 ) .

Our third example is the Klein bottle K , using the triangulation given in Fig. 2.4
on p. 29: analogously to the case of the torus, Fig. 2.4 shows two 1-cocycles
138 4 Products

a, b C 1 (K ) whose cohomology classes, again denoted by a and b, form a basis of

H 1 (K ) Z2 Z2 . The following equations hold in C 2 (K ):
 
a  a = {4, 5, 6}, {5, 6, 7}, {6, 7, 8}, {7, 8, 9}, {4, 5, 9}
 
b  b = {2, 3, 6}, {3, 6, 8}
 
a  b = {6, 7, 8}
 
b  a = {7, 8, 9} .

In H 2 (K ) = Z2 , generated by [K ] , Proposition 2.4.4 implies

a  a = [K ] , b  b = 0 and a  b = b  a = [K ] .

Though H (T 2 ) and H (K ) are GrV-isomorphic, we see that they are not GrA-
isomorphic. Indeed, for a space X , consider the cup-square map

2
H (X ) H (X )

given by 2 (x) = x  x. Note that this map is linear, since the ground field is Z2 .
Our above computations show that 2 = 0 for X = T 2 but not for X = K . It does
not vanish either for X = RP 2 , as seen above. Now, it is classical that a connected
closed surface X is a connected sum of tori if X is orientable and a connected sum of
projective spaces otherwise. Hence, Proposition 4.2.1 implies that the orientability
of a connected surface may be seen on its cohomology algebra:

Proposition 4.2.3 Let M be a closed connected surface. Then M is orientable if

and only if its cup-square map H 1 (M) H 2 (M) vanishes.

Remark 4.2.4 As a consequence of Wus formula, we shall see in Corollary 9.8.5 that
Proposition 4.2.3 generalizes in the following way: a closed connected n-dimensional
manifold M is orientable if and only if the linear map sq1 : H n1 (M) H n (M)
vanishes.

Finally, we see that closed surfaces are distinguished by their cohomology algebra.

Proposition 4.2.5 Two closed surfaces are diffeomorphic if and only if their coho-
mology algebra are GrA-isomorphic.

Proof By Proposition 4.2.3, the cohomology algebra determines whether a closed

surface M is orientable or not. If M is orientable, then M is a connected sum of m
tori and, by Proposition 4.2.1 H 1 (M) Z2m2 . If M is not orientable, then M is a
connected sum of m projective planes and, by Proposition 4.2.1 H 1 (M) Zm 2. 
4.3 Two-Fold Coverings 139

4.3 Two-Fold Coverings

4.3.1 H 1 , Fundamental Group and 2-Fold Coverings

Let (Y, y) and (Y  , y  ) be two pointed spaces. Let [Y, Y  ] be the set of homotopy
classes of pointed maps from Y to Y  (the homotopies also preserving the base point).
Let F : [Y, Y  ] [Y, Y  ] be the obvious forgetful map.
Let (X, x) be a pointed topological space. We first define a map e : H 1 (X )
map(1 (X, x), Z2 ). Let a H 1 (X ). If c : S 1 X is a pointed map representing
[c] [S 1 , X ] = 1 (X, x), we set e(a)([c]) = H c(a) H 1 (S 1 ) = Z2 . As H c =
H c if c is homotopic to c , the map is well defined. Observe that map(1 (X, x), Z2 )
is naturally a Z2 -vector space, containing hom(1 (X, x), Z2 ) as a linear subspace.

Lemma 4.3.1 Let X be a connected CW-complex, pointed by x X 0 . Then the map

e is an isomorphism

e : H 1 (X )
hom(1 (X, x), Z2 ) .

Proof We first prove that the image of e lies in hom(1 (X, x), Z2 ). The multipli-
cation in 1 (X, x) = [S 1 , X ] may be expressed using the comultiplication : S 1
S 1 /S 0 S 1 S 1 . Then [c][c ] = [(c c ) ]. Using that H 1 (S 1 S 1 )
H 1 (S 1 ) H 1 (S 1 ) (see Proposition 3.1.51), one has
 
e(a)([c][c ]) = H e(a)([c]), e(a)([c ]) = e(a)([c]) + e(a)([c ])

for all a H 1 (X ). This proves that e([c][c ]) = e([c]) + e([c ]). The equality
e(a + b)([c]) = e(a)([c]) + e(b)([c]) is obvious, so e is a homomorphism.
Let us consider RP with its standard CW-structure of Example 3.4.5, with one
cell in each dimension, pointed by its 0-cell a. Van Kampens Theorem implies that
1 (RP , a) = Z2 . The fundamental group functor gives rise to a map

[X, RP ]
hom(1 (X, x), Z2 ) (4.3.1)

which is a bijection. Indeed, the bijectivity is established in the same way as, in Propo-
sition 3.8.3, the fact that :[X, RP ] H 1 (X ) is a bijection. The forgetful map F :
[X, RP ] [X, RP ] and the homomorphism e fit in the commutative diagram

[X, RP ]
F / [X, RP ]

(4.3.2)
 
hom(1 (X, x), Z2 ) o
e
H 1 (X )
140 4 Products

The map F is surjective: if f : X RP , any path from f (x) to a extends to

homotopy from f to a pointed map. This follows from the fact that (X, {a}) is cofi-
brant (see Proposition 3.4.1). Hence, the commutativity of Diagram (4.3.2) implies
that e (and F) are bijective. 

We now turn our attention to 2-fold coverings. The reader is assumed some famil-
iarity with the theory of covering spaces, as presented in many textbooks (see e.g.
[179, Chap. 2] or [82, Sect. 1.3]).
Let X be a connected CW-complex, pointed by x X 0 . Two covering projections
pi : X i X are equivalent if there exists a homeomorphism h : X 1 X 2 such that
p2 h = p1 . Denote by Cov2 (X ) the set of equivalence classes of 2-fold coverings
of X .
Let p : X X be a 2-fold covering. Choose x p 1 ({x}). Then p (1 ( X , x))

is a subgroup of index 2 of 1 (X, x). Let Grp2 (1 (X, x)) be the set of such
subgroups. A subgroup of index 2 being normal, the subgroup p (1 ( X , x))
does
not depend on the choice of x p 1 ({x}). We thus get a map

Cov2 (X )
Grp2 (1 (X, x))

which is a bijection (see, e.g., [82, Theorem 1.38]). For example, the trivial 2-fold
covering {1} X X corresponds to the whole group 1 (X, x) which is of index
1 2. An element H Grp2 (1 (X, x)) is the kernel of a unique homomorphism
1 (X, x) 1 (X, x)/H  Z2 . This gives a bijection

Grp2 (1 (X, x))
hom(1 (X, x), Z2 ) .

If f : X RP is a continuous map, one can form the pullback diagram

f
X / S

p p (4.3.3)
 f

X / RP .

In detail, X = {(u, z) X S | f (u) = p (z)}, with p(u, z) = u and

f(u, z) = z. We say that the covering projection p is induced from p by
the map f and write p = f p . Observe that p correspond to the subgroup
ker 1 f . Thus, homotopic maps induce equivalent coverings and we get a map
ind : [X, RP ] Cov2 (X ). These various maps, together with those of (4.3.2) sit
in the commutative diagram
4.3 Two-Fold Coverings 141

[X, RP ]
ind / Cov2 (X )
O LLL
LLL
L
LL& 
F H 1 (X ) Grp2 (1 (X, x)) (4.3.4)
PPP
PPeP
P
PPP

' 
/ hom(1 (X, x), Z2 )
[X, RP ]

The commutativity of Diagram (4.3.4) implies the following proposition.

Proposition 4.3.2 ind : [X, RP ] Cov2 (X ) is a bijection.

Let p : X X be a 2-fold covering. A continuous map f : X RP

such that p is equivalent to f p is called a characteristic map for the covering p.
Proposition 4.3.2 implies the following corollary.

Corollary 4.3.3 Let X be a connected CW-complex. Then, any 2-fold covering

admits a characteristic map. Two such characteristic maps are homotopic.

Let p : X X be a 2-fold covering. The correspondence which, over each x X ,

exchanges the two points of p 1 (x) defines a homeomorphism : X X , which is
an involution (i.e. = id) without fixed point. Also, is a deck transformation, i.e.
p = p. We call the deck involution of p. For the covering p : S RP ,
one has (z) = z.

Lemma 4.3.4 A continuous map f : X RP is a characteristic map for the

2-fold covering p : X X if and only if there exists a commutative diagram

f
X / S

p p
 f

X / RP ,

where f is a continuous map such that f (v) = f(v).

Proof Let X X be the covering induced by f (see (4.3.3)). If f is a charac-

teristic map for p, there is a homeomorphism g : X X such that p g = p.
Therefore, g satisfies g = g. As f (v) = f(v), the map f = f g satisfies
the requirements of Lemma 4.3.4. Conversely, given f, let g : X X given by
g(v) = ( p(v), f(v)). The map g satisfies p g = p and g = g. Hence, g is
surjective and is a covering projection. Since both p and p are 2-folds coverings, g
is a homeomorphism and p is equivalent to p. 
142 4 Products

Example 4.3.5 The inclusion i : RP n RP is covered by the -equivariant

map i : S n S . By Lemma 4.3.4, the map i is characteristic map for the covering
S n RP n . In particular, the identity of RP is a characteristic map for the covering
S RP .

4.3.2 The Characteristic Class

Diagram (4.3.4) together with Proposition 4.3.2 provides a bijection

w : Cov2 (X )
H 1 (X ) . (4.3.5)

This associates to a 2-fold covering p : X X its characteristic class w( p)

H 1 (X ). For instance, the characteristic class w( p ) for the covering p : S
RP is the non-zero element H 1 (RP ) = Z2 . Indeed, as H 1 (RP ) = Z2 ,
the set Cov2 (RP ) has two elements and the trivial covering corresponds to 0. As
S is connected, p is not the trivial covering, hence w( p ) = . The following
lemma is obvious.

Lemma 4.3.6 Let p : X X be a 2-fold covering over a CW-complex. Then

(1) if f : X RP is a characteristic map for the covering p, then
w( p) = H f (w( p )) = H f ().
(2) if g : Y X is a continuous map, then w(g p) = H g(w( p)).
(3) p is the trivial covering if and only if w( p) = 0.

Let us give geometric descriptions of the characteristic class w( p). Choose a

set-theoretic section b : X X of p and let B = b(X ) X . We consider B
as a singular 0-cochain of X . Using the coboundary : C 0 ( X ) C 1 ( X ), we get

B C 1 ( X ) which is the connecting 1-cochain for B: for S1 ( X ), (B), 
=1
if and only if the (non-oriented) path connects a point in B to a point in X B
(see Example 3.1.4). Observe that (B) B) = (
= (X (B)), where is the deck
involution of p. Hence

(B), 
= (B), C ()
= C (B), 

=  C (B), 
(B)), 
= (

= (B),  .

Thus, (B), depends only on p S1 (X ). This permits us to define a singular

1-cochain wb ( p) C 1 (X ) by the formula

wb ( p),  = (B), 

where S1 ( X ) is any lifting of S1 (X ).

4.3 Two-Fold Coverings 143

Proposition 4.3.7 wb ( p) is a 1-cocycle representing w( p) H 1 (X ).

Proof Let wb = wb ( p). Let 2 S2 (X ). If 2 S2 ( X ) is a lifting of 2 , then the

2 ) are liftings of those in (2 ). Therefore
1-simplexes in (

(wb ), 2  = wb , (2 ) = (B), 2 ) = 0 ,
( (4.3.6)

which proves that wb is a cocycle.

We next prove that the cohomology class wb H 1 (X ) of wb does not depend
on the set-theoretic section b. Let b : X X  another such section, giving B  =
b (X ) C 0 ( X ). Define r C 0 (X ) by

+ B  , x
r, {x} = B, x ,

where x is a chosen element in p 1 ({x}). If x is another choice, one has B, x =

B, x = B  , x
+ 1 and B  , x + 1 in Z2 , so r is well defined. Let S1 (X ) with
end points u and v. Let S1 ( X ) be a lifting of with end points u and v.
Then

wb + wb ,  = (B)  ), 
+ (B
+ B  , u
= B, u + B  , v
+ B, v
= (r ),  .

This proves that wb = wb + (r ) and thus [wb ] = [wb ]. Denote by w( p) H 1 (X )

the cohomology class [wb ].
We can now prove that w( p) = w( p  ) if p  : X  X is a 2-fold covering

equivalent to p. Indeed, if h : X X  is a homeomorphism such that p  h = p,

then, wb ( p) = wh b ( p ), which implies that w( p) = w( p  ).

Choosing a characteristic map f : X RP for p, we now have w( p) =
w( where p : X X is the induced covering of Diagram (4.3.3). Choose
p),
a set-theoretic section b0 : RP S of p and set B0 = p (RP ) and
w 0 = [wb0 ( p )] H 1 (RP ). This gives rise to a set-theoretic section b of p
by the formula b(x) = (x, b0 f (x)). It satisfies B = b(X ) = f1 (B0 ), where

f : X S is the map covering f , as in (4.3.3). Let S1 (X ) with a lifting
S1 ( X ). Then, f is a lifting of f in S1 (S ) and we have

C f (wb0 ),  = wb0 , f  = 0 (B0 ), f 

= 0 (B0 ), C f()
= C f 0 (B0 ), 

=  C f(B0 ), 
f1 (B0 )), 
= ( B),
= ( 
= wb ( p),
 .

Hence,

w( = C f (w 0 ) .
p) (4.3.7)
144 4 Products

Together with Lemma 4.3.6, Eq. (4.3.7) reduces the proof of Proposition 4.3.7, to
showing that w 0 = w( p ). As 0 = w( p ) H 1 (RP ) = Z2 , it is enough to
prove that w 0 = 0. By Eq. (4.3.7) again, it is enough to find a covering q : Y Y
over some CW-complex Y for which w(q) = 0.
We take for q the double covering q : S 1 S 1 . Let : 1 S 1 given by
(t, 1 t) = e2it . As (0,
1) = (1,
0), one has w(q),
 = 0. But is a 1-
cocycle representing the generator of H 1 (S 1 ) = Z2 (see Corollary 3.2.4). Hence,
w(q)
= 0. 

Remark 4.3.8 Let p : X X be a two-fold covering over a CW-complex X . To

describe the characteristic class w( p) in the cellular cohomology of X , choose a
section b : X 0 X and see B = b(X 0 ) X 0 as a cellular 0-cochain of X (for the
cellular decomposition induced from that of X ). Let : 1 (X ) I X be a global
characteristic maps for the 1-cells of X . Let : 1 (X ) I X be the lifting of
for which (, 0) B. Consider the cellular 1-cochain w B C 1 (X ) defined, for
1 (X ), by

1 if (, 1) / B
w B , 
0 otherwise.

Note that w B ,  = (B), where is any one-cell of X above , which, as

in (4.3.6), proves that w B is a cellular cocycle. We claim that [w B ] H 1 (X ) cor-
responds to w( p) H 1 (X ), under the identification of H 1 (X ) and H 1 (X ) as the
same subgroup of H 1 (X 1 ) (see (3.5.5)). We can thus suppose that X = X 1 . We can
also suppose that X is connected. If T is a maximal tree of X 1 , then the quotient map
X 1 X 1 /T is a homotopy equivalence by Proposition 3.4.1 and Lemma 3.1.44.
The covering p is then induced from one over X 1 /T , so we can assume that X 1
is a bouquet of circles indexed by 1 . For each one cell , a characteristic map
: D 1 X 1 gives a singular 1-simplex of X 1 (identifying D 1 with 1 ). If is
a lifting of , one has

1 if is a loop
w B ,  = wb ,  = (4.3.8)
0 otherwise,

where b : X X is a set theoretic section of p extending b. As {[ ] | 1 }

is a basis for H 1 (X 1 ), Eq. (4.3.8) implies that [w B ] H 1 (X 1 ) corresponds to
w( p) H 1 (X ).

4.3.3 The Transfer Exact Sequence of a 2-Fold Covering

Let p: X X be a 2-fold covering projection with deck involution . To each

singular simplex : m X , one can associate the set of the two liftings of
4.3 Two-Fold Coverings 145

into X . This defines a map from Sm (X ) to Cm ( X ), extending to a linear map

tr : Cm (X ) Cm ( X ). The map tr is clearly a chain map. By Sect. 2.3, this gives
risen to two GrV-morphisms

tr : Hm (X ) Hm ( X ) and tr : H m ( X ) H m (X )

satisfying tr (a),  = a, tr (). The linear maps tr and tr are called the transfer
homomorphisms for the covering p.
The transfer homomorphism in cohomology and the characteristic class w( p)
H 1 (X ) are related by the following exact sequence.

Proposition 4.3.9 (Transfer exact sequence) The sequence

H p tr w( p) H p
H m (X ) H m ( X ) H m (X ) H m+1 (X )

is exact. It is functorial with respect to induced coverings.

Proof The sequence

tr C p
0 C (X ) C ( X ) C (X ) 0 .

is clearly an exact sequence of chain complexes and it is functorial with respect to

induced coverings. By Kronecker duality, it gives a short exact sequence of cochain
complexes
C p
0 C (X ) C ( X ) C (X ) 0 .
tr
(4.3.9)

By Proposition 2.6.2, this gives rise to a connecting homomorphism d : H (X )

H +1 (X ) and a functorial long exact sequence

H p H p
H m (X ) H m ( X ) H m (X ) H m+1 (X ) .
tr d

It just remains to identify d with w( p)  .

To construct the connecting homomorphism d we need a GrV-section of tr in
Sequence (4.3.9). Choose a set-theoretic section b : X X of p. If : m X is
a singular 1-simplex of X , define b () : m X to be the unique lifting of with
b ()(1, 0, . . . , 0) b(X ). This defines a map b : S(X ) S( X ). If a C m (X ),
we consider a as a subset of Sm (X ) and so its direct image b (a) Sm ( X ) is an
m-cochain of X . This determines a GrV-morphism b : C (X ) C ( X ) which is
a section of tr . By Eq. (2.6.2), the connecting homomorphism d is determined by
the equation

C p d (a),  =  b (a),  , (4.3.10)

146 4 Products

for all a C m (X ) and Sm ( X ), where : C ( X ) C +1 ( X ) is the coboundary.

The equality C p d (a),  = d (a), C p() together with (4.3.10) shows that
 b (a),  depends only on C p() = p = . Therefore, by taking
instead of if necessary, we may assume that / b (Sm (X )). Then, the faces i
of are not in b (Sm (X )), except possibly for i = 0 and

 b (a),  = b (a), () = b (a), 0  . (4.3.11)

The number b (a), 0  equals 1 if and only if (0, 1, 0, . . . , 0) b(X ) and

0 a. In other words, if and only if the front and back faces of satisfy 1
wb ( p) (see Sect. 4.3.2) and m1 a. Hence, Eqs. (4.3.10) and (4.3.11) imply that

= b (a), 0  = wb ( p)  a, 
d (a), 

for all a C m (X ) and Sm (X ). This proves that d () = wb ( p)  in C (X ).

By Proposition 4.3.7, this implies that d () = w( p)  in H (X ). 

An important application of the transfer exact sequence is the determination of

the cohomology ring of RP n .

4.3.4 The Cohomology Ring of R P n

Let Z2 [a] be the polynomial ring over a formal variable a in degree 1. This is an
object of GrA, as well as its truncation Z2 [a]/(a n+1 ), the quotient of Z2 [a] by the
ideal generated by a n+1 . By Proposition 3.8.3, H 1 (RP ) = Z2 , generated by the
class . Therefore, there is a GrA-morphism Z2 [a] H (RP ) sending a k to k ,
where the latter denotes the cup product of k copies of . The composition Z2 [a]
H (RP ) H (RP n ) factors by a GrA-morphism Z2 [a]/(a n+1 ) H (RP n ).

Proposition 4.3.10 The above GrA-morphisms

Z2 [a] H (RP ) and Z2 [a]/(a n+1 ) H (RP n )

are GrA-isomorphisms. In particular, the GrA-morphism H (RP ) H (RP n ),

induced by the inclusion, is surjective.

Proof As S is contractible [82, Example 1.B.3 p. 88], the transfer exact sequence
of the covering p : S RP shows that the cup product with w( p )

H 1 (RP ) gives an isomorphism H (RP )
H +1 (RP ). In particular, w( p )
is the generator of H (RP ). This proves the statement for RP .
1

For RP n we use the covering p : S n RP n . The transfer exact sequence

proves at once that the cup product with w( p) H 1 (RP n ) gives an isomorphism

H m (RP n )
H m+1 (RP n ) for 0 m < n 1. As RP n has a CW-structure with
4.3 Two-Fold Coverings 147

one k-cell for 0 k n, the end of the transfer exact sequence of the covering
p : S n RP n involves the Z2 -vector spaces

w( p) H p tr
0 H n1 (RP n ) H n (RP n ) H n (S n ) H n (RP n ) 0
           
dim=1 dim1 dim=1 dim1

Thus, the cup product with w( p) is also an isomorphism H n1 (RP n )
H n (RP n ).
This proves the proposition for RP . n 

4.4 Nilpotency, Lusternik-Schnirelmann Categories

and Topological Complexity

Let X be a topological space. A subspace U of X is categorical if the inclusion

U  X is homotopic to a constant map. The Lusternik-Schnirelmann category
cat (X ) is the minimal cardinality of an open covering of X with categorical sub-
spaces. Some authors (see, e.g. [35]) adopt a different normalization for the Lusternik-
Schnirelmann category, equal to one less than the definition above.
For a survey paper about the Lusternik-Schnirelmann category, see [108]. Amongst
its properties, cat (X ) is an invariant of the homotopy type of X . For example,
cat (X ) = 1 if and only if X is contractible and cat (S n ) = 2. More generally, one
has the following result (see [108, Proposition 1.2] for a more general statement and
a different proof).

Proposition 4.4.1 Let X be a connected CW-complex of dimension n. Then

cat (X ) n + 1.

Proof By induction on the dimension of X , the statement being obvious if dim X = 0.

By Proposition 3.4.1, we can write X as the union of two open sets X = C Z ,
where C is the disjoint union of the open n-cells of X and Z retracting by defor-
mation on X (n1) . By induction hypothesis, Z admits an open covering with n
categorical subspaces. If n 1, C is a categorical open set of X , which proves that
cat (X ) n + 1. 

Let X be a topological space and B be a vector subspace of H (X ). The nilpotency

class nil B of B is the minimal integer m such that

 B = 0 .
B  
m

If no such integer exists, we set nil B = .

Proposition 4.4.2 Let X be a topological space. Then nil H >0 (X ) cat (X ).

148 4 Products

Proof Let U1 , . . . , Um X be open subspaces of X which are categorical. By the

homotopy property, the homomorphism H >0 (X ) H >0 (Ui ) induced by the inclu-
sion vanishes. Hence, the exact sequence of the pair (X, Ui ) implies that the restric-
tion homomorphism H >0 (X, Ui ) H >0 (X ) is surjective. Then, in the diagram


m 
m
H >0 (X, Ui ) // H >0 (X )
i=1 i=1
,
 
 
H >0 (X, U1 Um ) / H (X )

which is commutative by the functoriality of the cup product, the upper horizontal
arrow is surjective. But, if X = U1 Um , the lower left vector space vanishes.
This proves that nil H >0 (X ) m. 

Corollary 4.4.3 The Lusternik-Schnirelmann category cat (RP n ) of RP n is equal

to n + 1.

Proof As seen in Example 3.4.5, the projective space RP n is a CW-complex of

dimension n. Therefore, cat (RP n ) n + 1 by Proposition 4.4.1. On the other hand,
by Proposition 4.4.2, cat (RP n ) nil H >0 (RP n ) and nil H >0 (RP n ) = n + 1 by
Proposition 4.3.10. 

Another classical consequence of Proposition 4.4.2 is the vanishing of the cup

products in a suspension.

Corollary 4.4.4 Let Y be a topological space. Then, all cup products in H >0 (Y )
vanish.

Proof As Y is the union of two cones, cat (Y ) 2, which proves the
corollary. 

The Lusternik-Schnirelmann category admits several generalizations, for instance

the category of a map (see, e.g. [108, Sect. 7]). Here, we introduce the category
cat (X, A) of a topological pair (X, A). A subspace U of a topological space X is
A-categorical if the inclusion U  X is homotopic to a map with value in A. Then
cat (X, A) is defined to be the minimal cardinality of an open covering of X with
A-categorical subspaces. For instance, X is path-connected, then

cat (X, A) cat (X ) = cat (X, pt) (4.4.1)

Lemma 4.4.5 cat (X, A) is an invariant of the homotopy type of the pair (X, A).

Proof Let f : (X, A) (X  , A ) be a homotopy equivalence of pair. It suffices

to prove that, if U  is a subset of X  which is A -categorical, then U = f 1 (U  )
4.4 Nilpotency, Lusternik-Schnirelmann Categories and Topological Complexity 149

is A-categorical in X . This will imply that cat (X, A) cat (X  , A ). Homotopy

equivalence being an equivalence relation, we also get cat (X  , A ) cat (X, A).
Let g : (X  , A ) (X, A) be a homotopy inverse of f . Let t : U  X  be a
homotopy with 0 (v) = v and 1 (U  ) A . Then, the map t : U X defined
by t (u) = g t f (u) is a homotopy satisfying 0 (u) = g f (u) and 1 (U ) A.
As g f is homotopic to id(X,A) , this proves that U is A-categorical. 
Proposition 4.4.2 generalizes in the following statement.
Proposition 4.4.6 Let (X, A) be a topological pair with X path-connected. Then

nil B cat (X, A) ,

where B = ker{H (X ) H (A)} .

Proof As X is path-connected, B H >0 (X ). Let U1 , . . . , Um X be open sub-
spaces of X which are A-categorical. Then, the homomorphism H (X ) H (Ui )
factors:

H (X ) / H (Ui )
JJJ 9
JJJ
JJ
%
H (A)

Therefore, if a B, then a is in the image of H >0 (X, Ui ). The proof of

Proposition 4.4.6 is then the same as that of Proposition 4.4.2. 
This category for pairs is related to the topological complexity, a notion of mathe-
matical robotics introduced by Farber [54, 55]. Let Y be a topological space and PY
be the space of continuous paths : I Y , endowed with the compact-open topol-
ogy. Let : PY Y Y be the origin-end map: () = ((0), (1)). A motion
planning algorithm is a section of . It is not possible to find a continuous motion
planning algorithm unless Y is contractible [54, Theorem 1]. The topological com-
plexity TC (Y ) is the minimal cardinality of an open covering U of Y Y such that
: PY Y Y admits a continuous section over each U U. Let Y be the diago-
nal subset of Y Y . The following proposition is the contents of [55, Corollary 18.2].
Proposition 4.4.7 TC (Y ) = cat (Y Y, Y ).
In consequence, TC (Y ) is an invariant of the homotopy type of Y .
Proof Let U Y Y . Suppose that a continuous section s : U PY of
exists. Then (y, y  , t) = (s(y, y  )(t), y  ) satisfies (y, y  , 0) = (y, y  ) and
(y, y  , 1) = (y  , y  ) Y , showing that U is Y -categorical. Conversely, if
c(t) = (c1 (t), c2 (t)) Y Y is a path from (y, y  ) to (u, u) Y , then the path
c1 c21 joins y to y  . This process being continuous in (y, y  ), it provides a sections
of over Y -categorical subsets of Y Y . 
150 4 Products

Proposition 4.4.7 together with (4.4.1) implies that TC (Y ) cat (Y Y ). The

inequality cat (Y ) TC (Y ) also holds true [55, Lemma 9.2], but is not a conse-
quence of Proposition 4.4.7.
If Y is path-connected, Propositions 4.4.7 and 4.4.6 give the inequality

TC (Y ) nil ker H j , (4.4.2)

where j : Y Y Y denotes the inclusion. We shall see with the Knneth

theorem that there is a commutative diagram (see Remark 4.6.1):

H j
H (Y Y ) / H (Y )

. (4.4.3)
 
 / H (Y )
H (Y ) H (Y )

(to use the Knneth theorem, we need that Y is of finite cohomology type). Under
the cross product, the image of ker H j in the ring H (Y ) H (Y ) is the ideal of
the divisors of zero for the cup product. The inequality (4.4.2) thus corresponds to
[54, Theorem 7].
For results concerning the topological complexity of the projective space, see the
end of Sect. 6.2.2.

4.5 The Cap Product

Let K be a simplicial complex. Choose a simplicial order on K . We define the cap
product

C p (K ) Cn (K ) Cn p (K )

to be the unique bilinear map such that

a  {v0 , . . . , vn } = a, {v0 , . . . , v p } {v p , . . . , vn } (4.5.1)

for all a C p (K ) and all {v0 , . . . , vn } Sn (K ), with v0 < v1 < < vn (this
makes sense if n p; otherwise, the cap product just vanishes). If a C p (K ),
b C n p (K ) and Cn (K ) the following formula follows directly from the
definitions

a  b,  = b, a   . (4.5.2)

Lemma 4.5.1 If a C p (K ) and Cn (K ), then

(a  ) = (a)  + a  () .
4.5 The Cap Product 151

Proof Let q = n p and b C q (K ). Denote  and  by just  and .

Using (4.5.2), one has

(a  b),  = (a  b, () = b, a  () . (4.5.3)

In the other hand

(a  b),  = (a)  b,  + a  (b),  (4.5.4)

= b, (a)   + (b), a  
= b, (a)   + b, (a  ) .

Equations (4.5.3) and (4.5.4) imply that

b, (a  ) = b, (a)  + a  () .

for all b C q (K ). By Lemma 2.3.3, this implies Lemma 4.5.1. 

Lemma 4.5.1 implies that Z (K )  Z (K ) Z (K ), B (K )  Z (K )

B (K ) and Z (K )  B (K ) B (K ). Therefore,  induces a map H p (K )

Hn (K ) Hn p (K ), or


H (K ) H (K )
H (K )

called the cap product (on simplicial cohomology). As in the case of the cup product
we drop the index from the notation because of the following proposition.

Proposition 4.5.2 The cap product on H (K ) H (K )
H (K ) does not depend
on the simplicial order .

Proof Let and  be two simplicial orders on K . Let a H p (K ) and Hn (K ).

For any b H n p (K ), Formula (4.5.2) and Proposition 4.1.3 imply that

b, a   = a  b, 
= a  b, 
= b, a   .

By Part (b) of Lemma 2.3.3, this implies that a  = a  in Hn p (K ). 


Proposition 4.5.3 The cap product H (K ) H (K )
H (K ) endows H (K )
with a structure of H (K )-module.

Proof By definition,  is bilinear and the equality 1  = is obvious. It remains

to prove that

(a  b)  = a  (b  ) (4.5.5)
152 4 Products

for all a H p (K ), b H q (K ) and Hn (K ). As the cup product is associative

and commutative (Corollary 4.1.4), one has, for any c H n pq (K ),

c, (a  b)   = c  (a  b), 
= (c  a)  b, 
= c  a, b  
= c, a  (b  ) .

By Lemma 2.3.3, this proves Eq. (4.5.5). 

Proposition 4.5.4 (Functoriality of the cap product) Let f : L K be a simplicial

map. Then, the formula

a  H f () = H f (H f (a)  )

holds in H (K ) for all a H (K ) and all H (L).

Proof Suppose that a H p (K ) and Hn (L). Using the functoriality of the cup
product established in Proposition 4.1.5, one has, for any b H n p (K ),

b, a  H f () = a  b, H f ()

= H f (a  b), 
= H f (a)  H f (b), 
= H f (b), H f (a)  
= b, H f (H f (a)  ) .

By Part (b) of Lemma 2.3.3, this proves Proposition 4.5.4. 

There are several version of the cap product in relative simplicial (co)homology.
Let (K , L) be a simplicial pair. Choose a simplicial order on K . We note first that
C (K )  C (L) C (L), whence a cap product

H p (K ) Hn (K , L)
Hn p (K , L) . (4.5.6)

One may also compose

j id 
H p (K , L) Hn (K ) H p (K ) Hn (K )
Hn p (K )

to obtain a cap product


H p (K , L) Hn (K )
Hn p (K ) . (4.5.7)
4.5 The Cap Product 153

The latter cap product may be post-composed with Hn p (K ) Hn p (K , L) and

get a cap product

H p (K , L) Hn (K )
Hn p (K , L) . (4.5.8)


As the restriction of C p (K ) Cn (K )
Cn p (K ) to C p (K , L) Cn (L) vanishes,
we obtain a cap product

H p (K , L) Hn (K , L)
Hn p (K ) . (4.5.9)

As for Formula (4.5.5), the equation

(a  b)  = a  (b  ) (4.5.10)

holds true in Hn pq (K ) for all a H p (K , L), b H q (K , L) and Hn (K , L).

The cap products (4.5.7) and (4.5.9) are used in (4.5.10).
More generally, suppose that L is the union of two subcomplexes L = L 1 L 2 .

Then, the restriction of C p (K ) Cn (K )
Cn p (K ) to C p (K , L 1 ) Cn (L) has
image contained in Cn p (L 2 ). This gives a cap product


H p (K , L 1 ) Hn (K , L)
Hn p (K , L 2 ) . (4.5.11)

The functoriality holds for a simplicial map f : (K  , L  ) (K , L) satisfying

f (L i ) L i for i = 1, 2: the formula

a  H f () = H f (H f (a)  ) (4.5.12)

holds in H (K , L 2 ) for all a H (K , L 1 ) and all H (K  , L  ). The proof is the

same as for Proposition 4.5.4.
The next two lemmas express the compatibility between these relative cap
products, the absolute one and the connecting homomorphisms for a simplicial
pair (K , L).

Lemma 4.5.5 Let (K , L) be a simplicial pair. Denote by i : L K and j :

(K , ) (K , L) the inclusions. Let x Hn (K , L). Then, for all integer p, the
diagram

H j H i
H p (K , L) / H p (K ) / H p (L) / H p+1 (K , L)

x x  x x
   
H j
Hn p (K ) / Hn p (K , L) / Hn p1 (L) H i
/ Hn p1 (K )

is commutative.
154 4 Products

Proof For the left hand square diagram, let a H p (K , L) and b H n p (K , L).
One has
b, H j (a)  x = b  H j (a), x

and

b, H j (a  x) = H j (b), a  x = H j (b)  a, x .

Hence, the left hand square diagram commutes if and only if b  H j (a) =
H j (b)  a, which was established in Lemma 4.1.7.
For the middle square diagram, let a H p (K ) and b H n p1 (L). One has

b, H i(a)  x = b  H i(a), x =  (b  H i(a)), x .

On the other hand:

b, (a  x) =  (b), a  x =  (b)  a, x .

The commutativity of the middle square diagram is thus equivalent to the formula
(b  H i(a)) = (b)  a holding true in H n p (K , L) for all a H p (K )
and b H n p1 (L). This formula was proven in Lemma 4.1.9. In the same way,
we see that the commutativity of the right hand square diagram is a consequence of
Lemma 4.1.9 (intertwining the role of a and b). 

Lemma 4.5.6 Let (K , L) be a simplicial pair. Denote by j : (K , ) (K , L) the

pair inclusion. Then, the equation

H j (a  ) = a  H j () .

holds true in Hn p (K , L) for all a H p (K ) and all Hn (K ).

Proof It is then enough to prove that b, H j (a  ) = b, a  H j () for all
b H n p (K , L). But,

b, H j (a  ) = H j (b), a  
= H j (b)  a, 
= H j (b  a),  by Lemma 4.1.8
= b  a, H j ()
= b, a  H j () . 

The cap product is also defined in the singular (co)homology of a space X . On

the (co)chain level, it is the unique bilinear map

C p (X ) Cn (X )
Cn p (X )
4.5 The Cap Product 155

such that

a  = a, p q

for all a C p (X ) and all Sn (X ), where the back and front faces p and q are
defined as in p. 131. If a C p (X ), b C n p (X ) and Cn (X ) the following
formula follows directly from the definition

a  b,  = b, a   . (4.5.13)

Therefore, as for the simplicial cap product, properties follows from those of the
cup product. The formula (a  ) = (a)  + a  () is proved as for

Lemma 4.5.1 and we get an induced bilinear map H p (X ) Hn (X ) Hn p (X ), or


H (X ) H (X )
H (X )

called the cap product in singular (co)homology. This cap product endows H (X )
with a structure of H (X )-module, as in Proposition 4.5.3 and is functorial for con-
tinuous maps f : Y X , as for Proposition 4.5.4.
For a topological pair (X, Y ), the three relative versions of the cap products:

H p (X ) Hn (X, Y )
Hn p (X, Y ) , (4.5.14)


H p (X, Y ) Hn (X )
Hn p (X, Y ) (4.5.15)

and

H p (X, Y ) Hn (X, Y )
Hn p (X ) (4.5.16)

hold true, as for (4.5.6)(4.5.9). When Y = Y1 Y2 , a relative cap product analogous

to (4.5.11)

H p (X, Y1 ) Hn (X, Y )
Hn p (X, Y2 ) . (4.5.17)

is available under some conditions, for instance if (Y, Yi ) is a good pair for i = 1, 2,
so one can use the small simplexes technique, as for the Mayer-Vietoris sequence
in Proposition 3.1.54. The functoriality formula (4.5.12) as well as the analogues of
Lemmas 4.5.5 and 4.5.6 hold true.
Finally, the simplicial and singular cap products are intertwined by the
isomorphisms

R, : H (K )
H (|K |) and R : H (|K |)
H (K )
156 4 Products

of Theorem 3.6.3. For any simplicial order , the equation

a  R, () = R, (R (a)  ) (4.5.18)

holds in H (|K |) for all a H (|K |) and all H (K ): The proof of (4.5.18) is
straightforward for a simplex of K .

4.6 The Cross Product and the Knneth Theorem

Let X and Y be topological spaces. Results computing H (X Y ) in terms of H (X )

and H (Y ) are known as Knneth theorems (or Knneth formulas). This generic name
comes from the thesis of Hermann Knneth in 1923 (see [40, pp. 5556]). To give an
example, when X and Y are discrete spaces, the cohomology rings are concentrated
in dimension 0 and

H 0 (X ) = Z2X , H 0 (Y ) = Z2Y , H 0 (X Y ) = Z2 X Y . (4.6.1)

The cross product of maps

Z2 X Y
Z2X Z2Y

defined by ( f g)(x, y) = f (x)g(y) is bilinear. The associated linear map

Z2 X Y
Z2X Z2Y (4.6.2)

is also called the cross product. The map (4.6.2) is clearly injective. It is not surjective
if both X and Y are infinite; for instance, if X = Y is infinite, it is easy to see that
the characteristic function of the diagonal in X X is not in the image of . On the
other hand, suppose that X or Y is finite (say Y ). Let F : X Y Z2 . For y Y ,
define Fy : X Z 2 by Fy (x) = F(x, y) and let y be the characteristic function
of {y}. Then

F= Fy y .
yY

Thus, if Y is finite, the cross product of (4.6.2) is an isomorphism.

Such finiteness conditions will occur in the statements of this section, under the
form that Y should be of finite cohomology type (see Definition p. 88).
Observe that, under the identification of (4.6.1), the cross product H 0 (X )
H (Y ) H 0 (X Y ) satisfies the formula
0

f g = X f  Y g ,

where X and Y are the projections of X Y onto X and Y .

4.6 The Cross Product and the Knneth Theorem 157

More generally, let X and Y be two topological spaces. Using the usual tensor
product of vector spaces over Z2 , we define the tensor product of the Z2 -algebras
(H (X ), +, ) and (H (Y ), +, ) as the Z2 -algebra (H (X ) H (Y ), +, )
defined by

[H (X ) H (Y )]m = H i (X ) H j (Y ) ,
i+ j=m

with the product

(a1 b1 ) (a2 b2 ) = (a1  a2 ) (b1  b2 ) . (4.6.3)

The projections X : X Y X et Y : X Y Y give GrA-morphisms

X : H (X ) H (X Y ) et Y : H (Y ) H (X Y ). This permits us to
define a bilinear map

H (X ) H (Y )
H (X Y )

by

a b = (a, b) = X (a)  Y (b) (4.6.4)

called the cross product. By the universal property of the tensor product (analogous
to that for vector spaces), this gives a GrV-morphism

H (X ) H (Y )
H (X Y ) ,

also called the cross product.

Remark 4.6.1 Let  : X X X be the diagonal map (x) = (x, x). The
composition

H (X ) H (X )
H (X X ) H (X ) (4.6.5)

is equal to the cup product (see also Diagram (4.4.3)) This relation, due to Lefschetz
(see [183, pp. 3841] for historical considerations), was quite influential: in some
books (e.g. [136, 179]), the cross product is introduced first using homological algebra
(the Eilenberg-Zilber theorem) and the cup product is defined via Formula (4.6.5).
Our opposite approach follows the viewpoint of [74, 82].

Under some hypotheses, the cross product may be defined in relative cohomol-
ogy. Let (X, A) and (Y, B) be topological pairs. The projections X and Y give
homomorphisms X : H (X, A) H (X Y, A Y ) and Y : H (Y, B)
H (X Y, B X ). Suppose that A or B is empty, or one of the pairs (X, A) or
158 4 Products

(Y, B) is a good pair. Then formula (4.6.4) defines a relative cross product

H (X, A) H (Y, B)
H (X Y, A Y X B) . (4.6.6)

Indeed, we must just check that the relative cup product


H (X Y, A Y ) H (X Y, B X )
H (X Y, A Y X B) .

is defined. By Lemma 4.1.14, it is enough to show that (A Y, X B) is excisive in

X Y . This is obvious if A or B is empty. Otherwise, suppose that one of the pair,
say (Y, B), is a good pair. Let V be a neighbourhood of B in Y which retracts by
deformation onto B. Let Z = A Y X B. Then, A (Y V ) int Z (A Y ).
By excision of A (Y V ) and homotopy, we get isomorphisms

H (Z , A Y )
H (A V X B, A V )
H (X B, A B) .

By Lemma 4.1.13, this implies that (A Y, X B) is excisive.

We first establish the functoriality of the cross product. In Lemmas 4.6.2 and 4.6.3
below, we assume that the conditions for the relative cross product to be defined are
satisfied.

Lemma 4.6.2 Let f : (X  , A ) (X, A) and g : (Y  , B  ) (Y, B) be maps of

pairs. Then, for all a H (X, A) and b H (Y, B) the following formula holds:

H ( f g)(a b) = H f (a) H g(b) .

Proof As X ( f g) = f X  and Y ( f g) = g Y  , one has

 
H ( f g)(a b) = H ( f g) H X (a)  H Y (b)
= H ( f g)(H X (a))  H ( f g)(H Y (b)))
= H X  H f (a)  H Y  H g(b)
= H f (a) H g(b) . 

Formula (4.6.3) provides a product on H (X, A) H (Y, B).

Lemma 4.6.3 The cross product H (X, A) H (Y, B) H (X Y, AY X

B) is multiplicative. In particular, the cross product H (X ) H (Y )
H (X Y )
is a GrA-morphism.

Proof
 
(a1 b1 ) (a2 b2 ) = (a1  a2 ) (b1  b2 )
= X (a1  a2 )  Y (b1  b2 )
= X (a1 )  X (a2 )  Y (b1 )  Y (b2 )
4.6 The Cross Product and the Knneth Theorem 159

= X (a1 )  Y (b1 )  X (a2 )  Y (b2 )

= (a1 b1 )  (a2 b2 )
= (a1 b1 )  (a2 b2 ) 

Remark 4.6.4 In the proof of Lemma 4.6.3, we have established that

(a1  a2 ) (b1  b2 ) = (a1 b1 )  (a2 b2 )

for all ai H (X, A) and b j H (Y, B).

Observe that the Kronecker pairing

,
[H (X ) H (Y )] [H (X ) H (Y )] Z2

given by

a b,  = a, b,  (4.6.7)

is a bilinear map (By convention, a,  = 0 if a H p () and Hq () with

p = q.)

Lemma 4.6.5 Let X and Y be topological spaces with Y of finite cohomology type.
Then, for all n N, the linear map
k
H p (X ) H q (Y )
[ H p (X ) Hq (Y )]
p+q=n p+q=n

given by k(a b) = a b,  is an isomorphism.

Proof It suffices to prove that k : H p (X ) H q (Y ) [H p (X ) Hq (Y )] is an

isomorphism for all integers p, q. As H r () Hr () via the Kronecker pairing,
this amounts to prove that, for vector spaces V and W , the homomorphism

k : V  W  [V W ] ,

given by k(r s)(v w) = r (v)s(w), is an isomorphism when W is finite dimen-

sional. This classical fact (true over any base field) is easily proven by induction
on dim W (see, e.g., [43, Chap. VI, Proposition 10.18] for a proof in a more general
setting). 

The following lemma permits us to define a Kronecker dual to the cross product,
called the homology cross product.
160 4 Products

Lemma 4.6.6 Let X and Y be topological spaces, with Y of finite cohomology type.
Then, there exists a unique GrV-homomorphism

: H (X Y ) H (X ) H (Y )

such that the equation

a b,  = a b, () (4.6.8)

holds true for all a H (X ), b H (Y ) and H (X Y ).

Proof Let H (X Y ). The uniqueness of () is guaranteed by Lemma 4.6.5,

using Lemma 2.3.3. For its existence, let M be a basis for H (Y ) and let M =
{m H (Y ) | m M} be the dual basis for the Kronecker pairing. We define
  
() = H X H Y (m )  m . (4.6.9)
mM

We must check that (4.6.8) holds true for all a H (X ), b H (Y ). It is enough to

check it for b = n with n M. As n , m = 1 if m = n and 0 otherwise, one has
  
a n , ()
=a n , mM H X H Y (m )  m
= mM a, H X H Y (m )  n , m
= H X (a), H Y (n )  
= H X (a)  H Y (n ), 
= a n ,  .

(Remark: the uniqueness of () shows that the right member of (4.6.9) does not
depend on the choice of the basis M.) 

Theorem 4.6.7 (Knneth Theorem) Let X and Y be topological spaces. Suppose

that Y is of finite cohomology type. Then, the cross product

: (H (X ) H (Y ), +, )
(H (X Y ), +, )

is a GrA-isomorphism and the homology cross product

: H (X Y ) H (X ) H (Y )

is a GrV-isomorphism.

The finiteness condition on one of the space (here Y ) is necessary in the cohomol-
ogy statement, as seen in the beginning of the section. It is used in the proof through
the following lemma.
4.6 The Cross Product and the Knneth Theorem 161

Lemma 4.6.8 Let V be a family of vector spaces over a field F. Let W be a finite
dimensional F-vector space. Then the linear map
   
: V W
V W
V V V V

given by

((v) w) = (v w)

is an isomorphism.

Proof The proof is by induction on n = dim W . The case n = 1 follows from the
canonical isomorphism T F T for any vector space T . The induction step uses
that, in the category of F-vector spaces, tensor and Cartesian products commute with
direct sums. 

Proof (Proof of the Knneth theorem) By Lemma 4.6.3, we know that the cross
product is a GrA-morphism. It is then enough to prove that it is a GrV-isomorphism.
Assuming that Y is of finite cohomology type, the proof goes as follows.
(1) We prove that the cross product is a GrV-isomorphism when X is a finite
dimensional CW-complex.
(2) By Kronecker duality, Point (1) implies that the homology cross product is a
GrV-isomorphism when X is a finite dimensional CW-complex. Any compact
subspace of X Y is contained in X n Y for some n N. Therefore, H (X Y )
is the direct limit of H (X n Y ) by Corollary 3.1.16. Also, H (X ) is the direct
limit of H (X n ). The homology cross product being natural by Lemma 4.6.2
and Kronecker duality, we deduce that is a GrV-isomorphism when X is any
CW-complex. By Kronecker duality, the cross product is a GrV-isomorphism
for any CW-complex X .
(3) If X is any space, there is a map f X : X X , where X is a CW-complex
and f is a weak homotopy equivalence, i.e. the induced map on the homotopy

groups f : ( X , u)
(X, f (u)) is an isomorphism for all u X (see

[82, p. 352] or Remark 3.7.5). As (A B, (a, b)) (A, a) (B, b),
the map f X id : X Y X Y is also a weak homotopy equivalence. But,
weak homotopy equivalences induce isomorphisms on singular (co)homology
(see [82, Proposition 4.21]). The diagram

/ H (X Y )
H (X ) H (Y )
H f X id H ( f X id)
 
/ H ( X Y )
H ( X ) H (Y )
162 4 Products

is commutative by Lemma 4.6.2. This proves that the cross product in singu-
lar cohomology is a GrA-isomorphism for any space X . The corresponding
diagram for the homology cross product, or Kronecker duality, proves that the
homology cross product is a GrV-isomorphism.
It thus remains to prove Point (1). We follow the idea of [82, p. 218]. Let us fix the
topological space Y . To a topological pair (X, A), we associate two graded Z2 -vector
spaces:

h (X, A) = H (X, A) H (Y ) and k (X, A) = H (X Y, A Y ) .

Using Proposition 3.9.2, Point (1) follows from the following lemma. 

Lemma 4.6.9 Let Y be a topological space of finite cohomology type. Then

(a) k and h are two generalized cohomology theories in the sense of Sect. 3.9, with
h ( pt) k ( pt) H (Y ).
(b) The cross product provides a natural transformation from h to k , restricting

to an isomorphism h ( pt) k ( pt).

Proof If f : (X, A) (X  , A ) is a continuous map of pairs, we define h f =

H f id H (Y ) and k f = H ( f idY ). This makes h and k functors from
TOP2 to GrV. The connecting homomorphism h : h (A) h +1 (X, A) and
k : k (A) k +1 (X, A) are defined by

h = id H (Y ) and k = : H (A Y ) H +1 (X Y, A Y ) ,

using the homomorphism of singular cohomology; h and k are then functorial

for continuous maps.
We now check that Axioms (1)(3) of p. 123 hold both for h and k . The homo-
topy and excision axioms are clear. For a topological pair (X, A), the long exact
sequence for k is that in singular cohomology for the pair (X Y, A Y ). The
exact sequence for h is obtained by tensoring with H (Y ) the exact sequence of
(X, A) for H . We use that a direct sum of exact sequences is exact and that, over
a field, tensoring with a vector space preserves exactness. The disjoint union axiom
holds trivially for k . For h , we use that H m (Y ) if of finite dimension for all m and
Lemma 4.6.8. Thus, both h and k are generalized cohomology theories.
We now check Point (b). Let f : (X  , A ) (X, A) be a continuous map of
pairs. We must prove that the following diagram

/ k (X, A)
h (X, A)
h f k f (4.6.10)
 
/ k (X  , A )
h (X  , A )
4.6 The Cross Product and the Knneth Theorem 163

is commutative. This amounts to show that

H f (a) y = H ( f idY )(a y) (4.6.11)

for all a H (X, A) and y H (Y ). This follows from Lemma 4.6.2.

For the second part of Point (b), we must show the commutativity of the diagram

h
h (A) / h +1 (X, A)

(4.6.12)
 k

k (A) / k +1 (X, A) .

This is equivalent to the commutativity of the diagram

id / H p+1 (X, A) H q (Y )
H p (A) H q (Y )
(4.6.13)



H p+q (A Y ) / H p+q+1 (X Y, A Y ) ,

for all p, q N. Here, we have introduced more precise notations, distinguishing the
connecting homomorphisms in singular cohomology : H (A) H +1 (X, A)
and : H (AY ) H +1 (X Y, AY ). We shall also distinguish the homomor-

phisms X : H (X ) H (X Y ) and X : H (A) H (A Y ) induced by the

projections onto A and X , as well as the homomorphisms Y : H (Y ) H (X Y )
and Y : H (Y ) H (A Y ) induced by the projections onto Y . Analogous nota-
tions are used for cochains. The commutativity of Diagram (4.6.13) is thus equivalent
to the formula
 
X (a)  Y (y) =

X (a)  Y (y) for all a H p (A) , y H q (Y ) .
(4.6.14)

Let a Z p (A) and y Z q (Y ) represent a and y. Let a C p (X ) be an extension of a

as a p-cochain of X . By the recipe of Lemma 3.1.17, (a) is a cocycle of C p+1 (X, A)

representing (a). Thus, the left hand member of (4.6.14) is represented by the
cocycle

X (a)
 Y (y)
. (4.6.15)

To compute the right hand member, we need an extension of X (a)  Y (y) as a

is an extension of X (a)
cochain of X Y . But, as cochains of X Y , X (a) and the
is an extension of Y (y).
cocycle Y (y) Therefore, X (a)
 Y (y) is an extension

of X (a)
 Y (y).
By Lemma 3.1.17, the right hand member of (4.6.14) is then
164 4 Products
 
represented by the cocycle X (a)
 Y (y)
. As (Y (y))
= 0, one has
 
X (a) = X (a)
 Y (y) = X (a)
 Y (y)  Y (y)
(4.6.16)

Comparing (4.6.15) and (4.6.16) proves Formula (4.6.14) and then the commutativity
of Diagram (4.6.12). 

Under some hypotheses, there are relative versions of the Knneth theorem, gen-
eralizing Theorem 4.6.7.

Theorem 4.6.10 (Relative Knneth theorem) Let (X, A) be a topological pair. Let
(Y, B) be a good pair such that Y and B are of finite cohomology type. Then, the
cross product

: H (X, A) H (Y, B)
H (X Y, A Y X B) (4.6.17)

is a GrA-isomorphism.

The classical proof of the Knneth theorem (see e.g. [179]) gives the more general
statement that (4.6.17) is an isomorphism if (X B, A Y ) is excisive in X Y and
(Y, B) is of finite cohomology type. If (X, A) and (Y, B) are CW-pairs, the condition
that (Y, B) is of finite cohomology type is also sufficient (see [82, Theorem 3.21];
see also Corollary 4.7.25 below).

Proof As (Y, B) is a good pair, the relative cross product (4.6.6) is defined. Let
Z = A Y X B and let p N. Let us consider the commutative diagram.

H p (X, A) H q1 (Y ) / H p+q1 (X Y, AY )

 

/ H p+q1 (X B, A B) o J
H p (X, A) H q1 (B)
H p+q1 (Z , AY )

 

H p (X, A) H q (Y, B) / H p+q (X Y, Z )

 

H p (X, A) H q (Y ) / H p+q (X Y, AY )

 

/ H p+q (X B, A B) o J
H p (X, A) H q (B)
H p+q (Z , AY )

The left column is the cohomology sequence for (Y, B) tensored by H p (X, A). It is
still exact since we work in the category of Z2 -vector spaces. The right column is
the cohomology sequence for the triple (X Y, Z , A Y ). The homomorphism J ,
induced by inclusion, is an isomorphism: if V is a neighbourhood of B in Y which
4.6 The Cross Product and the Knneth Theorem 165

retracts by deformation onto B, J is the composition

H (AY X B, AY )
H (AV X B, AV )
H (X B, A B) .

The left arrow is an isomorphism by excision of A (Y V ) and the right one

by the homotopy property. As Y and B are of finite cohomology type, the cross
products involving the absolute cohomology H (Y ) or H (B) are isomorphisms, as
established during the proof of Theorem 4.6.7. By the five-lemma, this proves that
the middle cross product is an isomorphism. 

4.7 Some Applications of the Knneth Theorem

4.7.1 Poincar Series and Euler Characteristic of a Product

One application of the Knneth theorem is the multiplicativity of Poincar series and
Euler characteristic.

Proposition 4.7.1 Let X and Y be spaces of finite cohomology type. Then, X Y

is of finite cohomology type and

Pt (X Y ) = Pt (X ) Pt (Y ) . (4.7.1)

If X and Y are finite complexes, then

(X Y ) = (X ) (Y ) . (4.7.2)

Proof Let ai = dim Hi (X ), bi = dim Hi (Y ). The Knneth theorem implies that

dim Hn (X Y ) = i+ j=n ai b j which proves (4.7.1). Equation (4.7.2) follows,
since is the evaluation of Pt at t = 1. Note that (4.7.2) also follows more
elementarily from the cellular decomposition of X Y (see Example 3.4.6). 

4.7.2 Slices

Let y0 Y . The slice inclusion s X : X X Y at y0 is the continuous map defined

by s X (x) = (x, y0 ). The slice inclusion sY : Y X Y at x0 X is defined
accordingly.
Using the bijection Y S0 (Y ), we see y0 Y as a 0-homology class [y0 ]
H0 (Y ). Hence, for b H 0 (Y ) the number b, y0  Z2 is defined.
166 4 Products

Lemma 4.7.2 Let s X : X X Y be the slice inclusion at y0 Y . Let a H m (X )

and b H n (Y ). Then,

b, y0  a if n = 0
H s X (a b) =
0 otherwise.

Proof One has X s X = id X , while Y s X is the constant map c onto y0 . Thus,

H c (b) = 0 if n = 0. When n = 0, H c (b) = b, y0 1. Thus,

H s X (a b) = H s X ( X (a)  Y (b)) = a  H c (b) = b, y0  a . 

Here are two corollaries of Lemma 4.7.2 which enable us to detect cohomology
classes via the slice homomorphisms.
Corollary 4.7.3 Let X and Y be path-connected topological spaces such that
H k (X ) = 0 for k < n. Then, the equation

a = 1 H sY (a) + H s X (a) 1

is satisfied for all a H n (X Y ).

Proof By the hypotheses and the Knneth theorem, the cross product provides an
isomorphism

: H 0 (X ) H n (Y ) H n (X ) H 0 (Y )
H n (X Y ) .

and H 0 (X ) Z2 H 0 (Y ). This implies that a = 1 v + u 1 for some

unique u H n (X ) and v H n (Y ). By Lemma 4.7.2, one has H sY (a) = v and
H s X (a) = u, which proves the corollary. 
The case n = 1 in Corollary 4.7.3 gives the following statement.
Corollary 4.7.4 Let X and Y be path-connected spaces. Then, the equation

a = 1 H sY (a) + H s X (a) 1

is satisfied for all a H 1 (X Y ).

4.7.3 The Cohomology Ring of a Product of Spheres

We first note the associativity of the cross product.

Lemma 4.7.5 Let X , Y and Z be three topological spaces. In H (X Y Z ), the
cross product is associative: (x y) z = x (y z) for all x H (X ), y Y
and z Z .
4.7 Some Applications of the Knneth Theorem 167

Proof We have to consider the various projections 12 : X Y Z X Y ,

23 : X Y Z Y Z , 1 : X Y Z X , etc. Also, 112 : X Y X ,
ij
etc. They satisfy j i j = j . Using the associativity and the functoriality of the
cup product, we get

(x y) z = 12 (112 (x)  212 (y))  3 (z) = 1 (x)  2 (y)  3 (z) .

In the same way, x (y z) = 1 (x)  2 (y)  3 (z). 

The cohomology of the sphere S d being concentrated in dimension 0 and d, one
has a GrA-isomorphism

H (S d ) (x of degree d) ,
Z2 [x]/(x 2 ) (4.7.3)

sending x to the generator [S d ] H d (S d ). Here, Z2 [x]/(x 2 ) denotes the quotient

of the polynomial ring Z2 [x], where x is a formal variable (here of degree d), by
the ideal generated by x 2 . The following proposition then follows directly from the
Knneth theorem.
Proposition 4.7.6 Let X be a topological space. The GrA-homomorphism

H (X )[x] (x 2 )
H (X S d ) (x of degree d) ,

induced by a  a 1, for a H (X ), and x  1 [S d ] , is a GrA-isomorphism.

Using Proposition 4.7.6 together with Lemma 4.7.5, we get the following propo-
sition.
Proposition 4.7.7 For i = 1 . . . , m, let xi be a formal variable of degree di . Then,
the GrA-homomorphism

H (S d1 S dm )
Z2 [x1 , . . . , xm ] (x12 , . . . , xm2 )

induced by

xi  1 1 [S di ] 1 1

is a GrA-isomorphism.

4.7.4 Smash Products and Joins

Let (X, x) and (Y, y) be two pointed spaces. The base points provide an inclusion

X Y X {y} {x} Y  X Y .
168 4 Products

The smash product X Y of X and Y is the quotient space

X Y = X Y X Y .

It is pointed by x y, the image of X Y in X Y .

Recall that (X, x) is well pointedif the pair (X, {x}) is well cofibrant. The following
lemma is the first place where the full strength of this definition is used.
Lemma 4.7.8 If (X, x) and (Y, y) are well pointed, so is (X Y, x y).
Proof By Lemma 3.1.40, the pair (X Y, X Y ) is well cofibrant and, by
Lemma 3.1.43, so is (X Y, {x y}). 
By Proposition 3.1.45 and (3.1.13), one has the isomorphisms

H (X Y, X Y ) H (X Y, {x y}) H (X Y ) . (4.7.4)

Proposition 4.7.9 Let (X, x) and (Y, y) be well pointed spaces. Then, the homo-
morphisms induced by the inclusion i : X Y X Y and the projection
p : X Y X Y give rise to the short exact sequence

H p
0 H (X Y ) H (X Y ) H (X Y ) 0 .
H i

Proof Using the isomorphism (4.7.4) and the exact sequence of Corollary 3.1.48, it
is enough to prove that H i is onto. Consider the commutative diagram

H (X )
id / H (X )
LLL H r 9
LLL X X rr
H j
LL r r
& rrr
H i / H (X Y )
H (X Y )
8 LLL
r r rrr LLL
L
rrr
H Y H jY L%
H (Y )
id / H (Y )

where X , Y are the projections and j X , jY the inclusions. We note that Y j X and
X jY are constant maps. By Proposition 3.1.51, the homomorphism H (X Y )
( H j , H j )
H (X ) H (Y ) is an isomorphism. Hence H i is onto.
X

Y

Remark 4.7.10 Using the relationship between the exact sequence of the pair
(X Y, X Y ) and that of Corollary 3.1.48, Proposition 4.7.9 implies that the homo-
morphism H i : H (X Y ) H (X Y ) is surjective, whence the short exact
sequence

H (X Y ) H (X Y ) 0 .
H i
0 H (X Y, X Y ) (4.7.5)
4.7 Some Applications of the Knneth Theorem 169

As (X, {x}) and (Y, {y}) are good pairs, the relative cross product

H (X, {x}) H (Y, {y})
H (X Y, X Y )

is defined by (4.6.6). Using the isomorphisms of (3.1.13), one constructs the com-
mutative diagram

/ H (X ) H (Y )
H (X, {x}) H (Y, {y})

(4.7.6)
 
/ H (X Y )
H (X Y, X Y )

The relative Knneth theorem 4.6.10

which defines the reduced cross product .
gives the following reduced Knneth theorem.
Proposition 4.7.11 Let (X, x) and (Y, y) be well pointed spaces, with Y of finite
cohomology type. Then, the reduced cross product

: H (X ) H (Y )
H (X Y )

is a multiplicative GrV-isomorphism.
For a pointed space (Z , z), Diagram (3.1.14) provides an injective homomorphism
H (Z ) H (Z ). Using this together with Proposition 4.7.9 (or Remark 4.7.10),
the Knneth theorem and its reduced form are summed up by the diagram

H (X ) H (Y ) / / H (X ) H (Y )

. (4.7.7)
 
H (X Y ) /
/ H (X Y )

Example 4.7.12 Proposition 4.7.11 says that H k (S p S q ) = 0 for k = p + q and

H p+q (S p S q ) = Z2 . Actually, S p S q is homeomorphic to S p+q by the following
homeomorphisms. Let Dr be the compact unit disk of dimension r with boundary
Dr = Sr 1 . Then Dr / Dr is homeomorphic to Sr and

q
 
S S
p
D / D D / D
p p
[ D p ] D q D p [ D q ]
q q




D D
p q
D p Dq D p Dq

D p D q (D p D q ) S p+q .

170 4 Products

Let (X, x) be a well pointed space. The smash product X S 1 is called the
reduced suspension of X , which has the same homotopy type than the suspension
 X . Indeed, let I = {0, 1}. The map

F :  X = (X I ) (X I ) X S 1

given by F(x, t) = [(x, e2it )] descends to a homeomorphism

F : 
X =  X/({x} I )
X S1 .

This homeomorphism preserves the base points, if we choose those to be [x]  X

and 1 S 1 . The pair (I, I ) is well cofibrant by Lemma 3.1.39. By Lemma 3.1.40, so
is the product pair (X I, X I {x} I ). By Lemma 3.1.43, the pair ( X, {x} I )
is well cofibrant. As {x} I is contractible, the projection F :  X  X X S1
is a homotopy equivalence by Lemma 3.1.44.
Let b be the generator of H 1 (S 1 ) = Z2 . By Propositions 4.7.11 and 3.1.49 and
the above, the three arrows in the following proposition are isomorphisms.

Lemma 4.7.13 The diagram

H n (X )
MMM
pppp
b MMM
ppp MM&
wpp
HF / H n+1 ( X )
H n+1 (X S 1 )

is commutative.

Proof As all these isomorphisms are functorial, it is enough to prove the lemma
for X = Kn . This is possible since the cellular decomposition of Kn given in
Proposition 3.8.1 has 0-skeleton Kn0 = {x0 }, so (Kn , x0 ) is a well pointed space
by Proposition 3.4.1. In this particular case, the statement is obvious since the three
groups are isomorphic to Z2 . 

The smash product gives a geometric interpretation of the cup product. Let a
H m (X ) and b H m (X ), given by maps f a : X Km and f m : X Kn to
Eilenberg-MacLane spaces. By Proposition 4.7.11, H m+n (Km Kn ) = Z2 , with
generator corresponding to g : Km Kn Km+n .

Proposition 4.7.14 The composed map

( fa , fb ) g
X Km Kn Km Kn
Km+n

represents the class a  b H m+n (X ).

4.7 Some Applications of the Knneth Theorem 171

Proof By Proposition 4.7.11, the generator of H m+n (Km Kn ) = Z2 is the reduced

n . By Diagram (4.7.7), it is send to m n in H m+n (Km Kn ).
cross product m
Now, the composed map of Proposition 4.7.14 coincides with the composition

 fa fb g
X
X X Km Kn Km Kn
Km+n .

Proposition 4.7.14 then follows from Remark 4.6.1. 

( fa , fb )
If we consider the composed map f : X Km Kn Km Kn ,
Proposition 4.7.14 gives the following corollary.

Corollary 4.7.15 The diagram

H m (Km ) H n (Kn ) / H m+n (Km Kn )

f a f b f
 
 / H m+n (X )
H m (X ) H n (X )

is commutative.

Let X and Y be two topological spaces. Their join X Y is the quotient of

X Y I by the equivalence relation (x, y, 0) (x, y  , 0) for y, y  Y and
(x, y, 1) (x  , y, 1) for x, x  X . This topological join is related to the simplicial
join in the following way: if K and L are locally finite simplicial complexes, then
|K L| is homeomorphic to |K | |L| (see [155, Lemma 62.2]).
The two open subspaces X Y [0, 1) and X Y (0, 1] of X Y I define
open subspaces U X and UY of X Y . The space U X retracts by deformation onto X
and UY retracts by deformation onto Y . Moreover, U X UY retracts by deformation
onto X Y { 21 }. The following diagram is homotopy commutative,

U X UY
incl / UX
O O
 
X
X Y / X

as well as the corresponding diagram for Y . Consider the homomorphism

X +Y
H k (X ) H k (Y ) H k (X Y ) .

If k > 0, then ( X + Y )(a, b) = a 1 + 1 b and, by the Knneth theorem, X + Y

is injective. As X Y is path-connected, the Mayer-Vietoris sequence for the data
172 4 Products

(X Y, U X , UY , U X UY ) splits and gives, for all integers k 0, the exact sequence

+
0 H k (X ) H k (Y ) H k (X Y )
H k+1 (X Y ) 0 .
X Y
(4.7.8)

Example 4.7.16 The join S p S q is homeomorphic to S p+q+1 . Considering

S p+q+1 R p+1 Rq+1 , a homeomorphism S p+q+1 S p S q is given by
(x, y)  [(x, y, |x|]. The reader can check (4.7.8) on this example, including the
case p = q = 0.
Observe that U X and UY are contractible in X Y . Hence, the Lusternik-
Schnirelmann category of X Y is equal to 2. By Proposition 4.4.2, the cup product
in H >0 (X Y ) vanishes.
When the Knneth theorem is valid, one sees that the cohomology ring of X Y is
isomorphic to that of (X Y ). Actually, under some hypotheses, these two spaces
have the same homotopy type (see [82, Ex. 24,p. 20]).

4.7.5 The Theorem of Leray-Hirsch

An important generalization of a product space is a locally trivial fiber bundle. A
map p : E B is a locally trivial fiber bundle with fiber F (in short: a bundle)
if there exists an open covering U of B and, for each U U, a homeomorphism

U : U F p 1 (U ) such that p (x, v) = x for all (x, v) U F. The space
E is the total space and B is the base space of the bundle. If A is a subspace of B, we
set E A = p 1 (A), getting a bundle p : E A A. If b B, we set E b = E {b} and
denote by i b : E b  E the inclusion. A fiber inclusion is an embedding i : F E
which is a homeomorphism onto some fiber E b . As elsewhere in the literature, we
i p
shall often speak about a (locally trivial) bundle F
E
B, meaning a locally
trivial bundle p : E B with fiber F together with a chosen fiber inclusion i.
If p : E B is a bundle, then the homomorphism p = H p : H (B)
H (E) provides a structure of graded H (B)-module on H (E).
A cohomology extension of the fiber is a GrV-morphism : H (F) H (E)
such that, for each b B, the composite map
H ib
H (F)
H (E) H (E b )

is a GrV-isomorphism. We do not require that is multiplicative. In the presentation

i p
of a bundle by a sequence F E B with B path-connected, a cohomology
extension of the fiber exists if and only if H i is surjective.
A cohomology extension of the fiber provides a morphism of graded H (B)-
modules

H (B) H (F)
H (E)
b) = p (a)  (b).
given by (a
4.7 Some Applications of the Knneth Theorem 173

Suppose that F is of finite cohomology type. As in Lemma 4.6.6, there is a unique

GrV-homomorphism
: H (E) H (B) H (F) .

such that the formula

u),  = b u, ()
(b (4.7.9)

holds true for all b H (B), u H (F) and H (E). As in the proof of
Lemma 4.6.6, we show that
  
() = H p (m )  m
mM

where M be a basis for H (F) and M = {m H (F) | m M} is the dual

basis for the Kronecker pairing.
p
Theorem 4.7.17 (Leray-Hirsch) Let E B be a locally trivial fiber bundle with
fiber F. Suppose that F is of finite cohomology type. Let : H (F) H (E) be
a cohomology extension of the fiber. Then, is an isomorphism of graded H (B)-
modules and is a GrV-isomorphism.

Proof By Kronecker duality, only the cohomology statement must be proven.

Let A B and let h (A) = H (A) H (F). The composition

A : H (F)
H (E)
H (E A )

is a cohomology extension of the fiber for the bundle p : E A A, giving rise to

A : h (A) H (E A ). We want to prove that B is an isomorphism. Considering
the commutative diagram

B
h (B) / H (E)


 
A0 (B) H (A) H (F)

  
 A 

A0 (B) h (A)
/ A0 (B) H (E A )

where is the linear map of Lemma 4.6.8, which is an isomorphism since H k (F)
is finite dimensional for all k, permits us to reduce to the case where the base is
path-connected.
174 4 Products

We now suppose that B is path-connected and that the bundle E B is trivial,

i.e. there exists a homeomorphism : B F E such that p = B , the
projection to the factor B. Since F is of finite cohomology type, one may use H
together with the Knneth formula to identify H (E) with H (B) H (F).
Fix an integer n and consider the vector subspace Pkn of H (B) H (F)
defined by

 
Pkn = H p (B) H q (F) | p + q = n and q k .
p,q

These subspaces provide a filtration

0 = P1
n
P0n Pnn = H n (E) . (4.7.10)

We denote by : H (F) H (F) the GrV-morphism defined by = H i .

The fiber inclusion i is a slice over one point. Let u H q (F). As B is path-connected,
Lemma 4.7.2 implies that
q
(u) = 1 (u) mod Pq1 . (4.7.11)

Hence, if a H p (B) with p + q = n, one has

u) = (a 1)  (u)
(a
= (a 1)  (1 (u)) n
mod Pn1 (4.7.12)
= a (u) n
mod Pn1

In particular, preserves the filtration (4.7.10). It thus induces homomorphisms

: Pkn /Pk1
n Pkn /Pk1
n . Moreover, one has a natural identification P n /P n
k k1
H nk k
(B) H (F) under which (a u) = a (u). This enables us to prove by
induction on k that : Pkn Pkn is an isomorphism, using the five-lemma in the
diagram

0 / Pn / Pn / H nk (B) H k (F) / 0
k1 k

id
  
0 / Pn / Pn / H nk (B) H k (F) / 0
k1 k

Indeed, the left vertical arrow is an isomorphism by induction hypothesis. Since

n = 0, the induction starts with k = 0, using (4.7.12). Therefore, the Leray-
P1
Hirsch theorem is true for a trivial bundle.
Let Bi , i = 1, 2, be two open sets of B with B = B1 B2 . Let B0 = B1 B2
and E i = p 1 (Bi ). The Mayer-Vietoris cohomology sequence for (B, B1 , B2 , B0 )
4.7 Some Applications of the Knneth Theorem 175

may be tensored with H k (F) and remains exact, since we are dealing with Z2 -vector
spaces. The sum of these sequences provides the exact sequence of the top line of
the commutative diagram

h k1 (B1 ) h k1 (B2 ) / h k1 (B ) / h k (B) /

0

1 2 0
  
H k1 (E 1 ) H k1 (E 2 ) / H k1 (E ) / H k (E) /
0

/ h k (B ) h k (B ) / h k (B )
1 2 0

1 2 0
 
/ H k (E ) H k (E ) / h k (E )
1 2 0

The bottom line is the Mayer-Vietoris sequence for the data (B, B1 , B2 , B0 ). By
the five-lemma, this shows that, if i are isomorphisms for i = 0, 1, 2, then is an
isomorphism.
What has been done so far implies that the is an isomorphism for a bundle of
p
finite type, i.e. admitting a finite covering U such that EU
U is trivial for U U.
By Kronecker duality, is an isomorphism in this case. As in Point (2) of the proof
of the Knneth theorem (p. 161), is the direct limit of A for A B such that
E A A is of finite type. Therefore, is an isomorphism and, by Kronecker duality,
is an isomorphism for any bundle. 

The Leray-Hirsch theorem also has the following version, in which the finite type
hypothesis is on the base rather than on the fiber. The proof, involving the Serre
spectral sequence, may be found in [141, Theorem 10].
p
Theorem 4.7.18 (Leray-Hirsch II) Let E B be a locally trivial fiber bundle
with fiber F. Suppose that B is path-connected and of finite cohomology type. Let :
H (F) H (E) be a cohomology extension of the fiber. Then, is an isomorphism
of graded H (B)-modules and is a GrV-isomorphism.

Here are a few corollaries of the above Leray-Hirsch theorems.

i p
Corollary 4.7.19 Let F E B be a locally trivial fiber bundle whose base
B is path-connected and whose fiber F (or base B) is of finite cohomology type.
Suppose that H i : H (E) H (F) is surjective. Then
(1) H p : H (B) H (E) is injective.
(2) ker H i is the ideal generated by the elements of positive degree in the image
of H p.
176 4 Products

Proof Let : H k (F) H k (E) be a cohomology extension of the fiber such that
1), the homomorphism H p is
H i is the identity of H (F). As H p(b) = (b
injective.
To prove (2), we note that any element in A H (E) may be written uniquely

as a finite sum A = kK (b k ak ). Let Ko = {k K | bk = 1}. As p i is
a constant map, one has H ( p i)(1) = 1 and H ( p i)(bk ) = 0 if bk has positive

degree. Therefore,

H i(A) = kK H i(H p(bb k)  (ak )) 
= kK  H i H p(bk )  H i((ak ))
= kK H ( p i)(bk )  ak
= kK0 ak .

Hence, H i(A) = 0 if and only if Ko = , which proves (2). 

As for Proposition 4.7.1, the Leray-Hirsch theorem implies the following result.
i p
Corollary 4.7.20 Let F E B be a locally trivial fiber bundle whose whose
base B is path-connected. Suppose that H i : H (E) H (F) is surjective.
Suppose that F and B are of finite cohomology type. Then, E is of finite cohomology
type and the Poincar series of F, E and B satisfy

Pt (E) = Pt (F) Pt (B) . (4.7.13)

Actually, Eq. (4.7.13) is equivalent to H i being surjective (see [15,

Proposition 2.1]). Here is another kind of corollaries of the Leray-Hirsch theorem.

Corollary 4.7.21 Let p : E B be a locally trivial fiber bundle with fiber F.

Suppose that H (F) = 0. Then, H p : H (B) H (E) is a GrA-isomorphism.

Remarks 4.7.22 (1) By Corollary 4.7.19, the existence of a cohomology extension

of the fiber implies that p : H (B) H (E) is injective. The converse is not true,
even if the map p has a section (see e.g. [72]).
(2) In the Leray-Hirsch theorem the isomorphism is not a morphism of algebras,
unless is multiplicative. It is possible that there exists cohomology extensions of
the fiber but that none of them is multiplicative (see Examples 4.7.45 or 7.1.16).
(3) The proof of the Leray-Hirsch theorem shows the following partial result. Let
: H k (F) H k (E) be a linear map defined for all k n. Suppose that, for
each b B, the composition H i b : H k (F) H k (E b ) is an isomorphism
for k n. Then, with the notation of the proof of Theorem 4.7.17, the linear map
: h k (B) H k (E) is an isomorphism for k n. For instance, we get the following
proposition.

Proposition 4.7.23 Let p : E B be a locally trivial fiber bundle with fiber F.

Suppose that H k (F) = 0 for all k m. Then, H p : H k (B) H k (E) is an
isomorphism for k m.
4.7 Some Applications of the Knneth Theorem 177

The Leray-Hirsch theorem admits a version for bundle pairs. A bundle pair with
fiber (F, F  ) is a topological pair (E, E  ) and a map p : (E, E  ) (B, B) such that
there exists an open covering U of B and, for each U U, a homeomorphism U :

U (F, F  ) ( p 1 (U ), p 1 (U ) E  ) such that p (x, v) = x for all (x, v)
U F. In consequence, p : E B is a bundle with fiber F and the restriction of p to
E  is a bundle with fiber F  . A cohomology extension of the fiber is a GrV-morphism
r el : H (F, F  ) H (E, E  ) such that, for each b B, the composite

r el H ib
H (F, F  ) H (E, E  ) H (E b , E b )

is a GrV-isomorphism. A cohomology extension of the fiber provides a morphism

of graded H (B)-modules

r el
H (B) H (F, F  ) H (E, E  )

given by r el (a b) = p (a)  r el (b).

Suppose that F is of finite cohomology type. As in Lemma 4.6.6, there is a unique
GrV-homomorphism

r el : H (E, E  ) H (B) H (F, F  ) .

such that the formula

r el (b u),  = b u, r el () (4.7.14)

holds true for all b H (B), u H (F, F  ) and H (E, E  ). The formula
  
r el () = H p r el (m )  m . (4.7.15)
mM

is satisfied for all H (E, E  ), where M is a basis for H (F, F  ) and M =

{m H (F, F  ) | m M} is the dual basis for the Kronecker pairing.

Theorem 4.7.24 (Leray-Hirsch relative) Let p : (E, E  ) (B, B) be a bundle

pair with fiber (F, F  ). Suppose that (F, F  ) is a well cofibrant pair and is of finite
cohomology type. Let r el : H (F, F  ) H (E, E  ) be a cohomology extension of
the fiber. Suppose that (E, E  ) is a well cofibrant pair. Then, r el is an isomorphism
of graded H (B)-modules and r el is a GrV-isomorphism.

The hypothesis that (E, E  ) is well cofibrant may be removed but this would
necessitate some preliminary work. Besides, this hypothesis is easily fulfilled in our
applications.
178 4 Products

Proof By Kronecker duality, only the cohomology statement must be proven. We

first reduce to the case where F  is a point. Let E = E/ where is the equivalence
relation
x y p(x) = p(y) and x, y E 

and let E  = E  /. Then the map p descends to a map p : ( E,

E  ) (B, B) which
is a bundle pair with fiber (F/F , y0 ), where y0 is the point given by F  in F/F  . In


particular, p : E  B is a homeomorphism. Consider the commutative diagram:

/ H (B) H (F, F  )
H (B) H (F/F  , y 0 )

r el r el (4.7.16)
 
/ H (E, E  )
E  )
H ( E,

We shall show that the horizontal homomorphisms, induced by the quotient maps,
are isomorphisms. Therefore, the right vertical arrow is bijective if and only if the
left one is.
The top horizontal homomorphism of Diagram (4.7.16) is an isomorphism by
Proposition 3.1.45 since (F, F  ) is a well cofibrant pair. To see that the bottom hori-
zontal map is also bijective, consider the commutative diagram

/ H (E/E  , [E  ])
E  , [ E  ])
H ( E/

 
H ( E, E  )
/ H (E, E  )

The top horizontal map is an isomorphism because the quotient spaces E/E  and
E  are equal. As (E, E  ) is well cofibrant, the right hand vertical map is bijective
E/
by Proposition 3.1.45. Also, Lemma 3.1.43 implies that ( E, E  ) is well cofibrant and
thus, the left hand vertical map is an isomorphism by Proposition 3.1.45. Now, the
diagram
r el
H (F/F  , [F  ]) / H ( E,
E  ) / H ( E b , E  )
b

  
r el
H (F, F  ) / H (E, E  ) / H (E b , E  )
3 b

shows that the bundle pair p inherits a cohomology extension of the fiber r el .
4.7 Some Applications of the Knneth Theorem 179

We are then reduced to the case F  = {y 0 } being a single point. Consider the
commutative diagram:

H (B) H (F, y 0 ) / / H (B) H (F) / / H (B) H (y 0 )

r el
  
H (E, E  ) / / H (E) / / H (E  )

The top line is the exact sequence of the pair (F, y 0 ) tensored by H (B). It is exact
since we are dealing with Z2 -vector spaces and its splits since {y 0 } is a retract of F  .
The bottom exact sequence of the pair (E, E  ) also splits since p : E B E 
is a retraction of E onto E  . We shall check below the existence of a cohomology
extension of the fiber : H (F) H (E), whence the middle vertical map .

The two maps are bijective by the absolute Leray-Hirsch theorem 4.7.17. By the
five-lemma, r el is then also an isomorphism.
The existence of a cohomology extension of the fiber : H (F) H (E) comes
from r el : H (F, y 0 ) H (E, E  ) when > 0, since H k (F, y 0 ) H k (F) and
H k (E b , E b ) H k (E b ) for k > 0. When k = 0, we consider the diagram

,
H 0 (F, y 0 ) / / H 0 (E, E  ) / / H 0 (E b , E  )
b
 r el
H j
  
/ H 0 (E) / / H 0 (E b )
H (F)
0 ,
X X
r p
  
 / H 0 (E  ) / H 0 (E  )
H 0 (y 0 ) 2 b

where j : (E, ) (E, E  ) denotes the inclusion. The retraction r : F {y 0 }

produces a section r : H 0 (y 0 ) H 0 (F) of the homomorphism induced by the
inclusion; this section provides an isomorphism H 0 (F) H 0 (F, y 0 ) H 0 (y 0 ).
As p : E  B is a homeomorphism, one gets a section p : H 0 (E  ) H 0 (E)
of the homomorphism induced by the inclusion. The homomorphism : H 0 (F)
H 0 (F, y 0 ) H 0 (y 0 ) H 0 (E) given by (a, b) = H j r el (a) + p  (b) com-
pletes the definition of the cohomology extension of the fiber in degree 0. 
Corollary 4.7.25 (Relative Knneth theorem II) Let X be a topological space and
(Y, C) be a well cofibrant pair which is of finite cohomology type. Then, the cross
product

: H (X ) H (Y, C)
H (X Y, X C)

is a GrV-isomorphism.
180 4 Products

Proof We see the projection 1 : (X Y, X C) (X, X ) as a trivial bundle

pair with fiber (Y, C). Then = H 2 : H (Y, C) H (X Y, X C) is a
cohomology extension of the fiber. As (Y, C) is a well cofibrant pair, so is (X
Y, X C). The cohomological result then follows from the relative Leray-Hirsch
theorem 4.7.24. 

4.7.6 The Thom Isomorphism

We start by some preliminary results.

Lemma 4.7.26 Let p : ( E, E) (B, B) be a bundle pair whose fiber ( F, F) is a

F) = 0 for k < r and that Hr ( F,
well cofibrant pair. Suppose that Hk ( F, F) = Z2 .

E) = 0 for k < r and there is a unique isomorphism : Hr ( E,
Then, Hk ( E, E)

H0 (B) such that, for each b B, the diagram

Hr ( E b , E b ) / / H ( E,
r E)

= (4.7.17)
 
H0 ({b}) / / H0 (B)

is commutative, where the horizontal homomorphisms are induced by the inclusions.

In Diagram (4.7.17), the left vertical isomorphism is abstract but well defined,
since both Hr ( E b , E b ) Hr ( F,
F) and H0 ({b}) are equal to Z2 .

Proof We first prove the uniqueness of , if it exists. Indeed, for each path-
connected component A of B, Diagram (4.7.17) implies that (H r ( E A , E A )) =
H0 (A) and, as H0 (A) = Z2 , the isomorphism is unique.
If the bundle pair is trivial, the lemma follows from the relative Knneth
theorem 4.7.25. Suppose that B = B1 B2 , where B1 and B2 are two open sets
with B1 B2 = B0 . Suppose that the conclusion of the lemma is satisfied for
( E i , E i ) for i = 0, 1, 2. Then, the Mayer-Vietoris sequence for the data ( E i , E i )
implies that Hk ( E, E) = 0 for k < r and gives the diagram

Hr ( E 0 , E 0 ) / H ( E , E ) H ( E , E ) / H ( E, / 0
r 1 1 r 2 2 r E)

.
  
H0 (B0 ) / H0 (B1 ) H0 (B2 ) / H0 (B) / 0

Diagram (4.7.17) for each b B0 implies that the left square is commutative.
Therefore, the middle vertical isomorphism descends to a unique homomorphism
4.7 Some Applications of the Knneth Theorem 181

E) H0 (B) making the right square commutative, which an isomor-

: Hr ( E,
phism by the five-lemma. It remains to prove the commutativity of Diagram (4.7.17)
for . Let b B. Without loss of generality, we may suppose that b B1 . Consider
the diagram

Hr ( E b , E b ) / H ( E , E ) / H ( E,
r 1 1 r E)

=
.
  
H0 ({b}) / H0 (B1 ) / H0 (B)

As both square commute, this gives the commutativity of Diagram (4.7.17) for .
We have so far proven the lemma when the bundle pair ( E, E) (B, B) is of
finite type. Let A be the set of subspaces A of B such that the bundle pair ( E A , E A )
(A, A) is of finite type. Each compact of B is contained in some A A and each
compact of E is contained in E A for some A A. By Proposition 3.1.29, this
provides isomorphisms

lim

Hr ( E A , E A ) Hr ( E,
E) and lim H0 (A) H0 (B) .

(4.7.18)
AA AA

Now, if A, A A with A A , Diagram (4.7.17) for each b A implies that the

diagram

Hr ( E A , E A ) / H ( E  , E  )
r A A

(4.7.19)
 
H0 (A) / H0 (A )

is commutative. We therefore get isomorphisms

lim

Hr ( E A , E A )
lim

H0 (A)
AA AA

which, together with the isomorphisms of (4.7.18), produce the required isomor-

E)
phism : Hr ( E, H0 (B). 

By Kronecker duality, Lemma 4.7.26 gives the following lemma.

Lemma 4.7.27 Let p : ( E, E) (B, B) be a bundle pair whose fiber ( F, F) is a

k
well cofibrant pair. Suppose that H ( F, F) = 0 for k < r and that H ( F, F) = Z2 .
r

E) = 0 for k < r and there is a unique isomorphism : H 0 (B)
Then, H k ( E,
E) such that, for each b B, the diagram
H ( E,
r
182 4 Products

H 0 (B) / / H 0 ({b})

= (4.7.20)
 
E)
H r ( E, / / H r ( E , E )
b b

is commutative, where the horizontal homomorphisms are induced by the inclusions.

E) (B, B) a bundle pair satisfying the hypotheses of Lemma 4.7.27.
Let p : ( E,
E) is called the Thom class of the bundle pair p. If B
The class U = (1) H r ( E,
E) = Z2 ,
is path-connected, the Thom class is just the non-zero element of H r ( E,
whence the following characterization of the Thom class.
E) which restricts
Lemma 4.7.28 The Thom class of p is the unique class in H r ( E,
to the generator of H r ( E b , E b ) for all b B.

Let  be a topological space having the same homology (mod 2) as the sphere
S k1 . For example,  = S k1 or a lens space with odd fundamental group. Let
p : E B be a bundle with fiber . Let E be the mapping cylinder of p:

E = (E I )
B {(x, 1) p(x)} .

E) (B, B)
Let C be the cone over . Then p extends to a bundle pair p : ( E,
with fiber (C, ), called the mapping cylinder bundle pair of p. As (C, )
is a well cofibrant pair (by Lemma 3.1.39) and H k (C, ) = 0 for k = r and
E) is defined.
H r (C, ) = Z2 , the Thom class U H r ( E,

Theorem 4.7.29 (The Thom isomorphism theorem) Let p : E B be a bundle

E) (B, B)
with fiber , where  has the homology of the sphere Sr 1 . Let p : ( E,

be its mapping cylinder bundle pair. Let U H ( E, E) be the Thom class. Then,
r

the homomorphisms

E) and : Hk ( E,
: H k (B) H k+r ( E, E) Hkr (B)

given by
(a) = H p(a)  U and () = H p(U  )

are isomorphism for all k Z.

Observe that Lemma 4.7.27 gives the result for k 0.

Proof As H j (C, ) = 0 for j = r , the homomorphism r el : H (C, )

E) sending the generator of H r (C, ) = Z2 onto the Thom class U is
H ( E,
E) and the fiber (C, ) are well
a cohomology extension of the fiber. Also, ( E,
cofibrant by Lemma 3.1.39. The relative Leray-Hirsch theorem 4.7.24 then provides
4.7 Some Applications of the Knneth Theorem 183

a GrV-isomorphism r el : H (B) H (C, ) E). Let be the
H ( E,
composite isomorphism

r el
E)
: H k (B) H k (B) H r (C, ) H k+r ( E,

satisfy, by definition of r el , the formula (a) = H p(a)  U . This proves the

cohomology statement.
For the isomorphism , let 0 = m Hr (C, ). Then {m} and {U } are
Kronecker dual bases for (co)homology of (C, ) in degree r . Theorem 4.7.24
and Formula (4.7.15) implies that the composite isomorphism

r el
E) Hk (B) Hr (C, ) Hk (B)
: Hk+r ( E,

satisfies () = H p(U  ). 
Let q : E B be a bundle with fiber F and let f : A B be a continuous
map. The induced bundle f q : f E A is defined by

f E = {(a, y) A E | f (a) = q(y)} , f q(a, y) = a ,

where f E is topologized as a subspace of A E. Then f q is a bundle over A

with fiber F. The projection onto E gives a map f : f E E and a commutative
diagram

f
f E / E

f q q .
 f

A / B

Let p : E B be a bundle with fiber , where  has the homology of the sphere
E) (B, B) be its mapping cylinder bundle pair. Let f : A
Sr 1 and let p : ( E,
f E) (A; A) is the mapping cylinder bundle pair of the
B be a map. Then ( f E,

induced bundle f E. The following lemma states the functoriality of the Thom class.
Lemma 4.7.30 If U H r ( E, E) is the Thom class of p, then H f(U )
f E) is the Thom class of f p.
H r ( f E,
Proof For a A, let consider the commutative diagram

H f
E)
H r ( E, / H r ( f E,
f E)

.
 
H f
H r ( E f (a) , E f (a) ) / H r (( f E)
a , ( f E)a )

184 4 Products

Both cohomology groups downstairs are equal to Z2 . The left vertical arrow sends
the Thom class U to the non-zero element. Therefore, H f(U ) goes, by the right
vertical arrow, to the non-zero element. As this is true for all a A, we deduce from
Lemma 4.7.28 that H f(U ) is the Thom class of f p. 
Let p : E B be a bundle with fiber , where  has the homology of the
E) (B, B) be its mapping cylinder bundle pair. As
sphere Sr 1 . Let p : ( E,
is a
C is contractible, Proposition 4.7.23 implies that H p : H (B) H ( E)
GrA-isomorphism. Therefore, there is a unique class e H (B) such that
r

H j (U ) = H p(e) , (4.7.21)

where U H r ( E, E) is the Thom class and j : ( E, ) ( E, E) is the pair

inclusion. The class e = e( p) is called the Euler class of the bundle p. If :
E) is the Thom isomorphism, one has the formula
H r (B) H 2r ( E,

(e) = U  U . (4.7.22)

Indeed:

(e) = H p(e)  U = H j (U )  U = U  U ,

the last equality coming from Lemma 4.1.7. The Euler class is functorial by the
following lemma.
Lemma 4.7.31 Let p : E B be a bundle with fiber , where  has the homology
of the sphere Sr 1 . Let f : A B be a map. If e H r (B) is the Euler class of p,
then H f (e) H r (A) is the Euler class of f p.
Proof This follows from the definition of the Euler class, Lemma 4.7.30 and the
commutativity of the diagram.

H p
E)
H r ( E, / H ( E)
o H r (B)

H f H f H f .
  
H f p
f E)
H r ( f E, / H r ( f E)
o H r (A)

The Euler class occurs in the Gysin exact sequence. 

Proposition 4.7.32 (Gysin exact sequence) Let p : E B be a bundle with fiber
, where  has the homology of the sphere Sr 1 . Let e H r (B) be its Euler class.
Then, there is a long exact sequence
H p e H p
H k1 (B) H k1 (E) H kr (B) H k (B) H k (E)

which is functorial with respect to induced bundles.

4.7 Some Applications of the Knneth Theorem 185

E) (B, B) be the mapping cylinder pair of p. One uses the

Proof Let p : ( E,
E) and the commutative diagram
cohomology sequence of the pair ( E,

H j
/ H k1 (E) / H k ( E,
E) / H k ( E)
/
O O
H p
e / H k (B)
H kr (B)

E) ( E,
where j : ( E, ) denotes the inclusion and is the Thom isomorphism.
The diagram is commutative since, for a H kr (B),
 
H j (a) = H j H p(a)  U
= H p(a)  H j (U )
= H p(a)  H p(e)
= H p(a  e) .

(The second equality is the singular analogue of Lemma 4.1.8). The functoriality of
the Gysin exact sequence comes from Lemma 4.7.30 and 4.7.31. 

Corollary 4.7.33 Let p : E B be a bundle with fiber , where  has the

homology of the sphere Sr 1 . If p admits a continuous section, then the Euler class
of p vanishes.

Proof In the following segment of the Gysin sequence:

e H p
H 0 (B) H r (B) H r (E) ,

the class 1 H 0 (B) is sent to the Euler class e. If p admits a section, then H p is
injective, which implies that e = 0. 

Remark 4.7.34 The vanishing of the Euler class of p : E B does not imply that
p admits a section. As an example, let p : S O(3) S 2 the map sending a matrix to
its first column vector. Then p is an S 1 -bundle, equivalent to the unit tangent bundle
of S 2 . The Gysin sequence gives the exact sequence

e H p
H 0 (S 2 ) H 2 (S 2 ) H 2 (S O(3)) 0

As SO(3) is homeomorphic to RP 3 , H 2 (S O(3)) = Z2 by Proposition 4.3.10 and all

the cohomology groups in the above sequence are equal to Z2 . Hence, e = 0. But it
is classical that S 2 admits no nowhere zero vector field [82, Theorem 2.28].

Proposition 4.7.35 Let p : E B be a bundle with fiber , where  has the

homology of the sphere Sr 1 . Let e H r (B) be its Euler class. Then, the following
assertions are equivalent.
186 4 Products

(1) e = 0.
(2) The restriction homomorphism H (E) H () is surjective.
(3) H r 1 (E) H r 1 (B) Z2 .

Proof Let p : ( E, E) (B, B) be the mapping cylinder pair of p. Identifying 

as the fiber over some point of B, we get a commutative diagram

0 / H r 1 ( E)
/ H r 1 (E) / H r ( E,
E) / H r ( E)

O
H p
 
/ H r (C, )
H r 1 (B) H r 1 ()

where the top line is the cohomology sequence of ( E, E) Z2 gen-

E). But H r ( E,
erated by the Thom class which, under the homomorphism H r ( E, E) H r ( E)

H r (B), goes to the Euler class. This proves the proposition. 

Let us consider the particular case of the Gysin sequence for an S 0 -bundle. Such
a bundle is simply a 2-fold covering = ( p : X X ). The Gysin sequence may
thus be compared to the transfer exact sequence of Proposition 4.3.9.

Proposition 4.7.36 Let = ( p : X X ) be a 2-fold covering (an S 0 -bundle).

Then, the Gysin and the transfer exact sequences of coincide, i.e. the diagram

H p tr w()
/ H k (X ) / H k ( X ) / H k (X ) / H k+1 (X ) /
O O O O
id id id id
   
H p e( p)
/ H k (X ) / H k ( X ) / H k (X ) / H k+1 (X ) /

is commutative. In particular, the Euler class e() H 1 (X ) and the characteristic

class w( p) H 1 (X ) are equal.

Proof By Corollary 4.3.3, is induced from = ( p : S RP ) by a

characteristic map f : X RP . Both the Gysin and transfer exact sequences
being functorial with respect to induced bundles, it suffices to prove the proposition
for . This is trivial since the vector spaces occurring in the diagram are either
equal to 0 or Z2 . 

The Thom isomorphism is classically used for vector bundles. Recall that a (real)
vector bundle of rank r is a map p : E B together with a R-vector space
structure on E b = p 1 (b) for each b B, satisfying the following local triviality
condition: there is an open covering U of B and for each U U, a homeomorphism

U : U Rr p 1 (U ) such that, for all (b, v) U Rr , p (b, v) = b and
U : {b} Rr E b is a R-linear isomorphism. In consequence, p is a bundle
4.7 Some Applications of the Knneth Theorem 187

with base B = B(), total space E = E() and fiber Rr . The map 0 : B E
sending b B to the zero element of E b is called the zero section of (it satisfies
p 0 (b) = b).
An Euclidean vector bundle is a vector bundle together with a continuous map
v  |v| R0 defined on E() whose restriction to each fiber is quadratic and
positive definite. Such a map is called an Euclidean structure (or Riemannian metric)
on . It is of course the same as defining a positive definite inner product on each
fiber which varies continuously. Vector bundles with paracompact basis admit an
Euclidean structure, [105, Chap. 3,Theorems 9.5 and 5.5]. If = ( p : E B) is an
Euclidean vector bundle, the restriction of p to

S(E) = {v E | |v| = 1} and D(E) = {v E | |v| 1}

gives the associated unit sphere and disk bundles. These bundles do not depend on
the choice of the Euclidean structure on . Indeed, using the map (v, t)  tv from
S(E) I D(V ) together with the zero section, the reader will easily construct

 S(E))
a homeomorphism ( S(E), (D(E), S(E)) over the identity of B, where
 S(E)) (B, B) is the mapping cylinder bundle pair of S(E) B. Thus, the
( S(E),
Thom class U H r (D(E), S(E)) exists by Lemma 4.7.27 and, by Theorem 4.7.29,

gives rise to the Thom isomorphisms : H k (B) H k+r (D(E), S(E)) and

: Hk+r (D(E), S(E)) Hk (B). Let E 0 = E 0 (B) and D(E)0 = D(E)E 0 .
By excision and homotopy, one has

H (E, E 0 )
H (D(E), D(E)0 )
H (D(E), S(E)) .

Hence the Thom class may be seen as an element U () H r (E, E 0 ) and one has
the following theorem.
Theorem 4.7.37 (The Thom isomorphism theorem for vector bundles) Let = ( p :
E B) be a vector bundle of rank r with B paracompact. Let U () H r (E, E 0 )
be the Thom class. Then, the homomorphisms

: H k (B) H k+r (E, E 0 ) and : Hk (E, E 0 ) Hkr (B)

given by
(a) = H p(a)  U () and () = H p(U ()  )

are isomorphism for all k Z.

Let = ( p : E B) be a vector bundle of rank r . The map E I E given
by (v, t)  tv is a retraction by deformation of E onto the zero section of . Hence,
H p : H (B) H (E) is a GrA-isomorphism. Therefore, there is a unique class
e() H r (B) such that H j (U ()) = H p(e()), where j : (E, ) (E, E 0 ).
The class e() is called the Euler class of (it coincides with the Euler e(S(E))
defined above). Lemma 4.7.31 and Corollary 4.7.33 imply the following two lemmas.
188 4 Products

Lemma 4.7.38 Let = ( p : E B) be a vector bundle of rank r , with B

paracompact. Let f : A B be a continuous map. Then the equality e( f ) =
H f (e()) holds in H r (A).
Lemma 4.7.39 Let = ( p : E B) be a vector bundle of rank r , with B
paracompact. If admits a nowhere zero section, then e() = 0.
Let i = ( pi : E i Bi ) (i = 1, 2) be two vector bundles of rank ri , with Bi
paracompact. The product bundle 1 2 is the vector bundle of rank r1 + r2 given
by p1 p2 : E 1 E 2 B1 B2 . If B1 = B2 = B, the Whitney sum 1 2 is
the vector bundle of rank r1 + r2 over B given by 1 2 =  (1 2 ) where
 : B B B is the inclusion of the diagonal, (x) = (x, x). The behavior of
the Euler class under these constructions is as follows.
Proposition 4.7.40
(1) e(1 2 ) = e(1 ) e(2 ).
(2) e(1 2 ) = e(1 )  e(2 ).
Proof Using Euclidean structures on i the Thom class U (i ) may be seen as an
element of H ri (D(E i ), S(E i )). Let E = E 1 E 2 , B = B1 B2 and r = r1 + r2 .
Let ji : (D(E i ), ) (D(E i ), S(E i )) and j : (D(E), ) (D(E), S(E))
denote the inclusions. There are homeomorphisms of pairs making the following
diagram commutative

j
(D(E), ) / (D(E), S(E))

.
 j1 j2

(D(E 1 ), ) (D(E 2 ), ) / (D(E 1 ), S(E 1 )) (D(E 2 ), S(E 2 ))

In the same way, if b = (b1 , b2 ) B, there is a homeomorphism of pairs

     
D(E)b , S(E)b D(E 1 )b1 , S(E 1 )b1 D(E 2 )b2 , S(E 2 )b2 . (4.7.23)

By the relative Knneth theorem 4.6.10, the generator of H r (D(E)b , S(E)b ) =

Z2 is the cross product of the generators of H ri (D(E i )bi , S(E i )bi ). Using
Lemma 4.7.28, we deduce that

U (1 2 ) = U (1 ) U (2 ) . (4.7.24)

Using Lemma 4.6.2, one has

H ( p1 p2 )(e()) = H j (U ())
= H j (U (1 ) U (2 ))
= H j1 (U (1 )) H j2 (U (2 ))
= H p1 (e(1 )) H p2 (e(2 ))
= H ( p1 p2 )(e(1 ) e(2 )) .
4.7 Some Applications of the Knneth Theorem 189

As H ( p1 p2 ) is an isomorphism, this proves (1). Point (2) is deduced from (1)

using the definition of 1 2 and Remark 4.6.1:

e(1 2 ) = H (e(1 2 )) = H (e(1 ) e(2 )) = e(1 )  e(2 ) . 

The Thom class of a product bundle was computed in (4.7.24). For the Whitney
sum, we use the projections i : E(1 2 ) E(i ).
Proposition 4.7.41 Let 1 and 2 be two vector bundles over a paracompact basis.
Let U (i ) H ri (D(E i ), S(E i )) be the Thom classes (for an Euclidean structure).
Then

U (1 2 ) = H 1 (U (1 ))  H 2 (U (2 )) .

Proof Restricted to the fiber over b B, the right hand side of the formula gives the
cross product of the generators of H ri (D(E i )b , S(E i )b ). The latter is the generator
of H r (D(E)b , S(E)b ). The proposition thus follows from Lemma 4.7.28. 

4.7.7 Bundles Over Spheres

In this section, we study bundles = ( p : E S m ) over the sphere S m with fiber F.

If A S m we set E A = p 1 (A). Consider the cellular decomposition of S m with one
0-cell b and one m-cell with characteristic map : D m S m sending S m1 onto b.
We denote by : S m1 {b} this constant map. We identify F with E b , getting thus
an inclusion i : F  E. As D m is contractible, any bundle over D m is trivial [181,
Corollary 11.6]. Therefore, there exists a trivialization E D m F of the induced
bundle . The map (, ) are covered by a bundle maps : D m F E and

: S m1 F F. The latter satisfies, for each x S m1 , that : {x} F F
is a homeomorphism. Observe that

E = (D m F) F . (4.7.25)

Let x0 S m1 be the base point corresponding to 1 S 0 S m1 . By changing

the trivialization of if necessary, we shall assume that : {x0 } F F is the
projection onto F. The map is called the bundle characteristic map and the map
is called the bundle gluing map of the bundle .
Lemma 4.7.42 The bundle characteristic map : D m F E induces an
isomorphism

H : H (E, F)
H (D m F, S m1 F) .

Proof Consider the decomposition D m = B C, where B is the disk with center 0

and radius 1/2 and C the closure of D m B; let S = B C. As (C) is a disk around
190 4 Products

b, the bundle is trivial above (C): E (C) (C) F. As : B F E (B)

is a homeomorphism, the lemma follows from the commutative diagram

H
H (E, F) / H (D m F, S m1 F)
O O
excision excision

H (E, E (C) ) H (D m F, C F)
excision excision
 
H
H (E (B) , E (S) ) / H (B F, S F)


Proposition 4.7.43 Let p : E S m be a bundle with fiber F. There is a long exact
sequence

H i
J H i
H k1 (E) H k1 (F)
H km (F)
H k (E) H k (F) .

The exact sequence of Proposition 4.7.43 is called the Wang exact sequence.
Proof We start with the exact sequence of the pair (E, F)

H i H j
H k1 (E) H k1 (F)
H k (E, F) H k (E) (4.7.26)

where j : (E, ) (E, F) denotes the pair inclusion. The following commutative
diagram defines the homomorphism
and J .

H j
H k1 (F) / H k (E, F) / H k (E)
D
H

H k (D m F, S m1 F) (4.7.27)

O J
e
'
H km (F)

Here, H is an isomorphism by Lemma 4.7.42, e H m (D m , S m1 ) = Z2

is the generator and the map e is an isomorphism by the relative Knneth
theorem 4.7.25. 
We now give some formulae satisfied by the homomorphism
: H k1 (F)
H km (F). We start with the case m = 1 which deserves a special treatment. The
bundle gluing map : S 0 F F satisfy (1, x) = x and (1, x) = h(x) for
4.7 Some Applications of the Knneth Theorem 191

some homeomorphism h : F F. The decomposition of (4.7.25) amounts to say

that E is the mapping torus Mh of h:
 
E = Mh = [1, 1] F {(1, x) (1, h(x))} .

The bundle projection p : Mh S 1 is given by p(t, x) = exp(2it). The corre-

spondence x [(x, 0)] gives an inclusion j : F  Mh . Let e H 1 (S 1 ) = Z2
be the generator. Proposition 4.7.43 may be rephrased and made more explicit in the
following way.

Proposition 4.7.44 (Mapping torus exact sequence) Let h : F F be a homeo-

morphism. Then, there is a long exact sequence

H i
J H i
H k1 (Mh ) H k1 (F)
H k1 (F)
H k (Mh ) H k (F) ,

with
= id + H h.

Proof We use the exact sequence (4.7.26) with E = Mh and Diagram (4.7.27). It
remains to identify
with id + H h. Let i : {1} F S 0 F denote the
inclusions. Let : H k1 (F) H k1 (S 0 F) be the homomorphism such that
H i + (a) = a and H i (a) = H h(a). Consider the diagram.
/ H k1 (S 0 F) i / H k1 ({1} F) H k1 ({1} F)
H k1 (F)

+
  
H
H k (Mh , F) / H k (D 1 F, S 0 F) o e
H k1 (F)

where i = (H i + , H i ). Let  : H k1 (F) H k1 (F) be the composed

homomorphisms through the upper right or lower left corners. Then + = id + H h
and =
. The left square of the diagram being commutative by construction of
Mh , it then sufficesto prove the commutativity
 of the right square, that is = ,

where (a) = e H i + (a) + H i (a) . The homomorphisms and are both
functorial. As, by Sect. 3.8, a class a H k1 (F) is represented by a map F Kk1 ,
it suffices to prove that = for F = Kk1 . Observe that and are both
surjective and have the same kernel, the image of H k1 (D 1 F) H k1 (S 0 F).
As H k1 (Kk1 ) = Z2 , this proves that = when F = Kk1 . 

Example 4.7.45 Let h : S 1 S 1 be the complex conjugation. Then, Mh is home-

p
omorphic to the Klein bottle K and we get a bundle S 1 K S 1 . The homo-
morphism
of Proposition 4.7.44 satisfies
= id + H h = 0. By the mapping
torus exact sequence, we deduce that H (K ) H (S 1 ) is surjective (this can also
be obtained using a triangulation like on p. 29 and computations like on p. 138). A
cohomology extension of the fiber : H (S 1 ) H (K ) produces, by the Leray-

Hirsch theorem 4.7.17, a GrV-isomorphism : H (S 1 ) H (S 1 ) H (K ).
192 4 Products

But is not a morphism of algebra. Indeed, the square map x  x x vanishes in

H (S 1 ) H (S 1 ) while the cup-square map x  x  x does not vanish in H (K )
(see, Proposition 4.2.3).

When m > 1, some information about the homomorphism

: H k1 (F)
H km (F) may be obtained via the composition

/ H km (F) / e / H k1 (S m1 F) ,
H k1 (F)

where e H m1 (S m1 ) = Z2 is the generator. The map e is injective by the

Knneth theorem.

Proposition 4.7.46 Suppose that m > 1. Then

e
(a) = H (a) H p2 (a) ,

p2 : S m1 F F are the bundle gluing map and the projection onto F.

where ,

Proof As F is a retract of S m1 F, the cohomology sequence of the pair

(S m1 F, F) splits into short exact sequences and, by the Knneth theorem and
Lemma 4.7.2, there is a commutative diagram
+
0 / H km (F) / H km (F) H k1 (F) / H k1 (F) / 0
SSSS
SSSe

SSSS
id
 SS)  
/ H k1 (S m1 F, F) / H k1 (S m1 F) H i / H k1 (F) / 0
0

(4.7.28)

where i : F S m1 F is the slice inclusion at the base point x0 S m1 and

(a, b) = e a + 1 b. Recall that we assume the restriction of to {x0 } F to
coincide with the projection p2 . Therefore, the composition

H H p
H k1 (F) H k1 (S m1 F) H k1 (F)
2

vanishes. Using Diagram (4.7.28), we get a factorization

H H p2
H k1 (F) / H k1 (S m1 F)
MMM  nn6
M

MMM nn nnn
M& nnn e

H km (F)
4.7 Some Applications of the Knneth Theorem 193

which we introduce in the diagram

/ H k (E, F)
H k1 (F)

H +H p
2 H
 
/ H k (D m F, S m1 F) .

 H k1 (S m1 F) (4.7.29)
O O

e e
'
= / H km (F)
H km (F)

We claim that the two square of Diagram (4.7.29) are commutative. By

Diagram (4.7.27), this will imply that
 =
and will prove the lemma.
As is the restriction of ,
the naturality of the connecting homomorphism

implies that H = H . Since p2 extends to D n F, the homomorphism

H p2 : H k1 (F) H k1 (S m1 F) factors through H k1 (D m F) and thus

H p2 = 0. Hence, the top square is commutative. For the bottom one, let a
H k1 (F). By Sect. 3.8, a = H f () for some map f from F into the Eilenberg-
MacLane space Kk1 . The bottom square being functorial for the map f , it suffices
to prove its commutativity for F = Kk1 . As the source and range vector space are
both then isomorphic to Z2 , the commutativity holds trivially. 

As an exercise, the reader may adapt the proof of Proposition 4.7.46 to the case
m = 0, thus getting an alternative proof of Proposition 4.7.44. The main point is to
replace e (which has no meaning in H 0 (S 0 ) by the class of {1}.
The family of homomorphisms
: H k1 (F) H km (F) forms an endomor-
phism of H (F) of degree m 1 (it sends H q (F) to H qm+1 (F)).

Proposition 4.7.47 As an endomorphism of H (F),

satisfies

(a  b) =
(a)  b + a 
(b) .

Proof Proposition 4.7.46 may be rephrased as

H (a) = H p2 (a) + e
(a) = 1 a + e
(a) .

Therefore, if a H p (F) and b H q (F),

 b) = 1 (a  b) + e
(a  b)
H (a

and, using Lemma 4.6.3,

   

H (a)
 H (b) = 1 a + e
(a)  1 b + e
(b)
= 1 (a  b) + e (
(a)  b) + e (a 
(b))
= 1 (a  b) + e
(a)  b + a 
(b) .
194 4 Products

 b) = H (a)
As H (a
 H (b) and e is injective, this proves the
proposition. 

Remark 4.7.48 The material of this section was inspired by [207, Sect. 1, Chap. VII].
As in this this reference, the following fact can also be proved:
(1) The Wang exact sequence holds for Serre fibrations. It also has a generalization
to bundles over a suspension.
(2) A Wang exact sequence for homology exists.
Further properties of the Wang sequences are given in [207, Sect. 2, Chap. VII].

4.7.8 The Face Space of a Simplicial Complex

Let K be simplicial complex. Fix an integer d > 0. For each v V (K ), consider a

copy Svd of the sphere S d . It is pointed by e1 = (1, 0, . . . , 0) Svd . For S(K ),
consider the space

Fd () = {(z v ) | z v = e1 ifv
/ } Svd ,
vV (K )

 
which is homeomorphic to v Svd . The face space of K is the subset of d
vV (K ) Sv
defined by
 
Fd (K ) = Fd () Svd .
S (K ) vV (K )

Remark 4.7.49 Let K be a flag simplicial complex (i.e. if K contains a graph L

isomorphic to the 1-skeleton of an r -simplex, then L is contained in an r -simplex
of K ). Then the complex F1 (K ) is the Salvetti complex of the right-angled Coxeter
group determined by the 1-skeleton of K (see [30]).

The interest of the face space appears in the following proposition, based on a
algebraic theorem of Gubeladze.

Proposition 4.7.50 Let K and K  be two finite simplicial complexes. Let d be a pos-
itive integer. Then, K is isomorphic to K  if and only if H (Fd (K )) and H (Fd (K  ))
are GrA-isomorphic.

To explain the proof of Proposition 4.7.50, we compute the cohomology alge-

bra of Fd (K ) for a finite simplicial complex K . Let us number the vertices of K :
V (K ) = {1, . . . , m}. Consider the polynomial ring Z2 [x1 , . . . , xm ] with formal
variables x1 , . . . , xm which are
of degree d. If J {1, . . . , m}, we denote by x J
Z2 [x1 , . . . , xm ] the monomial jJ x j . Let I(K ) be the ideal of Z2 [x1 , . . . , xm ]
4.7 Some Applications of the Knneth Theorem 195

generated by the squares xi2 of the variables and the monomials x J for J
/ S(K )
(non-face monomials). The quotient algebra

d (K ) = Z2 [x1 , . . . , xm ] I(K )

is called the face exterior algebra (because u2 = 0 for all u d (K ); however,

because the ground field is Z2 , d (K ) is commutative).
Lemma 4.7.51 The ring H (Fd (K )) is isomorphic to d (K ).
Proof (Compare [59, Proposition 4.3].) Let  K = F(V (K )) be the full complex
over the set V (K ) = {1, . . . , m}. The simplicial inclusion K  K induces an
inclusion

j : Fd (K )  Fd ( K ) = Svd .
vV (K )

For {1, . . . , m}, the fundamental class [Fd ()] H(dim +1)d (Fd ()) deter-
mines a class [] H(dim +1)d (Fd ( K )) (by convention, [] is the genera-
tor of H0 (Fd ( K ))). If S(K ), [] is the image under H j of a class in
H(dim +1)d (Fd (K )), also called []. Let

A = {[] H (Fd (K )) | S(K ) {}} H (Fd (K ))

and

B = {[] H (Fd ( K )) | {1, . . . , m}} H (Fd ( K )) .

By the Knneth theorem and Corollary 3.1.16, H (Fd (K )) is generated by A and B

is a basis of H (Fd ( K )). It follows that A is a basis of H (Fd (K )) and that H j is
injective. By Kronecker duality, H j is surjective and the Kronecker-dual basis B 
of B is sent onto Kronecker-dual basis A of A by

 [] if S(K )
H j ([] ) = (4.7.30)
0 otherwise.

By the Knneth theorem again,

H (Fd ( K )) Z2 [x1 , . . . , xm ] (x12 , . . . , xm2 ) (4.7.31)

and, if {1, . . . , m}, then [] = x . By (4.7.30), ker H j is the Z2 -vector

space in H (Fd ( K )) with basis {x | / S(K )}. Using (4.7.31), we check that,
H (Fd ( K )), ker H j is the image
under the epimorphism Z2 [x1 , . . . , xm ]
of I(K ). 
The proof of Lemma 4.7.51 provides the following corollary.
196 4 Products

Corollary 4.7.52 The Poincar polynomial of the algebra d (K ) is


Pt (d (K )) = 1 + t (dim +1)d .
S (K )

The proof of Proposition 4.7.50 follows from Lemma 4.7.51 and the following
theorem of Gubeladze. For a proof, see [76, Theorem 3.1].
Theorem 4.7.53 (Gubeladze) Let K and K  be two finite simplicial complexes. Sup-
pose that d (K ) = Z2 [x1 , . . . , xm ]/I(K ) and d (K  ) = Z2 [y1 , . . . , ym  ]/I(K  )
are isomorphic as graded algebras. Then m = m  and there is a bijection

: {x1 , , . . . , xm }
{y1 , . . . , ym  }

such that (I(K )) = I(K  ).

4.7.9 Continuous Multiplications on K (Z2 , m)

A continuous multiplication : X X X on a space X is homotopy commutative
if the maps (x, y)  (x, y) and (x, y)  (y, x) are homotopic. A element u X
is a homotopy unit for if the maps x  (u, x) and x  (x, u) are homotopic
to the identity of X . Note that, if u0 X is a homotopy unit for and if X is
path-connected, then any u X is also a homotopy unit.
Let K K (Z2 , m) be an Eilenberg-MacLane space in degree m, with its class
0 = H m (K). Recall from Sect. 3.8, the map : [X, K] H m (X ) given by

( f ) = H f () is a bijection. In particular, if K and K are two Eilenberg-MacLane
spaces in degree m, there is a homotopy equivalence g : K K whose homotopy
class is unique.
Proposition 4.7.54 Let K be an Eilenberg-MacLane space in degree m.
(1) There exists a continuous multiplication on K admitting a homotopy unit and
which is homotopy commutative.
(2) Any two continuous multiplications on K admitting a homotopy unit are homo-
topic.
(3) Let (K, ) and K ,  ) be two Eilenberg-MacLane spaces in degree m with
continuous multiplications admitting homotopy units, Let g : Km K a
m
(unique up to homotopy) homotopy equivalence. Then, the diagram


K K / K

gg g


KK / K

commutes up to homotopy.
4.7 Some Applications of the Knneth Theorem 197

Proof Consider the class

p = 1 + 1 H m (K K) . (4.7.32)

Since [K K, K] is in bijection with H m (K K), one has p = H () for some

continuous map : K K K, which we see as a continuous multiplication. The
involution exchanging the coordinates on K K satisfies H ( p) = p and then
H ( ) = H . Hence, is homotopic to , which says that is homotopy
commutative.
Choose u K and let i 1 , i 2 : K K K be the slice inclusions i 1 (x) = (x, u)
and i 2 (x) = (u, x). By Lemma 4.7.2, i j H () = for j = 1, 2. Hence, i j is
homotopic to the identity, which proves that u is a homotopy unit. Point (1) is thus
established.
For Point (2), let is continuous multiplication on K admitting a homotopy unit u.
Let i 1 , i 2 : K K K be the slice inclusions i 1 (x) = (x, u) and i 2 (x) = (u, x).
As u is a homotopy unit, h i j is homotopic to the identity for j = 1, 2, and thus
H i j H (a) = a for all a H (X ). By Lemma 4.7.2, this implies that

H (a) = a 1 + 1 a + y y , (4.7.33)

where the degrees of y and y  are both positive. By the Knneth theorem, the cross
product gives an isomorphism isomorphism H m (K) H 0 (K) H 0 (K) H m (K)
H m (K). Therefore, H () = p, which says that the homotopy class of is well
determined.
For Point (3), let h : K K  be a homotopy inverse for g. Then, the formula
(x, y) = h (g(x), g(y)) is a continuous multiplication of K with a homotopy


unit. By (2),  is homotopic to  , which proves (3). 

Examples 4.7.55 The following classical multiplications occur in Eilenberg-

MacLane spaces Km K (Z2 , m) (or more generally on K (G, m) for an abelian
group G).
The loop space Km+1 is an Eilenberg-MacLane space in degree m [82, pp. 407
and ff.]. One can use the loop multiplication.
Using semi-simplicial techniques, Milnor has shown that there exists an Eilenberg-
MacLane space Km which is an abelian topological group [146, Sect. 3].

The following property of the multiplication of Proposition 4.7.54 will be useful

in Sect. 8.3.

Lemma 4.7.56 Let K be an Eilenberg-MacLane space in degree m. Let a H k (K)

for m k < 2m. Then

H (a) = a 1 + 1 a . (4.7.34)
198 4 Products

Proof This comes from (4.7.33) since

H k (K) H 0 (K) H 0 (K) H k (K)
H k (K K) (4.7.35)

for m k < 2m by the Knneth theorem. 

Remark 4.7.57 Together with the cup product, the map H makes H (K (Z2 , m))
a Hopf algebra (see [82, Sect. 3.C]). In this setup, an element a H k (K (Z2 , m))
satisfying (4.7.34) is called primitive.

Let X be a CW-complex. The multiplication on K = K (Z2 ; m) induces a

composition law

[X, K] [X, K] [X, K]

given by f g (x) = ( f (x), g(x)). It admits the following interpretation.

Proposition 4.7.58 Let X be a CW-complex. Then, the bijection : H m (X )

[X, K ] satisfies
(a) (b) = (a + b) .

for all a, b, H m (X ).

Proof Let f, g : X K represent (a) and (b). Then (a) (b) is represented
by the composition
( f,g)
X K K
K.

The two projections 1 , 2 : K K K satisfy 1 ( f, g) = f and 2 ( f, g) = g.

Using that H () = 1 + 1 (see the proof of Proposition 4.7.54), one has

(a) (b) = H ( f, g) H ()
= H ( f, g)( 1 + 1 )
= H ( f, g)(H 1 () + H 2 ())
= H f () + H g()
= (a) + (b) . 

4.8 Exercises for Chapter 4

4.1. Write the proof of Lemma 4.1.8.
4.2. As H (S 1 S 1 ) has 4 elements, the bouquet of two circle has 4 inequivalent
2-fold coverings by the bijection (4.3.5). For each of them, describe the total
space and the transfer exact sequence.
4.8 Exercises for Chapter 4 199

4.3. Same exercise as the previous one, replacing S 1 S 1 by the Klein bottle.
Compare with the discussion on p. 33.
4.4. Write the transfer exact sequence for a trivial 2-fold covering.
4.5. Let p : X X be finite covering with an odd number of sheets. Prove that
H p is injective.
4.6. Let M and N be closed surfaces, with M orientable and N non-orientable.
Prove that there is no continuous map f : M N which is of degree one.
4.7. Show that there are no continuous map of degree one between the torus T and
the Klein bottle K , in either direction. Same things for S 1 S 2 and RP 3 .
4.8. Let M be a closed topological manifold of dimension n. Let h : D n M
be an embedding of the closed disk D n into M. Form the manifold M as the
quotient of M int h(D n ) by the identification h(x) h(x) for x Bd D n .
[Hint: express M as a connected sum.]
Compute the ring H ( M).
4.9. Show that the cohomology algebras of (S 1 S 1 )  RP 2 and of RP 2  RP 2
 RP 2 are GrA-isomorphic. (It is classical that these two spaces are homeo-
morphic: see [136, Lemma 7.1]).
4.10. Using the triangulation of the Klein bottle given in Fig. 2.4, compute all the
simplicial cap products. Check the formula a  b,  = a, b  .
4.11. Show that the smash product and the join of two homology spheres is a homol-
ogy sphere.
4.12. Compute the cohomology ring of (a) X = RP RP (n times); (b)
Y = CP 2 CP 3 ; (c) Z = CP 2 CP 3 .
4.13. Write the Mayer-Vietoris cohomology sequence for the decomposition

S 1 S n = [(S 1 {1})) S n ] [(S 1 {1})) S n ] .

and describe its various homomorphisms. If a H 1 (S 1 ) and b H n (S n ) are

the generators, describe how the elements a 1, 1 b and a b behave with
respect to the homomorphisms of the Mayer-Vietoris sequence.
4.14. Show that the product of two perfect CW-complexes is a perfect CW-complex.
4.15. What is the Lusternik-Schnirelmann category of RP 2 RP 3 ?
4.16. What is the Lusternik-Schnirelmann category of the n-dimensional torus T n =
S 1 S 1 (n times)?
4.17. Prove the relevant functoriality property for the homology cross product.
4.18. For
r m a positive integer, s let B(m) be a bouquet of m circles. Let X =
i=1 B(a i ) and Y = j=1 B(b j ). Suppose that X and Y have the same
Poincar polynomial. Prove that r = s and that bi = a(i) for some permuta-
tion .
4.19. Cap product in the (co)homology of X Y . Let X and Y be topological
spaces, with Y being of finite cohomology type. Let a H (X ), b H (Y ),
H (X ) and H (Y ). Prove that the formula
 
(a b)  1 ( ) = (a  ) (b  ) (4.8.1)
200 4 Products

holds in H (X ) H (Y ), using the (co)homology cross products and of

Sect. 4.6.
4.20. Slices in homology. Let X and Y be topological spaces, with Y being of finite
cohomology type. Let y0 Y and let s X : X X Y be the slice inclusion
of X at y0 . Let H (X ). Prove that
 
H s X () = y0 ,

where y0 is seen as a 0-homology class of Y , using the bijection Y S0 (Y ).

p
4.21. Let F E B be a locally trivial bundle containing a subbundle F0
p0
E 0 B. Prove that the cohomology sequence of (E, E 0 ) is a sequence of

H (B)-modules.
4.22. Let K be a finite simplicial complex and let Fd (K ) its face complex for an
integer d > 0. What is the relationship between the Euler characteristic of
Fd (K ) and that of K ?
4.23. Let p : E S n be a bundle with fiber , where  has the homology of the
sphere S n1 . Let e H n (S n ) be its Euler class. Prove that

(a) e = 0 if and only if H (E) H (S 2n1 ).

(b) e = 0 if and only if there is a GrV-isomorphism : H (E) H (S n
S n1 ).
(c) if n > 2, prove that the GrV-isomorphism in (b) is a GrA-isomorphism.

4.24. Let i : Q M be the inclusion of a smooth submanifold of codimension

r in a smooth manifold M. Let be the normal bundle to Q. Suppose that
H r (M) = 0. Prove that the Euler class e() H r (Q) vanishes.
Chapter 5
Poincar Duality

5.1 Algebraic Topology and Manifolds

Manifolds studied by algebraic topology tools occur in several categories: smooth,

piecewise linear, topological, homology manifolds, etc. Below are a few words about
this matter.
Henri Poincars paper analysis situs [161], published 1895, is considered as the
historical start of algebraic topology (for the prehistory of the field, see [163]).
The aim of Poincar was to use tools of algebraic topology in order to distinguish
smooth manifolds up to diffeomorphism (which he called homeomorphism). So,
differential and algebraic topology were born together. The importance of studying
smooth manifolds up to diffeomorphism was reaffirmed throughout the twentieth
century by many great mathematicians (Thom, Smale, Novikov, Atiyah, etc.). It is
based on the deep role played by global properties of smooth manifolds in analysis,
differential geometry, dynamical systems and physics.
After the failure of defining homology using submanifolds (see [40, Sect. I.3]),
Poincar initiated a new approach [162], in which smooth manifolds are equipped
with a triangulation. This permitted him to define what will later become simplicial
homology. The existence and essential uniqueness of smooth triangulations were of
course a problem, solved only in 1940 by Whitehead [208, Theorems 7 and 8]. Also,
besides some developments in the twenties (Veblen, Morse), the real foundations of
differential topology arose only after 1935 with the works of H. Whitney. As a result,
homology was seen for 3 decades as combinatorial in nature and smooth manifolds
were not considered as the right objects of study. In the prominent book written
in 1934 by Seifert and Threlfall [174], smooth manifolds are not even mentioned,
but replaced by a simplicial counterpart, i.e. combinatorial or piecewise linear (PL)
manifolds (see definition in Sect. 5.2). Techniques analogous to those for smooth
manifolds were later developed in the PL-framework (see [104]). Polyhedral homol-
ogy manifolds were later introduced (see Sect. 5.2), whose importance may grow
with the development of computational homology. For even more general objects,
like ANR homology manifolds, see e.g. [206].

Springer International Publishing Switzerland 2014 201

J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_5
202 5 Poincar Duality

Topological manifolds have also long attracted the attention of topologists, mostly
to know whether they carry smooth or piecewise linear structures (see, e.g. [8, p. 235],
[132, p. 183]). Their status however remained mysterious until the 1960s. Kirby
and Siebenmann produced examples in all dimension 5 of topological manifolds
without PL-structures and developed many techniques to deal with these topological
manifolds [116]. The field of topological versus smooth manifolds developed very
much in dimension four, after 1980, with the work of M. Freedman and S. Donaldson.
Poincar duality is one of the most remarkable properties of closed manifolds. In
its strong form, it gives, for a compact n-manifold M, that H k (M) and Hnk (M) are
isomorphic under the cap product with the fundamental class [M]. This result can
be obtained in two contexts:

by working with homology manifolds, using simplicial topology and dual cells.
Taking its origin in the early work of Poincar, this was achieved around 1930 in the
work by Pontryagin et al. (see [40, Sect. II.4.C]). In the next sections, we follow this
approach, akin to the presentation of [155, Chap. 8]. This proves Poincar duality
for triangulable topological manifold, whence for smooth manifolds. Observe that
smooth manifolds techniques (Morse theory or handle presentations) give an iso-
morphism from H k (M) to Hnk (M) but not the identification of this isomorphism
with a cap product (see e.g. [120, Sect. VII.6]).
by working with topological manifolds, using Cech cohomology techniques (see,
e.g. [179, Sect. 6.2] or [82, Sect. 3.3]). This is not done in this book.

5.2 Poincar Duality in Polyhedral Homology Manifolds

A polyhedral homology n-manifold is a simplicial complex such that, for each

Sk (M), the link Lk() of in M is a simplicial complex of dimension n k 1
which has the homology of the sphere S nk1 . (Recall that our homology is mod2
by default; thus, in a broader context, these objects may more accurately be called
polyhedral Z2 -homology manifolds).

Remark 5.2.1 (1) Let X be topological space satisfying the following local prop-
erty: for any x X ,

Z2 j = n
H (X, X {x}) =
j
0 j = n .

Such a space is called a homology n-manifold. For instance, an n-dimensional

topological manifold is a homology n-manifold by (3.3.1). The following result
is proven in e.g. [155, Theorem 63.2]: if K is a simplicial complex such that |K |
is a homology n-manifold, then K is a polyhedral homology n-manifold.
(2) Special kind of polyhedral homology n-manifolds are PL-manifolds. A sim-
plicial complex M is a PL-manifold, or a combinatorial manifold if, for each
5.2 Poincar Duality in Polyhedral Homology Manifolds 203

Sk (M), the link Lk() of in M has a subdivision isomorphic to a subdivi-

sion of the boundary of the (nk)-simplex. PL-manifolds were the combinatorial
objects replacing smooth manifolds for algebraic topologists around 1930.
(3) A smooth manifold M admits a so-called C 1 -triangulation, making M a PL-
manifold. Two C 1 -triangulations have isomorphic subdivisions. This was proven
by Whitehead in [208, Theorems 7 and 8].
(4) By a result of Edwards (see [128]), any PL-manifold of dimension 5 admits
non-PL triangulations (which are then polyhedral homology manifolds by (1)
above). It is an open problem whether a closed topological manifold of dimension
5 admits a (possibly non-PL) triangulation. This is wrong in dimension 4 (see
[168, Sect. 5]).
(5) There are polyhedral homological manifolds M such that |M| is not a topological
manifold. For instance, the suspension of a homology n-manifold N which has
the mod 2 homology of S n (a homology sphere) is an (n + 1)-dimensional
homology manifold. But there are many PL-homology sphere (even for integral
homology) with non-trivial fundamental group [115]. More examples are given,
for instance, by lens spaces with odd fundamental groups.

Here are two first consequences of the definition of a polyhedral homology

n-manifold.

Lemma 5.2.2 Let M be a polyhedral homology n-manifold. Then

(1) any simplex of M is contained in some n-simplex of M.
(2) any (n 1)-simplex of M is a face of exactly two n-simplexes of M.

Proof If v is a vertex of M, then Lk(v) is n 1 dimensional, so M is n-dimensional.

Let Sk (M). If Lk() = , must be an n-simplex by the above. If Lk() is not
empty, it must contain a (n k 1)-simplex . Then, is contained in the join
which is an n-simplex. This proves (1).
If Sn1 (M) then Lk() is a 0-dimensional complex having the homology of
S 0 . Hence, Lk() consists of two points, which proves (2). 

Let M be a finite polyhedral homology n-manifold. It follows from Point (2) of

Lemma 5.2.2 that the n-chain Sn (M) is a cycle and represents a homology class
[M] Hn (M), called the fundamental class of M.

Theorem 5.2.3 (Poincar Duality) Let M be a finite polyhedral homology

n-manifold. Then, for any integer k, the linear map

[M] : H k (M)
Hnk (M)

is an isomorphism.

The proof of this Poincar duality theorem will start after Proposition 5.2.8. We
first give some corollaries of Theorem 5.2.3. By Kronecker duality, we get
204 5 Poincar Duality

Corollary 5.2.4 (Poincar Duality, weak form) Let M be a finite polyhedral homol-
ogy n-manifold. Then, for any integer k,

dim Hk (M) = dim Hnk (M) .

Thus, in the computation of the Euler characteristic of M, the Betti numbers

essentially come in pairs, which gives the following corollary.
Corollary 5.2.5 Let M be a finite polyhedral homology n-manifold. Then, the Euler
characteristic (M) satisfies the following:
(1) if n is odd, then (M) = 0.
(2) if n = 2m, then (M) dim Hm (M) (mod 2).
Expressed in terms of Poincar polynomial, Corollary 5.2.4 has the following
form.
Corollary 5.2.6 Let M be a finite polyhedral homology n-manifold. Then,

Pt (M) = t n P1/t (M) .

Another easy consequence of Poincar duality is the following.

Corollary 5.2.7 A finite polyhedral homology n-manifold which is connected is an
n-dimensional pseudomanifold.
Proof Let M be a finite polyhedral homology n-manifold. We may suppose that M
is non-empty, otherwise there is nothing to prove. By Lemma 5.2.2, M satisfies Con-
ditions (a) and (b) of the definition of an n-dimensional pseudomanifold. If M is
connected, then H 0 (M) = Z2 . By Poincar duality, this implies that Hn (M) = Z2 .
Using Proposition 2.4.5, we deduce that M is an n-dimensional pseudomanifold. 
By Corollary 5.2.7, a continuous map between connected finite polyhedral man-
ifolds of the same dimension has a degree (see (2.5.4)).
Proposition 5.2.8 Let f : M M be a continuous map of degree one between
connected finite n-dimensional polyhedral manifolds. Then H f : H (M )
H (M) is surjective and H f : H (M) H (M ) is injective
Proof By Kronecker duality, only the homology statement needs a proof. The
hypotheses imply that H f ([M ]) = [M]. By Proposition 4.5.4, this implies that
the diagram
H f
H k (M ) o H k (M)
[M ] [M]
 
H f
Hnk (M ) / Hnk (M)

is commutative for all integer k 0. This provides a section for H f . 

5.2 Poincar Duality in Polyhedral Homology Manifolds 205

D(7)

5
D(15)

157 7
15
156
6 3

16
s(156)
1

Fig. 5.1 Dual cells and the map s : S (M) S (M ) of (5.2.10). In the simplex notation, brackets
and commas have been omitted: 156 = {1, 5, 6}, etc.

The remainder of this section is devoted to the proof of Theorem 5.2.3. We shall
introduce several simplicial or homology-cell complexes having all the homology
of M. Let M be the barycentric subdivision of M, with the notations introduced
in p. 11. The simplicial complex M is endowed with its natural simplicial order

defined in (2.1.2). For S(M), define D() S(M ) by (Fig. 5.1)

D() = {t S(M ) | = min t} .

The simplicial subcomplex D() of M generated by D() is called the dual cell
of . Observe that dim D() = n dim(). The simplicial subcomplex D() =
Lk(,
D()) is called the boundary of D(). Its dimension is one less than that of

D(). We are interested in the topological spaces E() = |D()| and E()
= |D()|.

Lemma 5.2.9 Let Sk (M), where M is a polyhedral homology n-manifold.

Then,

(a) the space E() is a homology (n k)-cell with boundary E();

(b) D() is an (n k 1)-dimensional pseudomanifold.

Proof The space E() is compact. Observe that D() is the cone over D(),
with cone vertex .
Hence, E() is the topological cone over E(). Therefore,

(E(), E()) is a good pair and H (E()) = 0. It then suffices to prove that
206 5 Poincar Duality
 

H (E())
H S nk1 . We shall see below that D() and Lk(, M) are iso-

morphic simplicial complexes. As |Lk(, M) | = |Lk(, M)| and M is a polyhedral
homology n-manifold, this implies that
   

H (E()) H |Lk(, M) | H (Lk(, M)) H S nk1 .

The simplicial isomorphisms p : D()

Lk(, M) and q : Lk(, M) D()
are defined as follows.
 

Let V D() . This implies that and S(M). Hence, =
S(Lk(, M)). We set p( ) = .
Let V (Lk(, M) ). Then, S(Lk(, M)), whence S(M). We set
 
q =  .
We check that p and q are simplicial maps and are inverse of each other. This proves
Point (a).
To prove Point (b), let Sk (M). We leave as an exercise to the reader that a
simplicial complex K is an m-dimensional pseudomanifold if and only if K is so.
Therefore, by Point (a) above and its proof, it is enough to prove that L = Lk(, M)
is (n k 1)-dimensional pseudomanifold.
Let S(L). By Lemma 5.2.2, the simplex is contained in some n-simplex
of M, which is of the form . Therefore, Snk1 (L). Now, if
Snk2 (L), then is a common face of exactly two n-simplexes of M (by
Lemma 5.2.2). Hence, is a common face of exactly two (n k 1)-simplexes
of L. We have proven that L satisfies Conditions (a) and (b) of the definition of an
(n k 1)-dimensional pseudomanifold. By Proposition 2.4.5, L is an (n k 1)-
dimensional pseudomanifold. 

Lemma 5.2.9 permits us to see |M| as a homology-cell complex (see p. 108). The
r -skeleton |M|r is defined by

|M|r = E() . (5.2.1)
Ss (M)
snr

Indeed, the space |M | is the disjoint union of its geometric open simplexes

 
|M | = |t| |t|
tS (M )

and each |t| |t| is contained in a single open dual cell E() E(), the one
associated to for which = min t. This shows that |M|n = |M|. If Snr (M),

then E() = E() |M|r 1 ; if Snr (M) is distinct from , the open dual
cells of and are disjoint. This shows that |M|r +1 is obtained by from |M|r by
adjunction of the family of r -homology cells:
5.2 Poincar Duality in Polyhedral Homology Manifolds 207
 
|M|r +1 = |M|r Snr (M) E() ,

where is the attaching map

:
Snr (M) E()
Snr (M) E() |M|r .

We denote by X the space |M| endowed with this (regular) homology-cell struc-
ture. As noted in p. 108, the cellular homology H (X ) (defined with the homology
cells) is isomorphic to the singular homology H (|M|) of |M|. If Sk (M), then

E() is the union of those E( ) for which Sk+1 (M) has as a face. Using
Formula (2.2.5), this amounts to


E() = E( ) . (5.2.2)
()

On the other hand, since D() is a (n k 1)-dimensional pseudomanifold by

Lemma 5.2.9, Proposition 2.4.4 tells us that Hnk1 (D()) = Z2 is generated

by [D()]
= Snk1 (D())
and the generator H nk1 (D()) = Z2 is represented

by any cochain formed by a single (n k 1)-simplex. Hence, Hnk (D(), D()) =

Z2 is generated by [D()] = Snk (D()) and the generator H nk (D(), D()) =
Z2 is represented by any cochain formed by a single (n k)-simplex of D(). The
proof of Lemma 3.5.5 thus works and, using (5.2.2), : C nk (X ) C nk1 (X )
satisfies 

(D()) = [D( )] . (5.2.3)
()

As M is a finite simplicial complex, C k (M) is isomorphic to the vector space gen-

erated by Sk (M). Therefore, the correspondence  E() gives a linear map
 1 : C k (M) C nk (X ) which, by (5.2.3), satisfies


1 = 1 . (5.2.4)

1 is bijective, the induced map

As 

H nk (X )
1 : H k (M) (5.2.5)

is an isomorphism. Observe that this proves the weak form of Poincar duality of
Corollary 5.2.5.
To prove Theorem 5.2.3, we now need to identify 1 with a cap product. The cor-
respondence E()  [D()] provides a linear map  2 : C nk (X ) Cnk (M ).
By (5.2.2), 2 is a chain map, thus inducing a linear map 2 : H nk (X )

Hnk (M ).

Lemma 5.2.10 2 : H nk (X ) Hnk (M ) is an isomorphism.

208 5 Poincar Duality

Proof The r -skeleton X r of the homology-cell decomposition of X was given

in (5.2.1). Note that X r = |K r | where K r is the subcomplex of M given by

Kr = D() . (5.2.6)
Ss (M)
snr

Thus, K r is a simplicial complex of dimension r . We can use the simplicial

pairs (K r , K r 1 ) to compute the simplicial homology of M . Define C r (M ) =
Hr (K r , K r 1 ) with the boundary : C r (M ) Cr 1 (M ) given by the composi-
tion

Hr (K r , K r 1 ) Hr 1 (K r 1 ) Hr 1 (K r 1 , K r 2 ) .

One has = 0. Set H (M ) = ker /Image The correspondence E() 
.

[D()] gives an isomorphism 2 : H r (X )
Hr (M ). Note that

C r (M ) = Hr (K r , K r 1 )


= ker Cr (K r )/Cr (K r 1 )
Cr 1 (K r )/Cr 1 (K r 1 ) ,


Cr (K r )

whence
 
C r (M ) = Cr (K r ) | Cr 1 (K r 1 ) Cr (K r ) Cr (M ) .

The inclusion  : C (M )  C (M ) is clearly a morphism of chain com-

2
plexes. It induces a homomorphism  2 : H (M ) H (M ). As in the proof of
Theorem 3.5.3, we have the commutative diagram

Hr +1 (K r +1 , K r )
0 / Hr (K r +1 )
QQQ 77
QQQr +1 oo o
o
QQQ o o
QQ( ooo 
r +1 Hr (K r ) / / Hr (M )
j mmmm
mv
mm m
 mv mm
Hr (K r , K r 1 ) 0 (5.2.7)
QQQ
QQQr
QQQ
QQ( 
r Hr 1 (K r 1 )
v
j mmmmm
mmm
 vmmm
Hr (K r 1 , K r 2 )
5.2 Poincar Duality in Polyhedral Homology Manifolds 209

which permits us to compute H (M ). As 

is just the inclusion, the diagram
2

=
Hr (K r )/Imr +1 / ker r /Imr +1 / ker r /Imr +1

j =
  2

Hr (M ) o Hr (M )

is commutative, which proves that  2 is an isomorphism. Finally, the commutative

diagram
2
H r (X ) / Hr (M )
JJJ   2 ttt
9
JJJ2 t t
JJ$ t
tt
Hr (M )

shows that 2 is an isomorphism. 

We now need a good identification of the simplicial (co)homology of M with that
of M . Choose a simplicial order  on M. One has a simplicial map g : M M
given, for Sm (M), by

= max .
g() (5.2.8)

In the other direction, one has a chain map sd : Cm (M) Cm (M ) given, for
Sm (M), by

sd() = Sm ( ) . (5.2.9)

(This chain map is in fact defined for any subdivision and is called the subdivision
operator). Observe that, for any Sm (M), there exists a unique Sm (M )
such that C g( ) = . Indeed, if = {v0 , v1 , . . . , vm } with v0  v1   vm ,
then = { 0 , 1 , . . . , m }, where i is the barycenter of {v0 , v1 , . . . , vi }. The other
m-simplexes of are mapped to proper faces of . This defines a map

s : S(M) S(M ) (5.2.10)

by s() = . For S(M), one has

C g C sd() = C g s() = , (5.2.11)

which proves that H g H sd = id H (M) .

On the other hand, if t = { 0 , 1 , . . . , m } Sm (M ), with 0 m ,
then sd g(t) C (m ). As, also t C ( ), the correspondence t  C ( ) is
m m
an acyclic carrier for both sd g and idC (M ) . By Proposition 2.9.1, this implies that
210 5 Poincar Duality

H sd H g = id H (M ) . Therefore, g and sd induce isomorphisms in (co)homology

which are inverse of each other. In particular, H g et H g do not depend on the order
 since this is the case for sd.
It is straightforward that sd([M]) = [M ]. As g : M M is a simplicial map,
Proposition 4.5.4 gives the formula
 
H g H g(a)  [M ] = a  [M] ,

which is equivalent to the commutativity of the diagram.

[M]
H k (M) / Hnk (M)
O
H g H g (5.2.12)

[M ]
H (M )
k / Hnk (M )

The identification of the isomorphism 1 with the cap product with the fundamental
class then follows from the following lemma.

Lemma 5.2.11 The diagram

1
H k (M) / H nk (X )

H g 2
 
[M ]
H k (M ) / Hnk (M )

is commutative.

Proof Let Sk (M). The properties of the map s : S(M) S(M ) defined
in (5.2.10) imply that C g() = s() and max s() = (for the natural simplicial
order on M defined in (2.1.2)). The isomorphism  = 2 1 comes from the
morphism of cochain-chains  : C (M) Cn (M ) such that

() = [D()] = {t Snk (M ) | min t = }

= {t Snk (M ) | s() t Sn (M )} .

On the other hand, if = { 0 , . . . n } Sn (M ) with 0 1 n ,

Formula (4.5.1) gives

C g()  = s()  = s(), { 0 , . . . , k } { k , . . . , n .

5.2 Poincar Duality in Polyhedral Homology Manifolds 211

But

1 if s() = { 0 , . . . , k }
s(), { 0 , . . . , k } =
0 otherwise

1 if s() { k , . . . , n } Sn (M )
=
0 otherwise.

Therefore

C g()  [M ] = s()  [M ] = () . 

The proof of Poincar Duality Theorem 5.2.3 is now complete.

5.3 Other Forms of Poincar Duality

5.3.1 Relative Manifolds

A topological pair (X, Y ) such that


Z2 j = n
H j (X, X {x}) =
0 j = n .

for any x X Y is called a relative homology n-manifold. The condition is for

instance fulfilled if X Y is n-dimensional topological manifold, by (3.3.1).
A simplicial pair (M, A) is a relative polyhedral homology n-manifold if, for each
Sk (M) Sk (A), the link Lk() of in M is a simplicial complex of dimension
n k 1 which has the homology of the sphere S nk1 . For instance, (M, ) is a
relative polyhedral homology n-manifold if and only if M is a polyhedral homology
n-manifold.
The following result is proven in e.g. [155, Theorem 63.2].
Proposition 5.3.1 If (K , L) is a simplicial pair such that (|K |, |L|) is a relative
homology n-manifold, then K is a relative polyhedral homology n-manifold.
A topological pair (X, Y ) is triangulable if there exists a simplicial pair (K , L)
and a homeomorphism of pair h : (|K |, |L|) (X, Y ). Such a homeomorphism h
is called a triangulation of (X, Y ).
Theorem 5.3.2 (Lefschetz duality) Let (X, Y ) be a compact relative homology n-
manifold which is triangulable. Then, for any integer k, there is an isomorphism

 : H k (X, Y ) Hnk (X Y ) .

For a more general result, see [43, Proposition 7.2 in Chap. VII].
212 5 Poincar Duality

Proof Let (M, A) be a simplicial pair such that (|M|, |A|) is homeomorphic to
(X, Y ). By Proposition 5.3.1, (M, A) is a relative polyhedral homology n-manifold.
We shall construct an isomorphism

H nk (|M| |A|) ,
0 : H k (M, A) (5.3.1)

where H k (M, A) is the simplicial cohomology. The proof is close to that of

Theorem 5.2.3, so we just sketch the argument. For more details (see [155,
Theorem 70.2]).
Let M be the subcomplex of the first barycentric subdivision of M consisting
of all simplexes of M that are disjoint from A. As in the proof of Theorem 5.2.3,
consider the dual cell D() for each S(M) S(A) and its geometric realization
E() = |D()|. Lemma 5.2.9 holds for these dual cells and, as in the proof of
Theorem 5.2.3, they provide a structure of a homology-cell complex on |M |. Call
X the space |M | endowed with this homology-cell decomposition. As M is a finite
complex, then C k (M, A) is the vector space with basis Sk (M) Sk (A) (see p. 41).
As in (5.2.5), the correspondence  E() produces an isomorphism

H nk (X ) .
1 : H k (M, A) (5.3.2)

To get the isomorphism 0 from 1 , we use that |M | X is a deformation retract

of |M| |A| (see [155, Lemma 70.1]). 

Corollary 5.3.3 Let (X, Y ) be a connected compact relative homology n-manifold

which is triangulable. If Y = , then H n (X Y ) Hn (X Y ) = 0.

Proof By Kronecker duality, it is enough to prove that Hn (X Y ) = 0. By

Theorem 5.3.2, Hn (X Y ) H 0 (X, Y ) and, as X is path-connected, H 0 (X, Y ) = 0
if Y = . (Corollary 5.3.3 may also be obtained using cohomology with compact
supports: see [82, Theorem 3.35]). 

The following consequence of Corollary 5.3.3 is often referred to as the

Z2 -orientability of finite polyhedral homology n-manifolds (see e.g. [82, pp. 235
236]).

Corollary 5.3.4 Let M be a finite polyhedral homology n-manifold and let x M.

We denote by j : (M, ) (M, M {x}) the pair inclusion. Then

H j : Hm (M) Hm (M, M {x})

sends [M] onto the generator of Hn (M, M {x}) Z2 . In particular, if M is

connected, H j is an isomorphism.

Proof The fundamental class of M being the sum of those of its connected compo-
nents, it is enough to consider the case where M is connected. Corollary 5.3.4 then
follows from the exact sequences
5.3 Other Forms of Poincar Duality 213

H j
Hm (M {x}) Hm (M) Hm (M, M {x}) ,

using that Hm (M {x}) = 0 by Corollary 5.3.3. 

Let (X, Y ) be a compact triangulable relative homology n-manifold. Choose

a simplicial pair (M, A) such that (|M|, |A|) is homeomorphic to (X, Y ). Then,
(M, A) is a finite relative polyhedral homology n-manifold by Proposition 5.3.1.
Lemma 5.2.2 holds true for the simplexes of M which are not in A. As a consequence,
the n-chain Sn (M) Sn (A) is a cycle relative to A and represent a homology class
[M] Hn (M, A) called the fundamental class of (M, A). Under the isomorphism
between simplicial and singular homology of Theorem 3.6.3, the class [M] corre-
sponds to a singular class [X ] Hn (X, Y ) called fundamental class of (X, Y ). Let
i : X Y  X denote the inclusion. The isomorphism  of Theorem 5.3.2 is
related to the cap product with [X ] in the following way.

Proposition 5.3.5 Let (X, Y ) be a compact relative homology n-manifold which is

triangulable. Then the diagram

 / Hnk (X Y )
H k (X, Y )
OOO
OOO
[X ] OOO H i
' 
Hnk (X )

is commutative.

Proof As in the proof of Theorem 5.3.2, we choose a finite relative polyhedral

homology n-manifold (M, A) such that such that (|M|, |A|) is homeomorphic to
(X, Y ) and we use the same definitions and notations, such that X M . The
isomorphism 2 : H nk (X ) Hnk (M ) may be established as in Lemma 5.2.10.
The subdivision operator sd : Cm (M) Cm (M ) of (5.2.9) is defined, as well as
the simplicial map g : M M of (5.2.8), choosing for the latter a simplicial order
on M. They induced reciprocal isomorphisms on (co)homology. Ons has sd([M]) =
[M ], where [M ] Hn (M , A ) is the class of the relative cycle Sn (M ) Sm (A ).
The commutative diagram (5.2.12) becomes

[M]
H k (M, A) / Hnk (M)
O
H g H g (5.3.3)

[M ]
H (M , A )
k / Hnk (M )

If i : M M denotes the simplicial map given by the inclusion, the commutativity

of the diagram
214 5 Poincar Duality

1 2
H k (M, A) / H nk (X ) / Hnk (M )

H g H i
 
[M ]
H (M , A )
k / Hnk (M )

is proven as in Lemma 5.2.11. Finally, as mentioned in the proof of Theorem 5.3.2,

|M | X is a deformation retract of |M| |A|, hence a commutative diagram
involving simplicial and singular homology:
/ H (|M| |A|)
H (M )
H i H j . (5.3.4)
 
/ H (|M|)
H (M ) 
In the definition of a relative homology n-manifold (X, Y ), it is not required
that X itself is a homology n-manifold. If this is the case (and if X is compact
and triangulable), the fundamental class [X ] Hn (X ) is defined. To distinguish,
call [X ]r el Hn (X, Y ) the class of Proposition 5.3.5. If (M, A) is a simplicial
pair with (|M|, |A|) homeomorphic to (X, Y ), then H j ([M]) = [M]r el , where
j : (M, ) (M, A) (or j : (X, ) (X, Y )) denote the pair inclusion. Therefore
H j ([X ]) = [X ]r el .
Proposition 5.3.6 Let X be a compact homology n-manifold and let Y be a closed
subset of X . Assume that the pair (X, Y ) is triangulable. Then (X, Y ) is a relative
homology n-manifold and the diagram


H k (X, Y ) / Hnk (X Y )
OOO
OO[X
OOO]r el
H j H i
 OO' 
[X ]
H k (X ) / Hnk (X )

is commutative. Here, j : (X, ) (X, Y ) is the inclusion and is the Lefschetz

duality isomorphism of Theorem 5.3.2.
Proof Only the commutativity of the diagram requires a proof. The commutativity
of the upper triangle is established in Proposition 5.3.5. For the lower triangle, let
a H k (X, Y ) and u H nk (X ). One has

u, a  [X ]r el  = u  a, [X ]r el 
= u  a, H j ([X ]) as [X ]r el = H j ([X ])
= H j (u  a), [X ] (5.3.5)
= u  H j (a), [X ] by Lemma 4.1.8
= u, H j (a)  [X ] ,

which is, in formula, the commutativity of the lower triangle. 

5.3 Other Forms of Poincar Duality 215

5.3.2 Manifolds with Boundary

Let X be a compact topological n-manifold with boundary Y = Bd X . Then (X, Y )

is a compact relative homology n-manifold. As seen in the previous subsection, if
the pair (X, Y ) is triangulable, the fundamental class [X ] Hn (X, Y ) is defined.

Theorem 5.3.7 Let X be a compact topological n-manifold with boundary Y =

Bd X . Suppose that the pair (X, Y ) is triangulable. Then, for any integer k, the
linear maps

[X ] : H k (X )
Hnk (X, Y )

and

[X ] : H k (X, Y )
Hnk (X )

given by the cap product with [X ] Hn (X, Y ) are isomorphisms.

Theorem 5.3.7 is also true without the hypothesis of the triangulability of (X, Y ),
[82, Theorem 3.43].

Proof We first establish the isomorphism.

[X ] : H k (X, Y )
Hnk (X ) .

As X is a topological manifold, its boundary admits a collar neighbourhood, i.e. there

exists a embedding h : Y [0, 1) X , extending the identity on Y (see, e.g. [82,
Proposition 3.42]). Then, X h(Y [0, 1/2]) is a deformation retract of both X and
X Y . It follows that the inclusion X Y  X is a homotopy equivalence. Hence,
the result follows from Proposition 5.3.5.
The other isomorphism comes from the five lemma applied to the diagram

/ H k (X, Y ) / H k (X ) / H k (Y ) / H k+1 (X, Y )

[X ] [X ] [Y ] [X ]

   

/ Hnk (X ) / Hnk (X, Y ) / Hnk1 (Y ) / Hnk1 (X )

The commutativity of the above diagram comes from Lemma 4.5.5, since

([X ]) = [Y ] . (5.3.6)

Indeed, if (M, N ) be a finite a simplicial pair triangulating (X, Y ), the fundamental

class [M] is represented by the chain Sn (M) C (M) and ([M]) is represented
by (Sn (M)) = Sn1 (N ). 
216 5 Poincar Duality

Corollary 5.3.8 Let X be a compact triangulable topological n-manifold with

boundary Bd X = Y . Suppose that is Y = Y1 Y2 the union of two compact
(n 1)-manifolds with common boundary Y1 Y2 = Bd Y1 = Bd Y2 . Then, for any
integer k, the linear map

[X ] : H k (X, Y1 )
Hnk (X, Y2 )

given by the cap product with [X ] Hn (X, Y ) is an isomorphism.

Again, Corollary 5.3.8 is true without the hypothesis of triangulability (see [82,
Theorem 3.43]).
Proof Corollary 5.3.8 reduces to Theorem 5.3.7 by applying the five lemma to the
diagram

/ H k (X, Y ) / H k (X, Y1 ) / H k (Y, Y1 ) / H k+1 (X, Y ) /

[X ] [X ] [X ]

   
/ Hnk (X ) / Hnk (X, Y2 ) / Hnk1 (Y2 ) / Hnk1 (X ) /

The top line is the cohomology sequence for the triple (X, Y, Y1 ) and the bottom line
is the homology sequence for the pair (X, Y2 ). The isomorphism is the composition

[Y2 ]
: H k (Y, Y1 )
H k (Y2 , Bd Y2 ) Hnk1 (Y2 ) .

The commutativity of the above diagram is obtained as for those in the proofs
in Sect. 5.3.1. 
Here are some applications of the Poincar duality for compact manifolds with
boundary.
Proposition 5.3.9 Let X be a compact
 triangulable manifold
 of dimension 2n + 1,
with boundary Y . Let B = Image H n (X ) H n (Y ) . Then
(1) Let u H n (Y ). Then

u B u  B, [Y ] = 0 .

In particular, B  B, [Y ] = 0.
(2) dim H n (Y ) = 2 dim B.
For example, RP 2n is not the boundary of a compact manifold.
Proof We follow the idea of [133, Lemma 4.7 and Corollary 4.8]. Let i : Y X
denote the inclusion and let a, b H n (X ). Then,

H i(a)  H i(b), [Y ] = H i(a  b), [Y ] = a  b, H i([Y ]) = 0 ,

5.3 Other Forms of Poincar Duality 217

since H i([Y ]) = 0 by (5.3.6). This proves the implication of (1). Conversely,

suppose that u  B, [Y ] = 0 for u H n (Y ). Since B = ker( : H n (Y )
H n+1 (X, Y ), it suffices to prove that (u) = 0. Let v H n (X ). One has

0 = u  H i(v), [Y ]
= u  H i(v), [X ] by 5.3.6
= (u  H i(v)), [X ]
= (u)  v, [X ] by Lemma 4.1.9

This equality, holding for any v H n (X ), implies, by Theorem 5.3.12, that (u) = 0.
To prove (2), let us consider the linear map  : H n (Y ) H n (Y ) given by
(a)(b) = a  b, [Y ]). Let  B be the restriction of  to B. The map  is an
isomorphism by Theorem 5.3.13. By (1),  B (B) = A , where A = H n (Y )/B.
Thus, there is a quotient map  fitting in the commutative diagram

0 / B / H n (Y ) / A / 0

B  
(5.3.7)
  
0 / A / H n (Y ) / B / 0

is also an isomorphism. Therefore,

(whose rows are exact) and 

dim H n (Y ) = dim B + dim A = 2 dim B .

only that  B is an isomorphism; but

(Remark: the proof does not use the map ,
Diagram (5.3.7) will be useful later). 
Proposition 5.3.9 and Corollary 5.2.4 have the following consequence on the Euler
characteristic of bounding manifolds.
Corollary 5.3.10 Let Y be a closed triangulable n-manifold. If Y is the boundary
of a compact triangulable manifold, then (Y ) is even.

5.3.3 The Intersection Form

Let X be a compact topological n-manifold with boundary Y = Bd X . We assume

that the pair (X, Y ) is triangulable. From Theorem 5.3.7, the cap product with

[X ] Hn (X, Y ) induces isomorphisms H q (X ) Hnq (X, Y ) and H q (X, Y )

Hnq (X ). We denote by PD the inverse of these isomorphisms. Thus, if Hq (X )
and Hq (X, Y ), their Poincar dual PD() H nq (X, Y ) and PD()
H nq (X ) are the classes determined by the equations

PD()  [X ] = and PD()  [X ] = .

218 5 Poincar Duality

(The first equation uses the cap product of (4.5.16) and the second that of (4.5.14)).
This permits us to define two intersection forms on the homology of X .
(1) If Hq (X ) and Hnq (X ), we set

a = PD()  PD(), [X ] .
a
This defines the (absolute) intersection form H (X ) Hn (X )
Z2 .
(2) Similarly, if Hq (X ) and Hnq (X, Y ), the same formula defines the
r
(relative) intersection form H (X ) Hn (X, Y )
Z2 .
The name intersection form will be justified by Corollary 5.4.13.
Let j : (X, ) (X, Y ) denote the pair inclusion. For Hq (X ) and
Hnq (X ), the absolute and relative intersection forms are related by the formula

a = H j () r = H j () r .

Indeed:

H j () r = PD(H j ())  PD(), [X ]

= (H j (PD())  PD(), [X ] by Lemma 4.5.5
= PD()  PD(), [X ] by Lemma 4.1.7
= a

and the other equality is proven the same way.

The absolute and relative intersection forms coincide when Y is empty. Even
when Y = , we shall usually not distinguish between the two forms and just write
when the context makes it clear. In both cases, since a  b,  = a, b  
[see (4.5.2)], one has

= PD(),  = PD(),  . (5.3.8)

By Theorem 5.3.13, the relative intersection form is non-degenerate, i.e. induces

Hnq (X, Y ) for all q. If Y = , the absolute intersection
an isomorphism Hq (X )
form may be degenerate (example: X = S 1 D 2 ). In fact, if X is connected, it is
always degenerate for q = 0, since Hn (X ) = 0. However, one has the following
proposition.
Proposition 5.3.11 Suppose that X is connected and that Y is not empty. Then, the
following conditions are equivalent.

Hnq (X )
(a) The absolute intersection form induces an isomorphism Hq (X )
for 1 q n 1.
(b) Y = Bd X is a Z2 -homology sphere.
Proof Let j : (X, ) (X, Y ) denote the pair inclusion. By (5.3.8), the composed
homomorphism
5.3 Other Forms of Poincar Duality 219

Hq j
Hq (X ) / Hq (X, Y ) PD / H nq (X ) k / Hnq (X )

is just the absolute intersection form of X . Thus, (a) is equivalent to Hq j being an

isomorphism for 1 q n 1. By the exact homology sequence of (X, Y ) this is
equivalent to (b) if X is connected. 

5.3.4 Non Degeneracy of the Cup Product

Theorem 5.3.12 Let M be a finite polyhedral homology n-manifold. Then, for any
integer k, the bilinear map

 ,[M]
H k (M) H nk (M)
H n (M) Z2

H nk (M) .
induces an isomorphism H k (M)

Proof By Corollary 5.2.4, it suffices to prove that the linear map  : H k (M)
H nk (M) given by


 
a  b  a  b, [M]

is injective. Suppose that a ker . Then

0 = a  b, [M] = b, a  [M]

for all b H nk (M). By Point (a) of Lemma 2.2.3, we deduce that a  [M] = 0,
which implies that a = 0 by Theorem 5.2.3. 

The same proof, using Corollary 5.3.8, gives the following result.

Theorem 5.3.13 Let X be a compact triangulable topological n-manifold with

boundary Bd X = Y . Suppose that Y = Y1 Y2 is the union of two compact
(n 1)-manifolds Yi with common boundary Y1 Y2 = Bd Y1 = Bd Y2 . Then, for
any integer k, the bilinear map

 ,[X ]
H k (X, Y1 ) H nk (X, Y2 )
H n (X, Y ) Z2

H nk (X, Y2 ) .
induces an isomorphism H k (X, Y1 )
220 5 Poincar Duality

5.3.5 Alexander Duality

The first version of Alexander Duality was proven in a paper [7] of Alexander II
(18881971). This article pioneered several new methods and was very influential
at the time (see [40, p. 56]). In his paper, Alexander used the mod 2 homology.
Classical Alexander duality relates the cohomology of a closed subset A or S n to the
homology of S n A. We give below a version where S n is replaced by a homology
sphere (for instance a lens space with odd fundamental group).

Theorem 5.3.14 (Alexander Duality) Let (X, A) be a compact triangulable pair

with = A = X . Suppose that X is a relative homology n-manifold and has
its homology isomorphic to that of S n . Then, for all integer k, there is an isomor-
phism

H k (A) H nk1 (X A) .

Particular case of Alexander duality were encountered in Proposition 3.3.6 and

Corollary 3.3.7. For a version of Theorem 5.3.14 without the assumption of triangu-
lability (see [82, Theorem 3.44]).

Proof The case n = 0 being trivial, we assume n > 0. The pair (X, A) satisfies the
hypotheses of Lefschetz duality Theorem 5.3.2. This gives an isomorphism

 : H k+1 (X, A) Hnk1 (X A) .

Suppose that k = n, n 1. Since H (X ) H (S n ), the connecting homomorphism

: H k (A) H k+1 (X, A) is an isomorphism and Hnk1 (X A) H nk1 (X
A). This proves the result in this case.
Let (M, L) be a simplicial pair such that (|M|, |L|) is homeomorphic to (X, A).
As M is a relative polyhedral homology n-manifold by Proposition 5.3.1. As L is
a proper subcomplex of M, one has Hn (L) = 0, since Sn (M) is the only non-
vanishing n-cycle of M. Hence, H n (A) H n (A) = 0 by Kronecker duality. As,
H nk1 (X A) = H 1 (X A) = 0, the theorem is true for k = n.
When k = n 1, consider the diagram

H i / H n1 (A) / H n (X, A) / H n (X ) / 0
H n1 (X )

  [X ]
  
H j
0 / H (X A) / H0 (X A) / H0 (X ) / 0
0

where i : A X and j : (X A) X denote the inclusions. The bottom line is the

exact sequence of Lemma 3.1.10 and the commutativity of the right hand square is the
contents of Proposition 5.3.6. Then the homomorphism : H n1 (A) H 0 (X A)
exists, making the diagram commutative. If n > 1, H n1 (A) = H n1 (A) and, as
5.3 Other Forms of Poincar Duality 221

H n1 (X ) = 0, the map is an isomorphism by the five lemma. Finally, when n = 1,

then coker H i = H 0 (A) by Lemma 3.1.10 and induces an isomorphism from
H n1 (A) to H 0 (X A). 

5.4 Poincar Duality and Submanifolds

In this section, we assume some familiarity of the reader with standard techniques
of smooth manifolds, as exposed in e.g. [95].

5.4.1 The Poincar Dual of a Submanifold

Let M be a smooth compact n-manifold and let Q M be a closed smooth sub-

manifold of codimension r . Recall that smooth manifolds admit PL-triangulations
[208], so the fundamental classes [Q] Hnr (Q) and [M] Hn (M, Bd M) do
exist. We are interested in the Poincar dual PD(H i([Q])) H r (M, Bd M) (see
Sect. 5.3.3) of the class H i([Q]) Hnr (M), where i : Q M denotes the
inclusion. We write PD(Q) for PD(H i([Q])) and call it the Poincar dual of Q. It
is thus characterized by the equation

PD(Q)  [M] = H i([Q]) ,

Two simple examples are given in Fig. 5.2.

Example 5.4.1 For a more elaborated example, let Q be a smooth closed con-
nected manifold, seen as the diagonal submanifold of M = Q Q. Let A =
{a1 , a2 , . . . } H (Q) be an additive basis of H (Q). By Theorem 5.3.12, there is
a basis B = {b1 , b2 , . . . } of H (Q) which is dual to A for the Poincar duality, i.e.

1 2 3 1 1 2 3 1

5 6 5 6
4 4 4 7
PD(Q) PD(Q)
8 9 8 9
7 7 7 4
Q Q
1 2 3 1 1 2 3 1
W W

Fig. 5.2 The Poincar dual PD(Q) of a circle Q in the torus (left) or the Klein bottle (right). This
illustrates the localization principle of Remark 5.4.3: PD(Q) is supported in a tubular neighbourhood
W of Q
222 5 Poincar Duality

ai  b j , [Q] = i j . We claim that


PD(Q) = ai bi . (5.4.1)
i

Indeed, by the Knneth theorem, ai b j is a basis of H (M), so there are unique

coefficients i j Z2 such that

PD(Q) = i j ai b j .
i, j

Let  : Q M be the diagonal inclusion, a p A and bq B. As H (b p aq ) =

b p  aq (see Remark 4.6.1), one has

b p aq , H ([Q]) = H (b p aq ), [Q] = b p  aq , [Q] = pq . (5.4.2)

Without loss of generality, we may suppose that Q is connected. Let [Q] be the non
zero element of H dim Q (Q). One has [Q] [Q] , [M] = 1 and

b p aq , H ([Q]) = (b p aq )  PD(Q), [M]


= i j (b p aq )  (ai bi ), [M]
i, j

= i j (b p  ai ) (aq  bi ), [M] by Remark 4.6.4
i, j

= i j  pi [Q] q j [Q] , [M]
i, j
= pq . (5.4.3)

Thus, Equation (5.4.1) follows from (5.4.2) and (5.4.3).

The next two lemmas are recipes to compute PD(Q). Let us denote by =
(M, Q) the normal bundle of Q in M. A Riemannian metric provides a smooth
bundle pair (D(), S()) with fiber (Dr , Sr 1 ) and there is a diffeomorphism from
D() to a closed tubular neighbourhood W of Q in M. By excision,

H (M, M Q)
H (W, Bd W ) H (D(), S()) .

Hence, the Thom class U () H r (D(), S()) determines an element U (M, Q)

H r (M, M Q). Let j : (M, Bd M) (M, M Q) denote the pair inclusion.
Lemma 5.4.2 PD(Q) = H j (U (M, Q)).
Proof We first reduce to the case where Q is connected. Indeed, as Q is the finite
union of components Q i , with tubular neighbourhood Wi , then

 
H r (M, M Q)
H r (W, Bd W ) H r (Wi , Bd Wi ) H r (M, M Q i )
5.4 Poincar Duality and Submanifolds 223

 
and U (M, Q) = U (M, Q i ). On the other hand, PD(Q) = PD(Q i ). Thus, we
shall assume that Q is connected.
Let us consider the case M = D() and Q is the image of the zero section. As
Hnr (Q) = Z2 = Hn (D(), S()), the Thom isomorphism of Theorem 4.7.29 says
that U ()  [D()] = [Q]. This proves the lemma for any tubular neighbourhood
of Q, for instance W or a smaller tube W contained in the interior of W .
Let us choose a triangulation of M for which W and W are subcomplexes. The
class H j (U (M, Q)) H r (M) may then be represented by a simplicial cocycle q
Sr (W ). The n-simplexes of M involved in the computation of H j (U (M, Q)) 
[M] are then all simplexes of W . Therefore
 
H j (U (M, Q))  [M] = H i U (W, Q)  [W ] = H i([Q]) .

.

Remark 5.4.3 We see in the proof of Lemma 5.4.2 that the Poincar dual PD(Q) of
a submanifold Q M is supported in an arbitrary small tubular neighbourhood of
Q. This localization principle is illustrated in Fig. 5.2 for Q a circle in the torus or
the Klein bottle. For the analogous localization principle in de Rham cohomology,
see [19, Proposition 6.25].
Lemma 5.4.4 The image of PD(Q) under the homomorphism H r (M, Bd M)
H r (M) H r (Q) is equal to the Euler class of the normal bundle = (M, Q).
Proof We use the notations of the proof of Lemma 5.4.2. Let k : (M, ) (M, M
Q) denote the pair inclusion and let 0 : Q D() be the zero section. The various
inclusions give rise to the commutative diagram

H j
H r (M, M Q) / H r (M, Bd M) /2 H r (M)
LLL
H k LH
LLLi
L&
 
H r (W, BdW ) H r (W ) H r (Q)
8
rrrr
rr
 H j
 rr H 0
H r (D(), S()) / H r (D())

The Euler class e() H r (Q) is characterized by the equation H j (U ()) =

H p(e()), where p : D() Q is the bundle projection (see p. 184). Since
p 0 = id Q , the previous diagram and Lemma 5.4.2 yield

H i(PD(Q)) = H i H k(U (M, Q)) = H 0 H j (U ())

= H 0 H p(e()) = e() . 
 
We now compute ker H i : H (M) H (Q) .
224 5 Poincar Duality

Proposition 5.4.5 Let M be a smooth compact n-manifold and let i : Q  M be

the inclusion of a closed smooth submanifold Q of codimension r . Then

ker H i Ann (PD(Q)) = {x H (M) | x  PD(Q) = 0}

(for a class b, the ideal Ann (b) is called the annihilator of b). The above inclusion
is an equality if and only if H i is surjective.

Proof Let a H (M) and let q = PD(Q). Then

 
H i H i(a)  q) = a  H i([Q]) by (4.5.5)
= a  (q  [M]) by definition of q
= (a  q)  [M] by Proposition 4.5.4

Therefore, if H i(a) = 0, then a Ann (q) and the converse is true if and only if
H i is injective. By Kronecker duality (see Corollary 2.3.11), the latter is equivalent
to H i being surjective. 

Example 5.4.6 Let i be the standard inclusion of Q = RP k in M = RP k+r . By

Proposition 4.3.10, one has a commutative diagram

  k+r +1    k+1 
Z2 [a] a / / Z2 [a] a

 
  H i  
H RP k+r / / H RP k

where a is of degree 1. Then, via the above vertical isomorphisms,

      
ker H i = a k+1 = Ann a r = Ann PD RP k .

The following proposition states the functoriality of the Poincar dual for transver-
sal maps.

Proposition 5.4.7 Let f : M N be a smooth map between smooth closed mani-

folds. Suppose that f is transversal to a closed submanifold Q of N . Then
 
PD f 1 (Q) = H f (PD(Q)) .

Proof Let P = f 1 (Q). We consider the commutative diagram

5.4 Poincar Duality and Submanifolds 225

H f
H (N , N Q) / H (M, M P)

H i H j
 H f

H (N ) / H (M)

where the vertical arrow are induced by the inclusions i : (N , ) (N , N Q)

and j : (M, ) (M, M P). Then,

PD(P) = H j (U (M, P)) by Lemma 5.4.2

= H j H f (U (N , Q)) by transversality and Lemma 4.7.30
= H f H i(U (N , Q))
= H f (PD(Q)) by Lemma 5.4.2. 

5.4.2 The Gysin Homomorphism

Let (M, Q) be a pair of smooth compact manifolds, with Q closed. Let i : Q

M denote the inclusion. Set q = dim Q and m = dim M = q + r . The Gysin
homomorphism Gys : H p (Q) H p+r (M, Bd M) is defined for all p N by the
composed homomorphism

[Q] [M]
H p (Q) / Hq p (Q)
H i
/ Hq p (M) o H p+r (M, Bd M) .

The notation i ! and the terminology umkehr homomorphism are also used in the
literature.
For example, Gys(1) = PD(Q), the Poincar dual of Q. More generally:
 
Lemma 5.4.8 For a H p (M), one has Gys H i(a) = a  PD(Q).

Proof

(a  PD(Q))  [M] = a  (PD(Q))  [M])

= a  H i([Q]) 
= H i  H i(a) [Q] by Proposition 4.5.4
= Gys H i(a)  [M]

As  [M] is an isomorphism, this proves the lemma. 

Example 5.4.9 Let M be the total space of an r -disk bundle : M Q. We see

Q as a submanifold of M via there 0-section i : Q M. Let U H r (M, Bd M)
be the Thom class. Since U  [M] = H i([Q]), we see that U = PD(Q). For
b H p (Q), one has
226 5 Poincar Duality
 
Gys(b)  [M] = Gys H i H (b)  [M] since i = id Q
 
= H (b)  PD(Q)  [M] by Lemma 5.4.8
 
= H (b)  U  [M] since U = PD(Q)

= Thom(b)  [M] .

Since  [M] is an isomorphism, we see in this example that the Gysin homomor-
phism is identified with the Thom isomorphism.

Proposition 5.4.10 Let (M, Q) be a pair of smooth closed manifolds, with Q of

codimension r . Let W be a closed tubular neighbourhood of Q in M. There is a
commutative diagram

Gys
/ H p1 (M Q) / H pr (Q) / H p (M) / H p (M Q) /

=
   
/ H p1 (Bd W ) / H pr (Q) / H p (Q) / H p (Bd W ) /

where the vertical arrows are induced by inclusions. The horizontal lines are exact
sequences and the bottom one is the Gysin sequence of the sphere bundle Bd W Q.
Proof We start with the commutative diagram

/ H p1 (M Q) / H p (M, M Q) / H p (M) / H p (M Q) /

   
/ H p1 (Bd W ) / H p (W, Bd W ) / H p (W ) / H p (Bd W ) /

using the cohomology sequences of the pairs (M, M Q) and (W, Bd W ). To get
the diagram of the proposition, we use the identification

Thom /
H pr (Q)
H p (W, Bd W ) o
H p (M, M Q)

and H (W ) H (Q). Thus the bottom line is the Gysin sequence of Bd W Q.

It remains to identify the homomorphism H pr (Q) H p (M) with the Gysin
homomorphism. This amounts to the commutativity of the diagram.

H pr (Q)
Thom / H p (W, Bd W ) o H p (M, M int W ) / H p (M)

[Q] [W ] [M]
  
Hq p+r (Q) / Hq p+r (W ) / Hq p+r (M)
5.4 Poincar Duality and Submanifolds 227

The commutativity of the left square was observed in Example 5.4.9. That of the
right square may be checked using simplicial (co)homology for a triangulation of M
extending one of W . 
Proposition 5.4.11 Let f : M M be a smooth map between closed manifolds.
Let Q be a closed submanifold of codimension r in M. Suppose that f is transversal
to Q. Then, for all p N, the diagram

Gys
H p (Q) / H p+r (M)

H f H f
 
Gys
H p ( f 1 (Q)) / H p+r (M )

is commutative.
Proof Let Q = f 1 (Q). By transversality, f : Q Q is covered by a morphism
of vector bundle f : (M , Q ) (M, Q). Put a Riemannian metric on (M, Q)
and pull it back on (M , Q ), so that f is an isometry on each fiber. By standard
technique of Riemannian geometry, one can find a tubular neighbourhood W of Q
and a tubular neighbourhood W of Q and modify f by a homotopy relative to
Q so that f (W ) W , f (BdW ) BdW , f (M int W ) M int W and
f : W W coincides with f via the exponential maps of W and W . We thus get
a diagram.

H pr (Q)
Thom / H p (W, Bd W ) o H p (M, M int W ) / H p (M)

H f H f H f H f .
   
Thom / o / H p (M )
H pr (Q )
H p (W , Bd W )
H p (M , M int W )

The left square is commutative by construction and the functoriality of the Thom
isomorphism (coming from Lemma 4.7.30). The other squares are obviously com-
mutative. But, as seen in the proof of Proposition 5.4.10, the compositions from the
left end to the right end of the horizontal lines are the Gysin homomorphisms. 

5.4.3 Intersections of Submanifolds

Consider two closed submanifolds Q i (i = 1, 2) of the compact smooth n-manifold

M, Q i being of codimension ri . We suppose that Q 1 and Q 2 intersect transversally.
Then, Q = Q 1 Q 2 is a closed submanifold of codimension r = r1 + r2 .
Proposition 5.4.12 Under the above hypotheses

PD(Q) = PD(Q 1 )  PD(Q 2 ) .

228 5 Poincar Duality

Proof As (M Q 1 ) (M Q 2 ) = M Q, the cup product provides a bilinear map


H r1 (M, M Q 1 ) H r2 (M, M Q 2 )
H r (M, M Q) .

In virtue of Lemma 5.4.2, it suffices to prove that

U (M, Q 1 )  U (M, Q 2 ) = U (M, Q) . (5.4.4)

We may suppose that Q = for, otherwise, the proposition is trivially true, since
q = 0 and r > n.
If A is a submanifold of B, we denote by (B, A) the normal bundle of A in B.
Choose an embedding : D((B, A))  B parameterizing a tubular neighbour-
hood W (B, A). If V B the notation W (A, B)V means (D((B, A)V )). As Q 1
and Q 2 intersect transversally, one has

(M, Q) = (Q 1 , Q)|Q (Q 2 , Q)|Q .

Let b Q. One may choose convenient tubular neighbourhood parameterizations

so that W (M, Q){b} Q j = W (Q j , Q){b} . Let W1 = W (Q 1 , Q){b} Dr2 , W2 =
W (Q 2 , Q){b} Dr1 and W = W (M, Q){b} W1 W2 Dr . Let i : W Wi
be the projection. By Lemma 5.4.2:
the class U (M, Q 1 ) H r1 (M, M Q 1 ) restricts to the non-zero element a1
H r1 (W, W1 Bd W2 ) = Z2 ;
the class U (M, Q 2 ) H r2 (M, M Q 2 ) restricts to the non-zero element a2
H r2 (W, Bd W1 W2 ) = Z2 .
Hence, U (M, Q 1 )  U (M, Q 2 ) restricts to a1  a2 H r (W, Bd W ) = Z2 .
We have to prove that a1  a2 = 0. Let 0 = a i H ri (W j , Bd W j ) (i = j). Then

a1  a2 = H 1 (a 1 )  H 2 (a 2 ) = a 1 a 2 .

By the relative Knneth theorem 4.6.10, a 1 a 2 = 0 in H r (W, Bd W ). Hence,

U (M, Q 1 )  U (M, Q 2 ) restricts to the non-zero element of H r (W (M, Q){b} ,
Bd W (M, Q){b} ) for all b Q. By Lemma 5.4.2, this proves (5.4.4). 
An interesting case is when dim Q 1 + dim Q 2 = dim M. If Q 1 and Q 2 intersect
transversally, then Q 1 Q 2 is a finite collection of points. Let ji : Q i M denote
the inclusion. The following result says that the parity of this number of points
depends only on [Q i ] M = H ji ([Q i ]) and justifies the terminology of intersection
form.
Corollary 5.4.13 Let Q i (i = 1, 2) be two closed submanifolds of the compact
smooth n-manifold M, with dim Q 1 + dim Q 2 = n. Let qi = PD(Q i ). Suppose that
Q 1 and Q 2 intersect transversally. Then

 (Q 1 Q 2 ) q1  q2 , [M] mod 2 .

5.4 Poincar Duality and Submanifolds 229

In other words,

 (Q 1 Q 2 ) H j1 ([Q 1 )] H j2 ([Q 2 )) mod 2 ,

where denotes the (absolute) intersection form (see Sect. 5.3.3).

Proof One has

q1  q2 , [M] = 1, (q1  q2 )  [M]

= 1, [Q 1 Q 2 ]  (Q 1 Q 2 ) mod 2 .

Lemma 5.4.14 Let = ( p : E N ) be a smooth vector bundle over a closed

smooth manifold N . Let , : N E be two smooth sections  of which are

transversal. Let Q be the submanifold of N defined by Q = 1 (N ) (N ) .
Then, the Poincar dual of Q in N is the Euler class e() of .

Proof We us the following notation: if : Y X is a continuous map and Y a

closed manifold, we write [Y ] X = H ([Y ]) Hdim Y (X ); the map is usually
implicit, being an inclusion or an embedding obvious from the context.
Endow with an Euclidean structure and consider the pair (D, S) = (D(), S())
of the associated unit disk and sphere bundle. Using a homotopy in each fiber, we
can assume that (N ) and (N ) are contained in the interior of D. All the sections
of a bundle are homotopic. By Lemma 5.4.2 and its proof, the Thom class U of is
the Poincar dual in D of [N ] D = H ([N ]) = H ([N ]). By Proposition 5.4.12,
U  U is the Poincar dual in D of [Q] D . Let j : (D, ) (D, S) denote the pair
inclusion. As p = id N , one has

[Q] N = H p([Q] D )
 
= H p (U  U )  [D] cap product of (4.5.9)
= H p U  (U  [D]) by Formula (4.5.10)
= H p U  [N ] D 
= H p  H j (U )  [N ] D  by definition of the cap product (4.5.7)
= H p H p(e())  [N ] D by definition of the Euler class
= e()  H p([N ] D )
= e()  [N ] .

which proves the lemma. 

When, in Lemma 5.4.14, the rank of is equal to the dimension of the manifold N ,
then (N ) (N ) is a finite collection of point and one gets the following corollary.

Corollary 5.4.15 Let = ( p : E N ) be a smooth vector bundle of rank n over a

closed smooth n-manifold N . Let , : N E be two smooth sections of which
are transversal. Then
230 5 Poincar Duality
 
 (N ) (N ) e(), [N ] mod 2 .

The following corollary is a justification for the name Euler class.

Corollary 5.4.16 Let N be a smooth closed manifold. Then the following congru-
ences mod 2 hold:

e(T N ), [N ] (N ) dim H (N ) mod 2 .


Proof As (N ) = i (1) dim Hi (N ), the second congruence is obvious. Let
i

0 : N D(T N ) be the zero section and let : N D(T N ) be another smooth

section (i.e. a vector field on N ) which is transversal to 0 . By Corollary 5.4.15, the
number of zeros of is congruent mod 2 to e(T N ), [N ]. It then suffices to find
some vector field transversal to 0 for which we know that its number of zeros is
congruent mod 2 to (N ). Observe that, for a finite CW-complex X , the following
congruence mod 2 holds

(X ) dim H (X )  (X ) mod 2 ,

where (X ) is set of cells of X . For the required vector field, one can take thegradient
vector field = grad f of a Morse function f : N R. Then 01 (N )
0 (N ) = Crit f , the set of critical points of f . The transversality of with 0
is equivalent to f being a Morse function (see [95, Chap. 6]). By Morse theory,
N has then the homotopy type of a CW-complex X with  (X ) =  Crit f , [95,
Chap. 6, Theorem 4.1]. One can also use the classical vector field associated to a
C 1 -triangulation of N , with one zero at the barycenter of each simplex (see, e.g.
[180, pp. 611612]). 

We give below a second proof of Corollary 5.4.16, using the following lemma.

Lemma 5.4.17 Let  N be the diagonal submanifold of M = N N , with normal

bundle (M,  N ). Then, there is a canonical isomorphism of vector bundles

(M,  N ) T N .

Proof Let p1 , p2 : N N N be the projections onto the first and second factor.
For x N , consider the commutative diagram in the category of real vector spaces

0 / T(x,x) ( N ) / T(x,x) (N N ) / (x,x) (M,  N ) / 0

  
/ (Tx N ) / Tx N Tx N / Tx N / 0
0
5.4 Poincar Duality and Submanifolds 231

where the rows are exact and (v) = (T p1 (v), T p2 (v)). The map is an isomor-
phism and sends T(x,x) ( N ) onto (Tx N ). Hence, descends to the isomorphism

: (M,  N )
Tx N which, of course, depends continuously on x. 

Second proof of Corollary 5.4.16 We consider N as the diagonal submanifold of

M = N N , with the diagonal inclusion  : N M. The normal bundle (M, N )
is isomorphic to the tangent bundle of M by Lemma  5.4.17.
By (5.4.1), the Poincar dual of N is equal to i ai bi , where A = {a1 , a2 , . . . }
and B = {b1 , b2 , . . . } are bases of H (N ) dual one to the other for the Poincar
duality.

e(T M), [N ] = e((M,

N )), [N ]
= H 
 i ai bi , [N ] by Lemma 5.4.4
=  i ai  bi , [N ] dim H (N ) mod 2 .

Example 5.4.18 As (S n ) 0 mod 2, the Euler class of T S n vanishes by

Corollary 5.4.16. Let T 1 S n be the associated sphere bundle. By Proposition 4.7.35
and the Leray-Hirsch theorem, we get an isomorphism of H (S n )-module
     
H T 1 S n H S n H S n1 .

If n 3, Poincar duality implies that this isomorphism is a ring-isomorphism. This

is not true if n = 2 (see Remark 4.7.34).
Thus, for n 3, T 1 S n has the same cohomology ring as S n S n1 . However,
by [109, Theorem 1.12], these two spaces have the same homotopy type if and only
if there exists a map f : S 2n+1 S n+1 with Hopf invariant one (see Sect. 6.3). By
Theorem 8.6.7, such an f exists if and only if n = 1, 3, 7.

5.4.4 The Linking Number

Let Q and Q be two disjoint closed submanifold of a closed manifold

(say, in the
smooth category), with q = dim Q, q = dim Q and s = dim
. We assume that
(1) q + q = s 1.
(2)
is a Z2 -homology sphere, i.e. H (
) H (S s ).
(3) 1 Q , [Q] = 1 Q , [Q ] = 0. This condition is always satisfied when q and q
are not zero. If, say, q = 0, it means that Q has an even number of points, so
that [Q] H 0 (Q) = ker1 Q , .
Thanks to (2), Alexander duality (see Theorem 5.3.14) provides an isomorphism

A : H q (Q)
H sq1 (
Q) .
232 5 Poincar Duality

Note that s q 1 = q . By Condition (3), [Q] H q (Q) and [Q ] H q (Q ), so

we can define the linking number (sometimes called the linking coefficient) l(Q, Q )
of Q and Q in
by

l(Q, Q ) = A[Q], H i([Q ]) Z2 , (5.4.5)

where i : Q
denotes the inclusion. Although the asymmetry of the definition,
the equality l(Q, Q ) = l(Q , Q) will be proven in Proposition 5.4.25.
The linking number l(Q, Q ) was introduced in 1911 by Lebesgue [131, pp. 173
175], with a definition in the spirit of Proposition 5.4.22. Lebesgue called Q and Q
enlaces if l(Q, Q ) = 1. One year later, Brouwer [22, pp. 511520] refined
the idea when
, Q and Q are oriented, defining an integral linking number
whose reduction mod 2 is l(Q, Q ) (for the philosophy of Brouwers definition,
see Remark 2.5.10). More history and references about the linking numbers may be
found in [40, pp. 176179 and 185].
As Q is a submanifold of
, the isomorphism A may be described in the following
way, which will be useful for computations. Let V be a closed tubular neighbourhood
of Q in
Q and let X =
intV . The cohomology sequence of (
, X )

/ H sq1 (X ) / H sq (
, X ) / H sq (
)
H sq1 (
)

shows that the connecting homomorphism descends to an injection of

 
coker H sq1 (
) H sq1 (X ) H sq1 (X )

into H sq (
, X ). Let j : (V, Bd V ) (
, X ) denote the pair inclusion. Identifying
H (V ) with H (Q), one has the diagram

H sq1 (X ) o H q (Q)
A


 
H j [V ]
H sq (
, X ) / H sq (V, BdV ) / Hq (Q) (5.4.6)

 
[
]
H sq (
) / Hq (
)

whose columns are exact (the right hand one by Lemma 3.1.10) and whose bottom
part is commutative by Proposition 5.3.6. Then A is the unique isomorphism making
the top rectangle commutative (compare the proof of Theorem 5.3.14).
Remark 5.4.19 Diagram (5.4.6) uses the singular (co)homology. But, via a trian-
gulation of
, the various spaces may be the geometric realizations of simplicial
complexes (by abuse of notations we use the same letters, i.e.
= |
|, etc.).
5.4 Poincar Duality and Submanifolds 233

Then, Diagram (5.4.6) makes sense for simplicial (co)homology; the isomorphism
H sq (
, X ) H sq (V, BdV ) is just the simplicial excision (see Exercise 2.17).
Remark 5.4.20 Suppose that Q is connected or consists of two points. Then H q (

Q) H q (Q) Z2 . Therefore, l(Q, Q ) = 1 if and only if H i([Q ]) = 0 in
Hq (
Q). In this case, l(Q, Q ) determines H i([Q ]).
The following lemma shows that l(Q, Q ) is not always zero. We say that Q is
a meridian sphere for Q if Q is the boundary of a (s q)-disk  in
intersecting
Q transversally in one point.
Lemma 5.4.21 Let Q, Q and
satisfying (1)(3) above. Suppose that Q is a
meridian sphere for Q. Then l(Q, Q ) = 1.
Proof Let (Q,
) be the normal bundle of Q in
. A Riemannian
 metric provides a
smooth bundle pair (D(), S()) with fiber D mq , S mq1 and a diffeomorphism

: (D(), S()) (V, BdV ), where V is a tubular neighbourhood of Q in
.
By choosing the Riemannian metric conveniently, we may assume that  V is the
image by of a fiber D mq of D(). One has H i([Q ]) = H j H ([S mq1 ]),
where j : (V, BdV ) (
,
Q) denotes the pair inclusion, so l(Q, Q ) =
l(Q, (S mq1 )) by (5.4.5).
In Diagram (5.4.6), one has H H j A([Q]) = U , the Thom class of ,
as can be checked on each connected component of Q. Therefore,

l(Q, S) = A([Q]), H i([Q ])

= A([Q]), H j H ([S mq1 ])
= A([Q]), H j H ([D mq ])
=  H H j A([Q]), [D mq ]
= U, [D mq ]

and U, [D mq ] = 1 by Lemma 4.7.28. 

The following proposition gives a common way to compute a linking number,
related to the original definition of Lebesgue [131, pp. 173175]. Let Q, Q and

satisfying (1)(3) above. Suppose that there exists a compact manifold W with
Bd W = Q so that the inclusion of Q into
extends to a map j : W
which
is transverse to Q ( j needs not to be an embedding). Then j 1 (Q) is a finite number
of points in W .
 
Proposition 5.4.22 l(Q, Q ) =  j 1 (Q) mod 2 .
 
Proof Let k =  j 1 (Q) N. Let W0 be the manifold W minus an open tubular
neighbourhood of j 1 (Q). By (5.4.5), l(Q, Q ) depends only on the homology class
H i([Q ]) Hq (
Q) which, thanks to the map j (see Exercise 5.6), is the same
as that of k meridian spheres. The result then follows from Lemma 5.4.21. 
234 5 Poincar Duality

We now introduce some material for Lemma 5.4.23, which will enable us to
compute linking numbers using convenient singular cochains. Let Q, Q and

satisfying (1)(3) above. Let W be a closed tubular neighbourhood of Q in

Q
and let V be a closed tubular neighbourhood of Q in int W . We also consider the
symmetric data Q V W
Q, assuming that W W = . Let
B = {W, W ,
(V V )}; note that the small simplex theorem 3.1.34 holds true
for B.
Let c Z sq (W, W intV ) be a singular cocycle representing the Poincar dual
class of Q in H sq (W, W intV ) H sq (V, BdV ). We can see c as a cocycle of
sq
W and take its zero extension c CB (
), i.e.

c,
 if Ssq (W )
c,
 =
0 otherwise.

Since c vanishes on Ssq (W, W intV ), the cochain c is a B-small cocycle, i.e.
sq sq1
c Z B (
). We claim that we can choose a CB (
) such that a = c.
sq
Indeed, by (2) and the small simplex theorem 3.1.34, 0 = H sq (
) HB (
)
when q > 0. When q = 0, c represents the Poincar dual class of Q in HBs (
)
H s (
). But, by (3), [Q] represents 0 in H0 (
), so c represents 0 in HBs (
). A
sq 1
cochain a CB (
) with (a ) = c may be also chosen, using the symmetric

data Q V W
Q.
Finally, let Z q (
int V ) represent H i([Q ]) and let Z sB (
) represent
[
] in HsB (
) Hs (
).

Lemma 5.4.23 The following equalities hold true.

(a) l(Q, Q ) = a, .
(b) l(Q, Q ) = a  c , .

Proof We first establish some preliminary steps.

Step 1: If a1 , a2 C sq1 (
) satisfy (a1 ) = (a2 ), then a1 ,  = a2 , .
Indeed, one has then (a1 + a2 ) = 0. If s q 1 = q > 0, Condition (2) implies
that there exists b C sq2 (
) such that b = a1 + a2 . Hence,

a1 ,  + a2 ,  = a1 + a2 ,  = b,  = b,  = 0 .

If q = 0, then (a1 + a2 ) = 0 implies that a1 + a2 = 1 by Proposition 3.1.8, thus

a1 + a2 ,  = 0 by Condition (3).
Step 2: Let ci Z sq (W, W intV ) (i = 1, 2) be singular cocycles both repre-
sq1
senting the Poincar dual class of Q in H sq (W, W intV ). Let ai CB (
)
such that ai = ci as above. Then a1 ,  = a2 , . Indeed, there exists b
C sq1 (W, W intV ) such that (b) = c1 + c2 . Its zero extension b CB
sq1
(
)
= c1 + c2 and then (a1 + b)
then satisfies (b) = c2 . Thus
5.4 Poincar Duality and Submanifolds 235

a2 ,  = a1 + b,  by Step 1

= a1 ,   = 0 .
since b,

We can now start the proof of Lemma 5.4.23. Given Steps 1 and 2, it is enough
to prove (a) for a particular choice of c and a. We use Diagram (5.4.6) and see A as
an isomorphism from H q (Q) onto H sq1 (
intV ). Let a Z sq1 (
int V )
sq1
represent A([Q]). Let a CB (
) be its zero extension and let c = (a)
sq
Z B (
). By Lemma 3.1.17, c represents (A([Q]). Also, c is the zero extension of
the cocycle c Z sq (W, W int V ) which, by definition of A and Diagram (5.4.6),
represents the Poincar dual class of Q in H sq (W, W int V ). Therefore, since a
represents A([Q]) and Z q (
int V ) represents H i([Q ]), one has l(Q, Q ) =
a,  = a, .
To prove (b), consider the pair inclusions j1 : (
, ) (
,
int V ) and
j2 : (W , W int V ) (
,
int V ). Since Z sB (
), there exists a (unique)
Z s (W , W int V ) such that C j2 ( ) = C j1 (). As H j1 and H j2 are
isomorphisms, represents the generator of Hs (W , W int V ) = Z2 . Therefore,
c  = c  represents H i([Q ]) and, by (a),

l(Q, Q ) = a, c   = a  c ,  .

Remark 5.4.24 The proof of Lemma 5.4.23 in the simplicial category (see
Remark 5.4.19) is somewhat simpler. It uses only the tubular neighbourhoods Vi
and not the Wi s, and, of course, does not require the use of small simplex tech-
niques. Also, may be taken explicitly as Ss (
). Writing the details is left to the
reader as an exercise.

Lemma 5.4.23 will be used for the Hopf invariant (see Sect. 6.3.3). For the moment,
its main consequence is the following proposition.

Proposition 5.4.25 Let Q, Q and

satisfying (1)(3) above. Then

l(Q, Q ) = l(Q , Q) .

Proof By Point (b) of Lemma 5.4.23, one has l(Q, Q ) = a  c ,  and

l(Q , Q) = a  c,
. Then

l(Q, Q ) + l(Q , Q) = a  c + a  c,

= (a  a ),  = a  a , () = 0 . 
236 5 Poincar Duality

5.5 Exercises for Chapter 5

5.1. Prove that the product of two homology manifolds is a homology manifold.
5.2. Let M be a compact manifold with boundary such that H (M) = 0. Show that
the boundary of M is a homology sphere.
5.3. Check the Poincar duality (Theorem 5.3.7) for the manifolds S 1 I and the
Mbius band.
5.4. Show that there is no continuous retraction of a non-empty compact manifold
onto its boundary.
5.5. Let M be a closed triangulable manifold of dimension n. Prove that the homo-
morphism Hk (M { pt}) Hk (M) induced by the inclusion is an isomor-
phism for k < n.
5.6. Let M be a compact triangulable topological n-manifold with boundary
Bd M = N . Suppose that is N = N1 N 2 the union of two closed (n 1)-
manifolds. Let f : M X be a continuous map. Show that H f ([N1 ]) =
H f ([N2 ]) in Hn1 (X ). What happens if N2 = ?
5.7. Let f : M N be a map between closed n-dimensional manifolds of the
same dimension. Show that the degree of f may be computed locally, using a
topological regular value, like in Proposition 3.2.6.
5.8. Let
m be the orientable surface of genus m and let
n be the nonorientable
surface of genus n. For which m and n does there exist a continuous map of
degree one from
m to
n or from

n to
m ?
5.9. Let M be a closed manifold of dimension m which is the products of two
closed manifolds of positive dimensions. Does there exist a degree one map
f : M RP m ?
5.10. Let Q 1 and Q 2 be closed submanifolds of a closed manifold M (in the smooth
category). Suppose that Q 1 and Q 2 intersect transversally in an odd number
of points. Show that [Q 1 ] and [Q 2 ] represent non-zero classes in H (M).
5.11. Let i : Q M be the inclusion of a smooth closed submanifold of dimension
q in a smooth closed manifold M. Suppose that H i([Q]) = 0 in Hq (M).
Prove that the Euler class of the normal bundle to Q vanishes.
5.12. For A {0, 1, . . . , n}, let PA = {[x0 : : xn ] RP n | xi = 0 when i / A}.
What is the diffeomorphism type of PA ? Show that, if A B = {0, 1, . . . , n},
then PA and PB intersect transversally (what is the intersection?). How does
Proposition 5.4.12 apply in this example?
5.13. Poincar dual classes in a product. Let M1 and M2 be smooth compact
manifolds. Let Q i be a closed submanifold of Mi (i = 1, 2). Then Q 1 Q 2
is a closed submanifold of M1 M2 . Prove that

PD(Q 1 Q 2 ) = PD(Q 1 ) PD(Q 2 ) (5.5.1)

in H (M1 M2 ).
5.5 Exercises for Chapter 5 237

5.14. Poincar dual classes in a product II. Let M and M be smooth closed mani-
folds and let x M and x M . What are PD({x} M ) and PD(M {x }) in
H (M M )? Check that PD({x} M )  PD(M {x }) = PD({(x, x )}).
5.15. Let Q and Q be disjoint submanifolds of S 2 , where Q consists of two circles
and Q of four points. Using Proposition 5.4.22, compute the linking numbers
l(Q, Q ) and l(Q , Q) for the various possibilities.
5.16. Brouwers definition of the linking number. Let Q and Q be two disjoint closed
submanifolds of S n , of dimension respectively q and q satisfying p + q =
n 1. If Q (or Q ) is of dimension 0, it should consist of an even number of
points. See Q and Q as submanifolds of Rn via a stereographic projection of
S n { pt} onto Rn . Consider the Gauss map : P Q S n1 given by
xy
(x, y) = ||xy|| . Show that the degree of is equal to the linking number
l(Q, Q ) (see Sect. 5.4.4). [Hint: use Proposition 5.4.22.]
5.17. Write the proof of Lemma 5.4.23 in the simplicial category (see Remarks
5.4.19 and 5.4.24).

5.18. Let
be the unit sphere in Rq+1 Rq +1 . Let Q = S q {0}
and

Q = {0} S q
. Compute the linking number l(Q, Q ) in
.
Chapter 6
Projective Spaces

Coming from algebraic geometry, projective spaces and their Hopf bundles play an
important role in homotopy theory, as already seen in Sects. 3.8 and 4.3. The precise
knowledge of their cohomology algebra has interesting applications, like the Borsuk-
Ulam theorem, continuous multiplications in Rm and the Hopf invariant, which are
presented in Sect. 6.2.

6.1 The Cohomology Ring of Projective SpacesHopf Bundles

The cohomology ring of RPn for n was established in Proposition 4.3.10, using
the transfer (or Gysin) exact sequence for the double cover (S 0 -bundle) S n RPn . It
gives a GrA-isomorphism Z2 [a]/(an+1 ) H (RPn ). We give below a completely
different proof of this fact, which is based on Poincar duality (as RPn is a smooth
closed manifold, it can be triangulated as a polyhedral homology n-manifold: see p.
203). We shall also discuss the cases of complex and quaternionic projective spaces
CPn and HPn , and of the octonionic projective plane OP2 .

Proposition 6.1.1 The cohomology algebra of RPn (n ) is given by

H (RPn ) Z2 [a]/(an+1 ), H (RP ) Z2 [a] ,

with a H 1 (RPn ).

Proof We prove the first statement by induction on n. It is true for n = 1 since

RP1 = S 1 (and also proven for n = 2 at p. 137). Suppose, by induction, that it is
true in for RPn1 .
In Example 3.4.5 is given the standard CW-structure of RPn , with one k-cell for
each k = 0, 1, . . . , n. It follows that H k (RPn , RPn1 ) = 0 for k n 1 and
H n (RPn , RPn1 ) = Z2 . By Poincar duality, H n (RPn ) = Z2 , so the exact sequence

Springer International Publishing Switzerland 2014 239

J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_6
240 6 Projective Spaces

for the pair (RPn , RPn1 ) gives

0 H n1 (RPn ) H n1 (RPn1 ) H n (RPn , RPn1 ) H n (RPn ) 0 .

        
Z2 Z2 Z2

All that implies that the inclusion induces an isomorphism H k (RPn ) H k (RPn1 )
for k n 1. By the induction hypothesis and functoriality of the cup product,
H k (RPn ) = Z2 for k n 1, generated by ak .
By Poincar duality, Hn (RPn ) = Z2 with generator [RPn ] and, by
Theorem 5.3.12, the bilinear map

[RPn ]
H 1 (RPn ) H n1 (RPn ) H n (RPn ) Z2

is not degenerated. Therefore, a  an1 = 0, which proves that H (RPn )

Z2 [a]/(an+1 ).
Finally, by the standard CW-structure of RP , one has, for all integer n, that
H (RP , RPn ) = 0 for k < n. The first statement then implies the second.
k

Note that we have also proven that the standard CW-structure of RPn (n ) is
perfect. 

Corollary 6.1.2 The Poincar series of RPn (n ) are

1 t n+1 1
Pt (RPn ) = 1 + t + + t n = and Pt (RP ) = .
1t 1t

Remark 6.1.3 Proposition 6.1.1 and its proof show that the GrA-homomorphism
H j (RPn+k ) H j (RPn ) induced by the inclusion RPn  RPn+k is surjective
(k ). In particular, it is an isomorphism for j n.

Remark 6.1.4 The polynomial structure on H (RP ) implies the following fact: if
f : RP X is a continuous map with X a finite dimensional CW-complex, then
H f = 0. Actually, f is homotopic to a constant map. This result is a weak version
of the original Sullivan conjecture [188, p. 180], which lead to important researches
in homotopy theory (see, e.g. [171]) and was finally proven, in a more general form,
by Miller [143].

We now pass to the complex projective space CPn , the space of complex lines in
Cn+1 . Such a line is represented by a non-zero vector z= (z0 , . . . , zn ) Cn+1 {0},
and two such vectors z and z are in the same line if and only if z = z with

C = C {0}. If |z| = |z | = 1, then S 1 . Thus

 
CPn = (Cn+1 {0}) C = S 2n+1 S 1 .
6.1 The Cohomology Ring of Projective SpacesHopf Bundles 241

The image of (z0 , . . . , zn ) in CPn is denoted by [z0 : z1 : . . . : zn ]. As S 1 acts smoothly

on S 2n+1 , the quotient CPn is a closed smooth manifold and the quotient map

p : S 2n+1 CPn

is a principal S 1 -bundle [82, Example 4.44], called the Hopf bundle. In this simple
example, this can be proved directly. Consider the open set Vk Cn+1 {0} given
by Vk = {(z0 , . . . , zn ) Cn+1 | zk = 0}. Its image in CPn is an open set Uk , domain

of the chart k : Cn Uk given by

k (z0 , . . . , zn1 ) = [z0 : z1 : . . . : zk1 : 1 : zk : . . . : zn1 ] . (6.1.1)

p1 (Uk ) is given by
On the other hand, a trivialization k : Uk S 1

1
k (k (z0 , . . . , zn1 ), g) =  n1 (z0 , z1 , . . . , zk1 , g, zk , . . . , zn1 ) .
1+ i=0 |zi |2
(6.1.2)

It is also classical that CPn is obtained from CPn1 by attaching one cell of
dimension 2n,

CPn = CPn1 p D2n ,

with the attaching map p : S 2n1 CPn1 being the quotient map (see e.g. [82,
Example 0.6] or [155, Theorem 40.2]). This gives a standard CW-structure on CPn
with one cell in each even dimension 2n. For the direct limit CP , we get CW-
structure with one cell in each even dimension. For these CW-structure, the vector
space of cellular chains vanish in odd degree, so the cellular boundary is identically
zero. Therefore,

1 t 2(n+1) 1
Pt (CPn ) = 1 + t 2 + + t 2n = and Pt (CP ) = .
1t 2 1 t2

As CPn is a smooth manifold, the same proof as for Proposition 6.1.1, using
Poincar duality, gives Proposition 6.1.5. One can also adapt the proof of
Proposition 4.3.10, using the Gysin exact sequence of the Hopf bundle (see
Exercise 6.2):

e() Hp
H k1 (S 2n+1 ) H k2 (CPn ) H k (CPn ) H k1 (S 2n+1 ) .

Proposition 6.1.5 The cohomology algebra of CPn (n ) is given by

H (CPn ) Z2 [a]/(an+1 ), H (CP ) Z2 [a],

242 6 Projective Spaces

with a H 2 (CPn ). The class a is the Euler class of the Hopf bundle
S 2n+1 CPn .

If we replace the field of complex numbers by that of quaternions H, we get quater-

nionic projective space HPn :
 
HPn = (Hn+1 {0}) H = S 4n+3 S 3

(it is usual to take the right H-vector space structure on Hn+1 ). The space HPn is
obtained from HPn1 by attaching one cell of dimension 4n, with the attaching map
p : S 4n1 HPn1 being the quotient map. The map p is an S 3 -bundle called
the Hopf bundle. This gives a standard CW-structure on HPn with one cell in each
dimension 4k 4n. For the direct limit HP , we get CW-structure with one cell in
each dimension 4k and

1 t (4n+1) 1
Pt (HPn ) = 1 + t 4 + + t 4n = and Pt (HP ) = .
1 t4 1 t4

Proposition 6.1.6 is proven as Proposition 6.1.5, either using Poincar duality

(HPn is a smooth 4n-manifold), or the Gysin exact sequence of the Hopf bundle (see
Exercise 6.2).

Proposition 6.1.6 The cohomology algebra of HPn (n ) is given by

H (HPn ) Z2 [a]/(an+1 ), H (HP ) Z2 [a],

with a H 4 (HPn ). The class a is the Euler class of the Hopf bundle
S 4n+3 HPn .

Let K = R, C or H and let d = d(K) = dimR K. The space KP1 has a CW-
structure with one 0-cell and one d-cell and is then homeomorphic to S d . The quotient
maps S 2d1  KP1 thus give maps

h1,1 : S 1  S 1 , h3,2 : S 3  S 2 and h7,4 : S 7  S 4

called the Hopf maps. Note that h1,1 is just a 2-covering. Using the homeomorphism
Sd K = K {} given by a stereographic projection, these Hopf maps admit the
formula

vw1 if w = 0
hi,j (v, w) = (6.1.3)
otherwise.

This formula also makes sense for K = O, the octonions, whose multipli-
cation admits inverses for non zero elements. This gives one more Hopf map
h15,8 : S 15 S 8 . One can prove that h15,8 is an S 7 -bundle (see [82, Example 4.47]),
also called the Hopf bundle. Attaching a 16-cell to S 8 using h15,8 produces the
6.1 The Cohomology Ring of Projective SpacesHopf Bundles 243

octonionic projective plane OP2 (because of non-associativity of the octonionic

multiplication, there are no higher dimensional octonionic projective spaces).
Proposition 6.1.7 The cohomology algebra of OP2 is given by

H (OP2 ) Z2 [a]/(a3 )

with a H 8 (OP2 ) = Z2 . In particular, Pt (OP2 ) = 1 + t 8 + t 16 .

Proof By its cellular decomposition, H k (OP2 ) = Z2 for k = 0, 8, 16 and zero
otherwise. Let a H 8 (OP2 ) and b H 8 (S 8 ) be the non-zero elements. The mapping
cylinder E of h15,8 is the disk bundle associated to the Hopf bundle and OP2 has the
homotopy type of E D16 , with E D16 = S 15 . The Thom class of the Hopf bundle
h15,8
U H 8 (E, S 15 ) H 8 (OP2 , int D16 ) H 8 (OP2 ) ,

is not zero, so corresponds to a H 8 (OP2 ). The diagram

U / H 16 (E,

H 8 (E) S 15 )
O O

/ H 16 (OP2 , int D16 ) / H 16 (OP2 )

H 8 (OP2 ) 3
a

is then commutative by the analogue in singular cohomology of Lemma 4.1.7. This

proves that  a : H 8 (OP2 ) H 16 (OP2 ) is bijective. 
Remark 6.1.8 Proposition 6.1.7 may also be proved using Poincar duality, since
OP2 has the homotopy type of a closed smooth 16-manifold, in fact a homogeneous
space of the exceptional Lie group F4 (see [207, Theorem 7.21, p. 707]).
The computations of the cohomology algebra H (KP2 ) have the following con-
sequence.
Corollary 6.1.9 The Hopf maps

h1,1 : S 1 S 1 , h3,2 : S 3 S 2 , h7,4 : S 7 S 4 and h15,8 : S 15 S 8

are not homotopic to constant maps.

Using the Steenrod squares, we shall prove in Chap. 8 that no suspension of these
Hopf maps is homotopic to a constant map (see Proposition 8.6.1).
Proof One has S 1 h1,1 D2 RP2 . If h1,1 were null-homotopic, RP2 would have
the homotopy type of S 1 S 2 (see [82, Proposition 0.18]). But, in H (S 1 S 2 ), the
cup-square map vanishes by (4.2.2), which is not the case in H (RP2 ). The same
proof works for the other Hopf maps. 
244 6 Projective Spaces

We now consider the vector bundle K over KPn associated to the Hopf bundle,
where K = R, C, H and n . For n N, total space of K is

E(K ) = {(a, v) KPn Kn+1 | v a} ,

with bundle projection (a, v)  a. For instance, for n = 1, E(R ) is the Mbius band.
The correspondence (a, v) V defines a map E(K ) Kn+1 whose restriction to
each fiber is linear and injective. This endows K with an Euclidean structure, whose
unit sphere bundle is the Hopf bundle. The vector bundle K is called the Hopf vector
bundle or the tautological bundle over KPn . Its (real) rank is d = d(K) = dimR K.
By passing to the direct limit when n , we get a tautological bundle K over
KP . Propositions 6.1.5 and 6.1.6 (and 4.7.36 for K = R) gives the following result.

Proposition 6.1.10 For n , the Euler class e(K ) is the non-zero element ad
H d (KPn ) = Z2 .

The inclusions R C H induce inclusions

j1 j2
RPn
CPn
HPn, n .

The above proposition permits us to determine the GrA-homomorphism induced in

cohomology by these inclusions.

Proposition 6.1.11 H jd (a2d ) = ad2 .

Proof Observe that C is a complex vector bundle, so the multiplication by i C is

defined on each fiber. We notice that

i1 C = R i R R R . (6.1.4)

Then
H j1 (a2 ) = H j1 (e(C )) by Proposition 6.1.11
= e(j1 C ) by Lemma 4.7.31
= e(R R ) by (6.1.4)
= e(R )  e(R ) by Proposition 4.7.40
= a12 by Proposition 6.1.11.

The proof that H j2 (a4 ) = a22 is the same, using the multiplication by j H on the
fiber of H which is a quaternionic vector bundle. 

The Hopf bundles are sphere bundles over S p such that the total space is also a
sphere. We shall see in Proposition 6.3.5 that p = 1, 2, 4, 8 are the only dimensions
where such examples may occur.
6.2 Applications 245

6.2 Applications

6.2.1 The Borsuk-Ulam Theorem

A (continuous) map f : Rm Rn or f : S m S n such that f (x) = f (x) is

called an odd map.
Theorem 6.2.1 Let f : S m S n be an odd map. Then:
(1) n m.
(2) if m = n, then deg f = 1.
Proof If f is odd, it descends to a map f : RPm RPn with a commutative diagram

f
Sm / Sn
pm pn .
 f

RPm / RPn

The two-fold covering pn is induced from p : S RP by the inclusion

RPn  RP . By Lemma 4.3.6 and Proposition 4.3.10, the characteristic classes
w(pm ) H 1 (RPm ) and w(pn ) H 1 (RPn ) are the generators of these cohomology
groups and H f (w(pn )) = w(pm ). By Proposition 4.3.10 again, one has that 0 =
H f (w(pn )n+1 ) = w(pm )n+1 which implies that n m.
If m = n, observe that H pn : H n (RPn ) H n (S n ) is the zero homomor-
phism since pn is of local degree 2. The transfer exact sequence of (4.3.9), which is
functorial, gives the commutative diagram

tr / H n (RPn )
H n (S n )
Hf H f ,
 
tr / H n (RPn )
H n (S n )

proving that deg f = 1. 

As a corollary, we get the theorem of Borsuk-Ulam.
Corollary 6.2.2 (Borsuk-Ulam theorem) Let g : S n Rn be a continuous map.
Then, there exists z S n such that g(z) = g(z).
Proof Otherwise, the map f : S n S n1 defined by

g(z) g(z)
f (z) =
|g(z) g(z)|

is continuous and odd, which contradicts Theorem 6.2.1. 

246 6 Projective Spaces

A famous consequence is the ham sandwich theorem. For an early history of this
theorem, see [14].
Corollary 6.2.3 Let A1 , . . . , An be n bounded Lebesgue measurable subsets of Rn .
Then, there exists a hyperplane which bisects each Ai .
Proof Identify Rn by an isometry with an affine n-subspace W of Rn+1 not passing
through the origin, and thus see A1 , . . . , An W . For each unit vector v Rn+1 ,
consider the half-space Q(v) = {x Rn+1 | v, x > 0}. Let gi : S n R defined
by gi (v) = measure(Ai Q(v)). The maps gi are the coordinates of a continuous
map g : S n Rn . By Corollary 6.2.2, there is z S n such that g(z) = g(z),
which means that gi (z) = 21 measure(Ai ). Then, P(z) W is the desired bisecting
hyperplane. 

6.2.2 Non-singular and Axial Maps

A continuous map : Rm Rm Rk is called non-singular if

(1) (x, y) = (x, y) for all x, y Rm and all , R, and
(2) (x, y) = 0 implies that x = 0 or y = 0.
Non-singular maps generalize bilinear maps without zero divisors. They were
introduced in [61], from where the results of this section are extracted. For references
about the earlier literature, see also [61].
Non-singular maps are related to axial maps. A continuous map g : RPm
RPm RP , with  m, is called axial if the restriction of g to each slice is not
homotopic to a constant map. By Corollary 3.8.4, this is equivalent to ask that these
restrictions gx1 : {x} RPm RP or gx2 : RPm {x} RP to be homotopic
to the inclusion RPm  RP . Using Corollary 3.8.4, this is equivalent to ask that
H gxi (a ) = am , where aj is the generator of H 1 (RPj ). By Corollary 4.7.4, we deduce
that a continuous map g : RPm RPm RP is axial if and only if

H g(a ) = 1 am + am 1 . (6.2.1)

The name of axial map appeared in [6] where references about the earlier literature
on the subject may be found. It started with the work of Stiefel and Hopf [101].
Let : Rm Rm Rk be a non-singular map. By Point (2) of the definition,
we get a continuous map : S m1 S m1 S k1 defined by

(x, y)
(x,
y) = . (6.2.2)
|(x, y)|

Point (1) above implies that descends to a map

: RPm1 RPm1 RPk1 . (6.2.3)

6.2 Applications 247

For x RPm1 , the restriction of x to the slice {x} RPm1 is covered by two-fold
covering maps:

{x} S m1 / RPm1 S m1 / S k1

pm1 pk1
   .

{x} RPm1 / RPm1 RPm1 /2 RPk1
x

By Lemma 4.3.6 and Proposition 4.3.10, the characteristic classes w(pm1 )

H 1 (RPm1 ) and w(pk1 ) H 1 (RPk1 ) are the generators of these cohomology
groups and H x (w(pk1 )) = w(pm1 ). Hence, x is not homotopic to a constant
map. The same reasoning holds for the slices RPm1 {x}. Therefore, is axial.
Conversely, if g : RPm1 RPm1 RPk1 is an axial map, it induces on universal
covers a map g : S m1 S m1 S k1 satisfying g(x, y) = g(x,
y) = g(x,
y).
The map : Rm Rm Rk defined by


x y
(x, y) = |x| |y| g ,
|x| |y|

is a non-singular map. This proves the following lemma.

Lemma 6.2.4 The correspondence  provides a bijection between non-
singular maps Rm Rm Rk (up to multiplication by non-zero constants) and
axial maps RPm1 RPm1 RPk1 .

Let : Rm Rm Rk be a non-singular map. The restriction of to each slice

is odd. Hence, if a non-singular map : Rm Rm Rk exists, it follows form
Theorem 6.2.1 that k m. When m = k, the following proposition is attributed to
Stiefel. For other proofs (see [153, Theorem 4.7] or Remark 8.6.7).

Proposition 6.2.5 Let : Rm Rm Rm be a non-singular map. Then m = 2r .

In fact, by a famous result of J.F. Adams (see Remark 8.6.7), non-singular maps
Rm Rm Rm exist only if m = 1, 2, 4, 8.

Proof We consider the associated axial map : RPm1 RPm1 RPm1 and
denote by a the generator of H m1 (RPm1 ). The Knneth theorem implies that the
correspondence x  1 a and y  a 1 provides a GrA-isomorphism

H (RPm1 RPm1 ) .
Z2 [x, y] (x m , ym )

By (6.2.1), H (a)
= x + y. Therefore, (x + y)m = 0. As x m and ym also vanish,
one has

m
m 
m1
m
(x + y)m = i x i ymi = i x i ymi
i=0 i=1
248 6 Projective Spaces

This implies that mi 0 mod 2 for all i = 1, . . . , m 1 which, by Lemma 6.2.6,
happens only if m = 2r . 

For n N, denote its dyadic expansion in the form n = jJ(n) 2j where J(n)
N.
Lemma 6.2.6 (Binomial coefficients mod 2) Let m, r N. Then
m 
r 1 mod 2 J(r) J(m) .

In other words, mr 1 mod 2 if and only if the dyadic expansion of r is a
sub-sum of that of m.
j j
Proof In Z2 [x], the equation (1 + x)2 = 1 + x 2 holds, whence (1 + x)2 = 1 + x 2 .
Therefore


m
m  j  j
r x r = (1 + x)m = (1 + x)2 = (1 + x 2 ) .
r=0 jJ(m) jJ(m)

The identification of the coefficient of x r gives the lemma. 

The technique of the proof of Proposition 6.2.5 also gives a result of H. Hopf,
[101, Satz I.e].
Proposition 6.2.7 Let : Rm Rm Rk be a non-singular map. If m > 2r , then
k 2r+1 .
Proof We already know that k > m. As in the proof of Proposition 6.2.5, we consider
the associated axial map : RPm1 RPm1  RPk1 and use the same notations.
We get the equation (x + y)k = 0 in Z2 [x, y] (x m , ym ), which, as m > 2r , implies
 
that ki = 0 for all 1 i 2r . By Lemma 6.2.6, the dyadic expansion k = j kj 2j
must satisfy kj = 0 for j r, which is equivalent to k 2r+1 . 
Remark 6.2.8 There always exists a non-singular map : Rm Rm R2m1 (see
[61, Sect. 5]).
We finish this section by mentioning two results relating non-singular or axial
maps to the immersion problem and the topological complexity of projective spaces.
The following proposition was proven in [6].
Proposition 6.2.9 There exists an axial map g : RPn RPn RPk (k > n) if and
only if there exists an immersion of RPn in Rk .
We shall not talk about the large literature and the many results on the problem
immersing or embedding RPm in Rq (see however Proposition 9.5.23). Tables and
references are available in [36]. The existence of non-singular maps is also related
to the topological complexity of the projective space. The following is proven in
[61, Theorem 6.1].
6.2 Applications 249

Theorem 6.2.10 The topological complexity TC (RPn ) is equal to the smallest inte-
ger k such that there is a non-singular map : Rn+1 Rn+1 Rk .

Symmetric non-singular maps (i.e. (x, y) = (y, x) are, in some range, related to
embeddings of RPn in Euclidean spaces or to the symmetric topological complexity.
For results and references, see [68].

6.3 The Hopf Invariant

6.3.1 Definition

Let f : S 2m1 S m be a continuous map. The space Cf = D2m f S m is a

CW-complex with one cell in dimension 0, m and 2m. Consider the cup-square
map 2m : H m (Cf ) H 2m (Cf ), given by 2m (x) = x  x. The Hopf invariant
Hopf (f ) Z2 is defined by

1 if 2m is surjective for Cf .
Hopf (f ) =
0 otherwise.

The space Cf depends only on the homotopy class of f (see, e.g.[82, Proposition 0.18]),
then so does the Hopf invariant. A constant map has Hopf invariant 0. The compu-
tation of the cohomology ring of the various projective planes in Sect. 6.1 shows
that the 2-fold cover S 1 S 1 as well as the other Hopf maps h3,2 : S 3 S 2 ,
h7,4 : S 7 S 4 and h15,8 : S 15 S 8 have Hopf invariant 1.
Our Hopf invariant is just the mod 2 reduction of the classical integral Hopf
invariant defined in e.g. [82, Sect. 4.B]. The form of our definition is motivated by
extending the statements to the case m = 1, usually not considered by authors.
Note that Hopf defined his invariant in 193135 [99, 100], before the invention
of the cup product. He used linking numbers (see Sect. 6.3.4).

For m = 1, recall the bijection DEG: [S 1 , S 1 ]
Z given in (3.2.3).

Proposition 6.3.1 Let f : S 1 S 1 . Then

0 if DEG (f) 0 mod 4
Hopf (f ) =
1 otherwise.

Proof Let C = Cf . If DEG (f ) is odd, then deg(f ) = 1 by Proposition 3.2.9.

The computation of the cellular cohomology of C using Lemma 3.5.4 shows that
H (C) = 0, so 21 is surjective and Hopf (f ) = 1. If DEG (f ) = 2k, then,
H 1 (C) Z2 H 2 (C). Consider the 2-fold covering p : C C whose
250 6 Projective Spaces

characteristic class is the non zero element a H 1 (C). Its transfer exact sequence
looks like

H p a

0 H 0 (C) H 1 (C) H 1 (C)
tr
H 1 (C) H 2 (C)
           
Z2 Z2 Z2 Z2

By Van Kampens theorem, 1 (C) is cyclic of order 2k. Thus, 1 (C) is cyclic of

order k and H (C) = Z2 if k is even while it vanishes if k is odd. The proposition
1

thus follows from the above transfer exact sequence. 

6.3.2 The Hopf Invariant and Continuous Multiplications

A classical construction associates a map f : S 2m1 S m to a continuous

multiplication : S m1 S m1 S m1 . Let D = Dm and S = S m1 = D.
Divide S m into the upper and lower hemisphere: S m = B+ B Rm R, with
B+ B = S m Rm {0} = S. Using the decomposition

(D D) = D D D D = S D D S ,

the map f : (D D) S m is defined, for x, y S et t [0, 1], by

 
f (tx, y) = (t (x, y), 1 t 2 ) and f (x, ty) = (t (x, y), 1 t 2 ) .

For u, v, S m1 , we consider the hypothesis H(u, v) on :

H(u, v) : (u, x) = (u, x) and (x, v) = (x, v) for all x S m1 .

For example, H(u, v) holds for all u, v S m1 if = , the map associated

using (6.2.2) to a non-singular map : Rm Rm Rm . Also, H(e, e) is satisfied
if e is a neutral element for .
Example 6.3.2 Let : S 0 S 0 S 0 be the usual sign rule (S 0 = S =
{1}). Then D = [1, 1] and the map f : (D D) S 1 is pictured in
Fig. 6.1. One sees that f has degree 2. By Proposition 6.3.1, this implies that
Hopf (f ) = 1. Actually, the map f is topologically conjugate to the projection
S 1  S 1 /{x x}, so Cf is homeomorphic to RP2 .
The same exercise may be done for the other possible multiplications on S 0 . By
changing the sign of if necessary, we may assume that (1, 1) = 1. There are then
8 cases (Table 6.1). One sees that Hopf (f ) = 1 if and only if H(u, v) is satisfied
for some u, v S 0 . This is partially generalized in the following result.
Proposition 6.3.3 Let : S m1 S m1 S m1 be a continuous multiplication.
Suppose that H(u, v) is satisfied for some u, v S m1 . Then Hopf (f ) = 1.
6.3 The Hopf Invariant 251

(1, 1) (1, 1) f (A) = f (C)

A

f
B D 1 1

(1, 1) C (1, 1) f (B) = f (D)

Fig. 6.1 The map f for the usual sign rule

Table 6.1 The eight multiplications on S 0

(1, 1) (1, 1) (1, 1) (1, 1) DEG (f ) Hopf (f ) Satisfies
1 1 1 1 1 0 0
2 1 1 1 1 1 1 H(1, 1)
3 1 1 1 1 1 1 H(1, 1)
4 1 1 1 1 0 0
5 1 1 1 1 1 1 H(1, 1)
6 1 1 1 1 2 1 H(u, v) u, v
7 1 1 1 1 0 0
8 1 1 1 1 1 1 H(1, 1)

Proof The case m = 1 was done in Example 6.3.2. We may thus assume that m > 1.
The following proof is inspired by that of [81, Lemma 2.18]. Let f = f . Consider
the commutative diagram

 / H 2m (Cf )
H m (Cf ) H m (Cf )
O O

 / H 2m (Cf , S m )
H m (Cf , B+ ) H m (Cf , B )

 
 /
H (D D, S D) H (D D, D S)
m m
H (D D, (D D))
2m
O ggg3
ggg gg
1 2 ggggg
ggggg
H m (D, S) H m (D, S)

where : DD Cf is the characteristic map for the 2m-cell of Cf and = H .

The cross-product map at the bottom of the diagram is an isomorphism by the relative
Knneth theorem 4.6.10. Hence, Hopf (f ) = 1 if and only if the homomorphism
in the left column is an isomorphism. By symmetry, it is enough to prove
252 6 Projective Spaces

that : H m (Cf , B+ ) H m (D D, S D) is not zero. Consider the commutative

diagram
/ H m (S m , B+ ) / H m (B , S)
H m (Cf , B+ )
f f
  
/ H m ((D D), S D) / H m (D S, S S) .
H m (D D, S D)

The left horizontal maps are isomorphism since m > 1 and the right ones by excision.
It then suffices to prove that f : H m (B , S) H m (D S, S S) is not zero.
As the restriction of f to S S is equal to , one has a commutative diagram
/ H m (B , S)
H m1 (S)
f
 
/ H m1 (S S) / H m (D S, S S)
H m1 (D S)
O 2 nnnn
7
2 n n n
nnn
H m1 (S)

where the second line is the cohomology sequence of the pair (D S, S S).
Let u, v S m1 such that H(u, v) is satisfied. Let s1 , s2 : S S S be the
slice inclusions given by s1 (x) = (x, v) and s2 (x) = (u, x). The composition si
is thus an odd map. Therefore, Theorem 6.2.1 implies that H ( si )(a) = a. By
Corollary 4.7.3, one deduces that

H (a) = 1 a + a 1 . (6.3.1)

On the other hand, ker = Image 2 = {0, 1 a}. Therefore, f does not
vanish. 

In the proof of Proposition 6.3.3, the hypothesis on H(u, v) is only used to obtain
Eq. (6.3.1). Therefore, one has the following proposition.

Proposition 6.3.4 Let : S m1 S m1 S m1 be a continuous

multiplication. Suppose that H (a) = 1 a + a 1 for a H m1 (S m1 ). Then
Hopf (f ) = 1.

6.3.3 Dimension Restrictions

We shall prove in Corollary 8.6.4 that, if there exists a map f : S 2m1 S m with
Hopf invariant 1, then m = 2r . Actually, m = 1, 2, 4, 8 by a famous theorem of
Adams (see Theorem 8.6.6). This theorem implies the following result.
6.3 The Hopf Invariant 253

i
Proposition 6.3.5 Let S q
E
S p be a locally trivial bundle. Suppose that

H (E) H (S ). Then q = p 1 and p = 1, 2, 4 or 8.
p+q

Proof If p = 1 and q > 0, then H (E) is GrV-isomorphic to H (S 1 ) H (S q ) by

the argument of Example 4.7.45. Thus, we must have q = 0 and is a non-trivial
double cover of S 1 .
Let us suppose that p 2. If H (E) H (S p+q ), then H i is not surjective; oth-
erwise, H (E) is GrV-isomorphic to H (S p )H (S q ) by the Leray-Hirsch theorem.
The Wang exact sequences (see Proposition 4.7.43)

Hi 
H q (E) H q (S q )
H q+1p (S q ) .

then implies that q + 1 p = 0 (since p > 1). Therefore, q = p 1.

The bundle gluing map : S q S q S q (see p. 189) may thus be seen as
a continuous multiplication, to which a map f : S 2p1 S p may be associated
using (6.1.2). We shall prove that Hopf (f ) = 1. By Theorem 8.6.6, this implies that
p = 1, 2, 4 or 8.
Let a H q (S q ) be the generator. The restriction of to a slice {x} S q being a
homeomorphism, one has, using Lemma 4.7.2, that

H (a) = 1 a + (1 a)

for some Z2 . As  = 0 and p > 1, one gets from Proposition 4.7.46 that

0 = e (a) = H (a) +1a,

where e H p (S p ) is the generator. Therefore, H (a) = a 1 + 1 a. By
Proposition 6.3.4, this implies that Hopf (f ) = 1. 

Example 6.3.6 Consider the bundle S 1 E S 2 with gluing map (u, z) = uk z.

The total space E, obtained by gluing two copies of D S using the map ,
2 1 is
then a lens space with fundamental group of order k. Thus, if k is odd, E satisfies
the hypotheses of Proposition 6.3.5. Other famous examples are the bundles S 3
E S 4 which were used by J. Milnor to produce his exotic 7-spheres [145, Sect. 3].
Indeed, with a well chosen gluing map, the total space E is a smooth 7-manifold
homeomorphic but not diffeomorphic to S 7 .

6.3.4 Hopf Invariant and Linking Numbers

Let f : S 2m1 S m be a smooth map. Let y, y S m be two distinct regular values

of f . Then Q = f 1 ({y}) and Q = f 1 ({y }) are two disjoint closed submanifolds
of S 2m1 , both of dimension m 1. Therefore, their linking number l(Q, Q ) Z2
(see Sect. 5.4.4) is defined, at least if m > 1 (see also Remark 6.3.8).
254 6 Projective Spaces

Proposition 6.3.7 If m > 1, then Hopf (f ) = l(Q, Q ).

Actually, l(Q, Q ) is the original definition by H. Hopf of his invariant [99, 100].
Proposition 6.3.7 goes back to the work of Steenrod [182], after which the definition
of Hopf (f ) with the cup product in Cf was gradually adopted.

Proof Let  = S 2m1 . We consider the mapping cylinder Mf of f


S m ] {(x, 0) f (x) | x S 2m1 } .
Mf = [(S 2m1 I)

The correspondence (x, t)  f (x) descends to a retraction by deformation :

f
Mf S m . We identify  with the subspace  {1} of Mf . The mapping cone M
of f , defined as
M f = Mf  C Mf  D2m ,

where C D2m is the cone over , is homotopy equivalent to the CW-complex Cf .

We first introduce some material in order to compute l(Q, Q ) using
Lemma 5.4.23. Let W0 be a closed tubular neighbourhood of y in S m {y } and
let V0 be a closed tubular neighbourhood of y in int W0 (W0 and V0 are just m-
disks). Let y V0 W0 be a symmetric data for y with W0 W0 = . Let
B0 = {W0 , W0 , S m (V0 V0 )}. As y and y are regular values of f , we may assume,
provided W0 and W0 are small enough, that W = f 1 (W0 ) and V = f 1 (V0 ) are
nested tubular neighbourhoods of Q and that W = f 1 (W0 ) and V = f 1 (V0 ) are
nested tubular neighbourhoods of Q . Let B = f 1 (B0 ) = {W , W ,  (V V )}.
Let us briefly repeat the preliminary constructions for Lemma 5.4.23 (see p. 234
for more details), in our context, with the dimensions q = q = m1 and s = 2m1.
Let c0 Z m (W0 , W0 intV0 ) represent the Poincar dual class of {y} in H m (W0 , W0
intV0 ) H m (V0 , BdV0 ) and let c0 Z m (W0 , W0 intV0 ) represent the Poincar
dual class of {y } in H m (W0 , W0 intV0 ) H m (V0 , BdV0 ). Let c 0 , c 0 ZBm0 (S m )
be their zero extensions. Then c = C f (c0 ) Z m (W , W intV ) represent the
Poincar dual class of Q in H m (W , W intV ) H m (V , BdV ) and c = C f (c0 )
Z m (W , W intV ) represent the Poincar dual class of Q in H m (W , W intV )
H m (V , BdV ). Also, c = C f (c0 ) ZBm () and c = C f (c0 ) ZBm () are the zero
extensions of c and c . Choose a CB m1
() such that a = c . Let Z2m1 B ()
B
represent []B in H2m1 () H2m1 (). According to Lemma 5.4.23, one has
l(Q, Q ) = a  c , . We note that a  c ZB2m1 (). Indeed, for any
S2m1 (), one has
(a  c ),  = c  c ,  = 0 (6.3.2)

since the support of c is in W and that of c is in W . Therefore, (a  c ) = 0 and

a  c represents a cohomology classes |a  c | HB2m1 () (in this proof, we
use the notation | | for the cohomology class of a cocycle). The equality l(Q, Q ) =
a  c ,  is equivalent to

|a  c | = l(Q, Q ) []B , (6.3.3)
6.3 The Hopf Invariant 255

an equality holding in HB2m1 (), where []B is the generator of HB2m1 ()
H 2m1 () = Z2 .
Let B1 = 1 (B). The inclusion i :   Mf induces a morphism of cochain
complexes C i : CB (M ) C () whose kernel is denoted by C (M , ).
1
f B B1 f

Note that CB1 (Mf , ) = CB (M ) and so the inclusion C (M , )  C (M )
1
f B1 f B1 f
coincides with C j, the morphism induced by the pair inclusion j : (Mf , ) 
(Mf , ) (see Remark 3.1.26). One has the commutative diagram

(M , ) Cj (M ) Ci
0 / CB f / CB f / C () / 0
B
O t:
1 1

tt
t
C tt (6.3.4)
tt C f
(S m )
CB 0

where the top row is an exact sequence of cochain complexes. This sequence gives
rise to a connecting homomorphism sitting in the exact sequence

Hi Hj
HB2m1 (Mf ) / H 2m1 () / H 2m (Mf , ) / H 2m (Mf ) .
1 B B1 B1

As m > 1, one has HB2m1

1
(Mf ) H 2m1 (Mf ) = 0 and HB2m1 (Mf ) H 2m (Mf ) = 0.

Therefore : HB2m1 () HB2m1 (Mf , ) is an isomorphism. Let b = ([]B )
be the generator of HB2m1 (Mf , ). By (6.3.3), the linking number l(Q, Q ) is then
determined by the equation

(|a  c |) = l(Q, Q ) b . (6.3.5)

To compute (|a  c |), write M for the coboundary operator in CB (M ). Let

1
f

c 1 = C (c0 ) and c 1 = C (c0 ), both in ZB1 (Mf ). Let a1 C
m m1 (Mf ) such that
C i(a1 ) = a. By the commutativity of diagram (6.3.4), one has C i(a1  c 1 ) = a 
c . Then M (a1  c 1 ) is a cocycle in ker C i, so there is a unique u ZB2m1 (Mf , )
such that C j(u) = M (a1  c 1 ). As in Lemma 2.7.1, one has

|u| = (|a  c |) . (6.3.6)

The cohomology class |u| may be described in another way. As for (6.3.2), one
has c 1  c 1 = 0 for support reasons. Therefore,

(M (a1 ) + c 1 )  c 1 = M (a1 )  c 1 = M (a1  c 1 ) . (6.3.7)

Now, C i(M (a1 ) + c 1 ) = 0, thus there is a unique w ZBm1 (Mf , ) with C j(w) =
M (a1 ) + c 1 . The first cup product of (6.3.7) may be understood as relative cochain
256 6 Projective Spaces

cup product giving rise to a relative cohomology cup product


HB 1 (Mf , ) HB 1 (Mf )
HB 1 (Mf , )

analogous to that of Lemma 4.1.14 (in the case Y2 = ). Equation (6.3.6) is equiv-
alent to

|w|  |c1 | = (|a  c |) (6.3.8)

and, using (6.3.6), we get the equality

|w|  |c1 | = l(Q, Q ) b (6.3.9)

holding in HB2m1 (Mf , ).

Now, under the isomorphism H m (S m ) HBm0 (S m ) (due to the small simplex
theorem 3.1.34), the cohomology class |c0 | corresponds to the Poincar dual PD({y})
H m (S m ), which is [S m ] . Analogously, |c0 | also corresponds to [S m ] . Hence,

under the isomorphism H m (Mf ) HBm1 (Mf ), |c1 |, |c1 | and H j(|w|) all correspond

to H ([S ] ), that is the generator e of H m (Mf ) Z2 . The diagram
m

 / H 2m (Mf , )
HBm1 (Mf ) HBm1 (Mf , ) B1
O O

 / H 2m (Mf , )
H m (Mf ) H m (Mf , )
O O
(6.3.10)
 / H 2m (M
f ) H m (M
H m (M f , C) f , C)

 
 / H 2m (M
f)
H (Mf ) H m (M
m f)

is commutative, where the vertical arrows are the obvious ones or induced by
the inclusions (the commutativity of the bottom square is the content of the
singular analogue of Lemma 4.1.8). Let k : Mf M f denote the inclusion. Then

Hk (e) = e , the generator of H (Mf ) Z2 . Equation (6.3.9), obtained using the top
m

line of (6.3.10), becomes, using the bottom line

e  e = l(Q, Q ) b (6.3.11)
6.3 The Hopf Invariant 257

where b is the generator of H 2m (M f ) Z2 . But, as m > 1, the equality

e  e = Hopf (f ) b holds true, by definition of the Hopf invariant. Therefore,
Hopf (f ) = l(Q, Q ).

Remark 6.3.8 For a map f : S 1 S 1 with even degree, the equality Hopf (f ) =
l(Q, Q ) holds true, using Proposition 6.3.1 (see Exercise 6.11). When deg f is odd,
the linking number l(Q, Q ) is not defined. Indeed, both Q and Q have an odd number
of points and condition (3) of p. 231 is not satisfied.

6.4 Exercises for Chapter 6

6.1. What is the Lusternik-Schnirelmann category of KPn for K = C or H.

6.2. Compute the cohomology ring of CPn and HPn , using the Gysin exact
sequence for the Hopf bundles. [Hint: like in Sect. 4.3.4.]
6.3. For K = C or H, prove that H (KP ) H (RP ) is injective.
6.4. Let f : X Y be a map. The double mapping cylinder CCf of f is the
union of two copies of the mapping cylinder Cf of f glued along X. Compute
H (CCj ) where j is the inclusion of CP HP .
6.5. Prove that X = S 4 S 4 and Y = HP2 HP2 have the same Poincar polyno-
mial. Are H (X) and H (Y ) GrA-isomorphic?
6.6. For any positive integer n, construct a vector bundle of rank n over a closed
n-dimensional manifold such that e() = 0.
6.7. Let X be a CW-complex of dimension n = 1, 2, 4, 8 and let a H n (X). Prove
that there exists a vector bundle over X with e() = a.
6.8. Prove that there is no continuous injective map f : Rn Rk if n > k. [Hint:
use the Borsuk-Ulam theorem.]
6.9. Check the table of p. 251.
6.10. Show that the Hopf vector bundle over KP1 S d (d = dimR K) cannot be
the normal bundle of an embedding of S d into a manifold M of dimension 2d
with Hd (M) = 0.
6.11. Let f : S 1 S 1 be a smooth map with even degree. Show that the Hopf
invariant of f is equal to the linking number of the inverse image of two
regular values of f , as in Proposition 6.3.7. [Hint: use Proposition 6.3.1.]
6.12. Using the linking numbers and Proposition 6.3.7, show that the various Hopf
maps have Hopf invariant 1. [Hint: use Formula (6.1.3) and Exercise 5.18.]
6.13. Let g : S 2m+1 S m be a continuous map, as well as f : S m S m and
h : S 2m+1 S 2m+1 . Prove that Hopf (f g h) = deg(h) deg(f )2 Hopf (g).
(Remark: of course, deg(f )2 deg(f ) mod 2 but the formula is the one which
is valid for the cohomology with any coefficients.)
6.14. Let f : S 2m1 S m be a smooth map and let y S m be a regular value
for f . The closed (m 1)-manifold Q = f 1 ({y}) bounds a compact manifold
W (see Exercise 9.14). As S 2m1 pt is contractible, the inclusion of Q into
S 2m1 extends to a smooth map j : W S 2m1 (see Exercise 3.2). We
258 6 Projective Spaces

thus get a homomorphism H (f j) : Hm (W , Q) Hm (S m , {y}). As both

these homology groups are isomorphic to Z2 , this defines a degree for f j, as
in (2.5.4). Prove that Hopf (f ) = deg(f j).
Chapter 7
Equivariant Cohomology

In ordinary life, the symmetries of an object (like a ball or a cube) help us to apprehend
it. The same should happen in topology when studying spaces with symmetries, i.e.
endowed with actions of topological groups. Equivariant cohomology is one tool for
such a purpose.
Our aim here is mostly to develop enough material needed in the forthcoming
chapters. For instance, the definition and most properties of the Steenrod squares use
equivariant cohomology for spaces with involution. This case is treated in detail in
Sect. 7.1, at an elementary level and with ad hoc techniques. A second section deals
with -spaces for any topological group  (the proof of the Adem relations requires
-equivariant cohomology with  the symmetric group Sym4 ). Equivariant cross
products, treated in Sect. 7.4, will also be used. Only Sect. 7.3 is written uniquely for
its own interest, devoted to some simple form of localization theorems and Smith
theory. A final section presents the equivariant Morse-Bott theory, used in Sect. 9.5
to compute the cohomology of flag manifolds (see also Sect. 10.3.5). For further
reading on equivariant cohomology, see e.g. [9, 37, 103].

7.1 Spaces with Involution

An involution on a topological space X is a continuous map : X X such that
= id. The letter is usually used for all encountered involutions. We also use
the symbol for the non-trivial element of the cyclic group G = {id, } of order 2;
an involution on X is thus equivalent to a continuous action of G on X and a space
with involution is equivalent to a G-space, i.e. a space together with an action of
G. We often pass from one language to the other. If X is a G-space, its fixed point
subspace X G is defined by

X G = {x X | (x) = x}.

As G has only two elements, the complement of X G is the subspace where the action
is free.
Springer International Publishing Switzerland 2014 259
J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_7
260 7 Equivariant Cohomology

A continuous map f : X Y between G-spaces is a G-equivariant map,

or just a G-map if it commutes with the involutions: f = f . Two G-maps
f 0 , f 1 : X Y are G-homotopic if there is a homotopy F : X I Y connecting
them which is a G-map. Here, the involution on X I is (x, t) = ( (x), t). This
permits us to define the notion of G-homotopy equivalence and of G-homotopy type.
For instance, a G-space is G-contractible if it has the G-homotopy type of a point.
Let X be a G-space. A CW-structure on X with set of n-cells n is a G-CW-
structure if the following condition is satisfied: for each integer n, there is a G-action
on n and a G-equivariant global characteristic map n : n D n X , where
the G-action on n D n is given by (, x) = ( (), x). In particular, if n
satisfies () = , then restricted to {} D n is the identity. These cells are
called the isotropic cells; they form a G-CW-structure for X G . The other cells, the
free cells, come in pairs (, ()). A G-space endowed with a G-CW-structure is a
G-CW-complex, or just a G-complex. Observe that, if X is a G-complex, then the
quotient space X/G inherits a C W -structure (with set of n-cells equal to n /G) for
which the quotient map is cellular. A smooth G-manifold admits a G-CW-structure,
in fact a G-triangulation (see [106]).

Example 7.1.1 Let X = S n (n ) be the standard sphere endowed with the

CW-structure where the m-skeleton is S m and having two m-cells in each dimension
m n (see Example 3.4.4). This is a G-CW-structure for the free involution given
by the antipodal map z  z. The quotient space X/G is RP n with its standard
CW-structure.

Let X be a space with an involution . The Borel construction X G , also known

as the homotopy quotient, is the quotient space

X G = S G X = (S X ) (7.1.1)

where is the equivalence relation (z, (x)) (z, x). If X and Y are G-spaces and
if f : Y X is a continuous G-equivariant map, the map id f : S Y S X
descends to a map f G : YG X G . This makes the Borel construction a covariant
functor from the category TopG to Top,TopG where is the category of G-spaces
and G-equivariant maps. Using the obvious homeomorphism between (X I )G and
X G I , a G-homotopy between two G-maps f 0 and f 1 : X Y descends to a
homotopy between f G0 and f G1 . Hence, X G and YG have the same homotopy type if
X and Y have the same G-homotopy type.
Let p : S RP be the quotient map (this is a 2-fold covering projection). A
map p : X G RP is then given by p([z, x]) = p(z). Observe that p coincides
with the map f G : X G ptG = RP induced by the constant map X pt.

Example 7.1.2 Suppose that the involution is trivial, i.e. (x) = x for all x X .
The projection S X X then descends to X G X . Together with the map p,

this gives a homeomorphism X G RP X .
7.1 Spaces with Involution 261

Lemma 7.1.3 (1) The map p : X G RP is a locally trivial fiber bundle with
fiber homeomorphic to X .
(2) If f : Y X is a G-equivariant map, then the diagram

fG
YG / XG
FF w
FFp p ww
FF ww
F# {ww
RP

is commutative.
(3) If has a fixed point, then p admits a section. More precisely, the choice of a
point v X G provides a section sv : RP X G of p.
(4) The quotient map S X X G is a 2-fold covering admitting p as a charac-
teristic map.

Proof We use that p : S RP is a principal G-bundle, i.e. a 2-fold covering.

Denote by z = (z 0 , z 1 , . . . ) the elements of S . The set Vi = {z S | z i
= 0}
is an open subspace of S . As p is an open map, Ui = p(V i ) is an open set of
RP . A trivialization i : Vi Ui {1} is given by i (z) = ( p(z), z i /|z i |).

Using the group isomorphism {1} G, this gives a trivialization i : Vi

Ui G. Now, i id : Vi X Ui G X descends to a homeomorphism

p 1 (Ui ) Ui (G G X ). Here, G G X denotes the quotient of G X by
the equivalence relation (g, (x)) (g , x). But the map x  (id, x) provides a
homeomorphism from X onto G G X . This proves Point (1). This also shows that,
over p 1 (Ui ), the map S X X G looks like the projection G p 1 (Ui )
p 1 (Ui ). Therefore, S X X G is a 2-fold covering, with the product involution
(z, x) = (z, (x)) as deck transformation. The diagram

proj S
S X / S

 p

XG / RP

is commutative and proj S ( (y)) = proj S (y). By Lemma 4.3.4, this implies
that p is a characteristic map for the covering S X X G . Point (4) is thus
established.
Point (2) is obvious from the definitions. For Point (3), let v X G . By Point (2),
the inclusion i : {v}  X gives rise to a commutative diagram

iG
{v}G / XG
HH w
HH p ww
.
HH ww
H# {ww
RP
262 7 Equivariant Cohomology

Hence, i G provides a section sv : RP X G of p, depending on the choice of the

fixed point v. 
The projection q : S X X seen in Example 7.1.2 descends to q : X G
X/G.

Lemma 7.1.4 Let X be a free G-space such that X is Hausdorff. Then, the GrA-
morphism H q : H (X/G) H (X G ) is an isomorphism. Moreover, if X is a free
G-complex, then the map q : X G X/G is a homotopy equivalence and the map
p : X G RP is homotopic to the composition of q with a characteristic map for
the covering X X/G.

Proof If X is Hausdorff, such a projection X X/G is a 2-fold covering and X/G

is Hausdorff. Over a trivializing open set U X/G, this covering is equivalent to
G U U . Then (G U )G S U since any class has a unique representative
q
of the form (z, id, u) S G U . Hence, X G X/G is a locally trivial
bundle with fiber S . As H (S ) = 0, the map q : H (X/G) H (X G )
is a GrA-isomorphism by Corollary 4.7.21. Actually as S is contractible [82,
example 1.B.3 p. 88], the homotopy exact sequence of the bundle [82, Theorem 4.41
and Proposition 4.48] implies that q is a weak homotopy equivalence. If X is a
G-complex, then X/G is a C W -complex. Also, S X is a free G-complex and
thus X G is a CW-complex. Therefore, a weak homotopy equivalence is a homotopy
equivalence by the Whitehead theorem [82, Theorem 4.5]. Also, again since S
is contractible and X/G is a CW-complex, a direct proof that q is a homotopy
equivalence is available using [42, Theorem 6.3].
Let f : X/G RP be a characteristic map for the covering X X/G. The
diagram
f
S X / X / S

covering covering
 q
 f

XG / X/G / RP

is commutative and the upper horizontal arrows commute with the deck involu-
tions. By Lemma 4.3.4, this implies that f q is a characteristic map for the cov-
ering S X X G . By Lemma 7.1.3, so is p. By Corollary 4.3.3, two char-
acteristic maps of a covering are homotopic. Therefore, the maps p and f q are
homotopic. 

Corollary 7.1.5 Let X be a finite dimensional G-complex. Then, the following con-
ditions are equivalent.
(1) X has a fixed point.
(2) The morphism H p : H (RP ) H (X G ) is injective.

Proof If X has a fixed point, then p admits a section by Lemma 7.1.3, so H p is

injective. If X has no fixed point, then X is a free G-complex and, by Lemma 7.1.4,
7.1 Spaces with Involution 263

H (X G ) H (X/G). Also, X/G is a finite dimensional CW-complex, so H p is

not injective. 

Let X be a space with an involution and let Y X be an invariant subspace. Then

YG X G . The (relative) G-equivariant cohomology HG (X, Y ) is the cohomology
algebra

HG (X, Y ) = H (X G , YG ).

We shall mostly concentrate on the absolute case HG (X ) = H (X G ) = HG (X, ).

The map p : X G RP induces a GrA-homomorphism p : H (RP )
HG (X ). By Proposition 6.1.1, H (RP ) is GrA-isomorphic to the polynomial ring
Z2 [u], where u is a formal variable in degree 1. Hence, the GrA-homomorphism p
gives on HG (X ) a structure of Z2 [u]-algebra. In particular, HG ( pt) = Z2 [u].
As an important example, let us consider the case of a G-space Y with Y = Y G ,
i.e. the involution is trivial. As seen in Example 7.1.2, we get an identification
YG = RP Y . By the Knneth theorem,

HG (Y ) Z2 [u] H (Y ) H (Y )[u]. (7.1.2)

The GrA-homomorphism H (Y ) HG (Y ) induced by the projection RP Y

Y corresponds to the inclusion of the ring of constants H (Y ) into H (Y )[u].
The functoriality of the Borel construction and of the cohomology algebra,
together with Point (2) of Lemma 7.1.3, says that, if f : Y X is a G-equivariant
map between G-spaces, then H f G : HG (X ) HG (Y ) is a GrA-homomorphism
commuting with the multiplication by u. We are then driven to consider the category
GrA[u] whose objects are graded Z2 [u]-algebras and whose morphisms are GrA-
homomorphism commuting with the multiplication by u. Hence, the correspondence
X  HG (X ) is a contravariant functor from TopG to GrA[u]. If f : Y Y is a G-
equivariant map between trivial G-spaces (i.e., any continuous map), the following
diagram is commutative:

/ H (Y )
H (Y )[u] G

H f [u] HG f (7.1.3)
 
/ H (Y ) .
H (Y )[u] G

Choosing a point z S provides, for each G-space X , a map i z : X X G

defined by i z (x) = [z, x]. As S is path-connected, the homotopy class of i z does
not depend on z. Therefore, we get a well defined GrA-homomorphism

: HG (X ) H (X ) (7.1.4)
264 7 Equivariant Cohomology

given by = H i z for some z S . We can call the forgetful homomorphism

(it forgets the G-action). Observe that is functorial. Indeed, if f : X X is a
G-equivariant map, the diagram

iz
X / XG

f fG
 i f (z)

X / X
G

is commutative, and so is the diagram

X
H ( X ) o HG ( X )

H f HG f (7.1.5)
 
X
H (X ) o HG (X ) .

If Y is a G-space with Y = Y G , we get, using the GrA[u]-isomorphism of (7.1.2),

the commutative diagram

/ H (Y )
H (Y )[u] G
KKK t
KKK tt
ev0 KKK tt (7.1.6)
% tz tt
H (Y )

where ev0 is the evaluation of a polynomial at u = 0, i.e. the unique algebra homo-
morphism extending the identity on H (Y ) and sending u to 0.
We now explain how some information on HG (X ) may be obtained form transfer
exact sequences. Observe that = H where : X X G is the composition

slice
: X X S X G .

As S is contractible [82, example 1.B.3 p. 88], the slice inclusion is a homotopy

equivalence. By Lemma 7.1.3, the map X S X G is then a 2-fold covering. Its
characteristic map is p : X G RP and, by Lemma 4.3.6, its characteristic class
coincides with u HG1 (X ). Therefore, the transfer exact sequence of the covering
(see Proposition 4.3.9) gives the exact sequence

u tr u
HGm1 (X ) HGm (X )
H m (X ) HGm (X ) HGm+1 (X ) .
(7.1.7)
7.1 Spaces with Involution 265

Denote by (u) the ideal of HG (X ) generated by u and by

Ann (u) = {x HG (X ) | ux = 0}

the annihilator of u. The information carried by Sequence (7.1.7) may be concentrated

in the following short exact sequence of graded Z2 [u]-modules:

tr
0 HG (X )/(u)
H (X ) Ann (u) 0. (7.1.8)

A G-space X is called equivariantly formal if : HG (X ) H (X ) is surjective.

For instance, X is equivariantly formal if the G-action is trivial. See 7.2.9 for a
discussion of this definition in a more general setting.
Proposition 7.1.6 For a G-space X , the following conditions are equivalent.
(1) X is equivariantly formal.
(2) HG (X ) is a free Z2 [u]-module.
(3) Ann (u) = 0.
Proof That (2) (3) is obvious and (3) (1) follows from (7.1.8). For (1) (2),
choose a GrV-section : H (X ) HG (X ) of (as is surjective). Then is
p
a cohomology extension of the fiber for the fiber bundle X X G RP . As

RP is path-connected and of finite cohomology type, the Leray-Hirsch theorem II
(Theorem 4.7.18) implies that HG (X ) is a free Z2 [u]-module generated by (B),
where B is a Z2 -basis of H (X ). 
Remark 7.1.7 As noted before, H (X ) Z2 [u] is isomorphic, as a Z2 [u]-algebra,
to H (X )[u]. If X is equivariantly formal, the Leray-Hirsch theorem II (Theo-
rem 4.7.18) thus provides an isomorphism of Z2 [u]-modules between HG (X ) and
H (X )[u]. This isomorphism depends on the choice of a GrV-section : H (X )
HG (X ) of : HG (X ) H (X ) and is not, in general, an isomorphism of algebras.
However, as in the case of a trivial G-action, Diagram (7.1.6) is commutative.
Corollary 7.1.8 Let X be a G-space. Suppose that r : HG (X ) HG (X G ) is
injective. Then X is equivariantly formal.
Proof Since HG (X G ) is a free Z2 [u]-module, the hypothesis implies that Ann
(u) = 0. 
The converse of Corollary 7.1.8 is true in some cases (see Proposition 7.3.9), but
not in general. For example, S is equivariantly formal since H (S ) H ( pt).
But HG (S ) = H (RP ) by Lemma 7.1.4 and HG ((S )G ) = HG () = 0.
For X a G-space, let r : HG (X ) HG (X G ) and r : H (X ) H (X G ) be
the GrA[u]-homomorphism induced by the inclusion X G  X . One can compose
tr : H (X ) HG (X ) with r .
Proposition 7.1.9 r tr = 0.
266 7 Equivariant Cohomology

Proof As G acts trivially on X G , one has HG (X G ) H (RP ) H (X G ). The

commutative diagram

r / H (S X G ) o
H (S X ) H (S ) H (X G )
tr tr tr id
  
r / H (RP X G ) o
H (S G X ) H (RP ) H (X G )

proves the proposition since tr : H (S ) H (RP ) is the zero map (even in

degree 0, by the transfer exact sequence). 
To say more about the image of : HG (X ) H (X ), define

H (X )G = {a H (X ) | H (a) = a} H (X ).

As H is a GrA-morphism, H (X )G is a Z2 -graded subalgebra of H (X ).

Lemma 7.1.10 (HG (X )) H (X )G .
Proof Let z S and b HG (X ). Then,

H (b) = H H i z (b) = H (i z )(b) = H i z (b) = (b). 

The reverse inclusion in Lemma 7.1.10 may be wrong, as shown by the following
example.
Example 7.1.11 Let be the antipodal map on the sphere X = S n , making X a
free G-complex, as seen in Example 7.1.1. The equality H n (X ) = H n (X )G holds
iz q
true since H n (X ) = Z2 . For any z S , the composition X XG X/G
coincides with the quotient map X X/G = RP n . By Lemma 7.1.4, the map
q is a homotopy equivalence. But, since S n RP n is of local degree 2, the
homomorphism H (RP ) H (S ) vanishes. Thus = 0. In this example, the
n n n n

non-existence of fixed points is important (see Proposition 7.1.12).

Let X be a G-space. The reduced equivariant cohomology H G (X ) is the GrA[u]-
algebra defined by
 
H G (X ) = coker H pG : HG ( pt) HG (X ) (7.1.9)

where pG : X pt denotes the constant map to a point (which is G-equivariant).

Warning: H G (X )
= H (X G ). Here are some examples.
(1) H G ( pt) = 0.
(2) Let X = S n with the antipodal involution. By Lemma 7.1.4, X G has the homo-
topy type of RP n and H p : H (RP ) HG (X ) is surjective. Therefore,
H G (X ) = 0.
7.1 Spaces with Involution 267

(3) If Y is a space with trivial G-action, one has a natural GrA[u]-isomorphism

H G (Y ) = H (Y )[u]/Z2 [u] H (Y )[u] (7.1.10)

(4) If X is equivariantly formal, we get, as in Remark 7.1.7, an isomorphism of

Z2 [u]-modules between H G (X ) and H (X )[u]. This isomorphism depends on
the choice of a section of : HG (X ) H (X ) and is not, in general, an
isomorphism of algebra.
Any G-equivariant map f : Y X satisfies p f = p, so H G is a contravariant
functor from TopG to GrA[u]. One checks that the homomorphisms : HG (X )
H (X )G and tr : H (X ) HG (X ) descend to : H G (X ) H (X )G and to

tr : H (X ) H G (X ).
The equivariant reduced cohomology will be further developed in a more general
setting (see 7.2.10 in the next section). Here, we shall only prove the following
proposition, which plays an important role in the construction of the Steenrod squares
in Chap. 8. Note that H (X )G contains the classes a + (a) for all a H (X ).
Proposition 7.1.12 Let X be a G-space with X G
= . Suppose that H j (X ) = 0
for 0 j < r . Then, : H Gr (X ) H r (X )G is an isomorphism. Moreover,

1 (a + (a)) = tr (a) for all a H r (X ).

Proof The last assertion follows from the main one since tr (a) = a + (a) by
definition of the transfer. We shall prove that the sequence

H p
0 HGr ( pt) HGr (X )
H r (X )G 0 (7.1.11)

is exact. This will prove the main assertion.

For 0 k , let

Z k = S k G X.

Thus, X G = Z and, as in Lemma 7.1.3, there is a natural locally trivial bundle

p : Z k RP k with fiber X .
Choosing a point z S 1 provides a map i z : X Z 1 Z k , defined by i z (x) =
[z, x], which induces a GrA-homomorphism : H (Z k ) H (X ) (independent of
z) given by = H i z . As in Lemma 7.1.10, one proves that (H (Z k )) H (X )G .
We shall prove, by induction on k, that the sequence

H p
0 H r (RP k ) H r (Z k )
H r (X )G 0 (7.1.12)

is exact for each k 1. Since any compact subset of X G = Y is contained in Z k

for some k, the exactness of (7.1.11) will follow, using Corollary 3.1.16.
Observe that, in Sequence (7.1.12), the homomorphism H p is injective since the
choice of a G-fixed point in X provides a section of p. It is also clear that H p = 0.
268 7 Equivariant Cohomology

We start with k = 1. The space Z 1 is the mapping torus of . The mapping torus
exact sequence of Proposition 4.7.44 is of the form

 J 
H r 1 (X )
H r 1 (X )
H r (Z 1 )
H r (X )
H r (X ) ,


where  = id + H . Hence, ker(H r (X ) H r (X )) = H r (X )G . If r 2, then
H (X ) = 0, which proves that : H (Z 1 ) H r (X )G is an isomorphism and thus
1 r

Sequence (7.1.12) is exact (for k = 1). If r = 1, then  : H 0 (X ) H 0 (X ) is the

null-homomorphism, since X is path-connected. Then, ker(H 1 (Z 1 ) H 1 (X )G )
Z2 which implies that Sequence (7.1.12) is also exact when k = r = 1.
Take, as induction hypothesis, that Sequence (7.1.12) is exact for k =  1 1.
We have to prove that it is exact for k = . The space Z  is obtained from Z 1 by
gluing D  X using the projection S 1 X Z 1 . Let e be the generator of
H  (D  , S 1 ) = Z2 . Using excision and the relative Knneth theorem, we get the
commutative diagram

H (RP  , RP 1 ) / H (D  , S 1 ) o e
H  ( pt)

  
H (Z  , Z 1 )
/ H (D  X, S 1 X ) o e
H  (X )

This diagram, together with the cohomology sequences for the pairs (RP  , RP 1 )
and (Z  , Z 1 ) gives the commutative diagram:

H r  ( pt) / H r (RP  ) / H r (RP 1 ) / H r +1 ( pt)

 
H p H p
   
H r  (X ) / H r (Z  ) / H r (Z 1 ) / H r +1 (X ) . (7.1.13)

 
= / H r (X )G
H r (X )G

where the two long lines are exact. The induction step follows by comparing the two
middle columns. The argument divides into four cases.

Case 1  < r . As  2, one has 0 < r  r 2. By hypothesis, H j (X ) = 0 for

1 j < r . Therefore the left and right columns vanish and the two middle columns
are isomorphic.
Case 2  = r . Since  2, one has H 1 (X ) = 0 and the right column vanishes. Also,
H r (RP 1 ) = 0 and the left vertical arrow is an isomorphism (since H 0 (X ) = 0).
The induction step follows. Observe that H 0 (X ) H r (Z  ) is injective.
7.1 Spaces with Involution 269

Case 3  = r + 1. The left column vanishes. By step 2, Diagram (7.1.13) continues

on the right by injections

H 0 ( pt) / / H r +1 (RP  )

H p .
 
H 0 (X ) / / H r +1 (Z  )

Hence, the two middle columns are isomorphic.

Case 4  > r + 1. The left and right columns vanish for dimensional reasons, so
the two middle columns are isomorphic.
Remark 7.1.13 The Serre spectral sequence for the bundle X X G RP
provides a shorter proof of the exactness of sequences (7.1.11) and (7.1.12). This
will be used to prove the more general Proposition 7.2.17 in the next section.
Example 7.1.14 Linear involution on spheres. Let S n be the standard sphere equipped
with an involution O(n + 1). In Rn+1 , the equality

x + (x) x (x)
x= +
2 2

gives the decomposition Rn+1 = V+ V with V being the eigenspace for the
eigenvalue 1. As is an isometry, the vector spaces V+ and V are orthogonal.
Therefore, two elements ,  O(n + 1) of order 2 are conjugate in O(n + 1)

if and only if dim(S n ) = dim(S n ) . We write S np , (1 p n) for the sphere
S equipped with an involution O(n + 1) such that dim(S n ) = p. Hence,
n

(S np ) S p . The equivariant CW-structure on S n (see Example 3.4.5) provides a

G-CW-structure on S np for all p n.
n is just the antipodal map and, by Lemma 7.1.4, (S n )
The involution on S1 1 G
RP . For p 0, the inclusion S p = (S np )G  S np gives rise to GrA[u]-morphisms
n

r : HG (S np ) HG ((S np )G ) = HG (S p ) H (S p )[u]

and

r : H G (S np ) H G ((S np )G ) = H G (S p ) H (S p )[u].

If n 1, then H j (S np ) = 0 for 0 j < n and Proposition 7.1.12 asserts that :

H Gn (S np ) H n (S np )G = H n (S np ) is an isomorphism (this is also true if n = p = 0).
Let a H n (S n ) and b H p (S p ) be the generators.
Proposition 7.1.15 When p 0 the GrA[u]-morphisms r and r are injective.
Moreover, r 1 (a) = b u n p .
270 7 Equivariant Cohomology

Proof The proposition is trivial if n = p, so we can suppose that n > p 0. Using

the commutative diagram

0 / H ( pt) / HG (S np ) / H (S n ) / 0
G G p

= r r ,
  
0 / H ( pt) / H (S p ) / H (S p ) / 0
G G G

the five-lemma technique show that r is injective if and only if r is injective. Thus
we shall prove that r is injective.
We first prove that r is injective when p = 0. One can see S0n as the suspension
n , with (S n )G = { , } S 0 . Then, X = S n is the union of the G-
of S1 0 + 0
equivariantly contractible open sets X + = S0n { } and X = S0n {+ }, with
n . Hence, X has the
intersection X 0 having the G-equivariant homotopy type of S1 G
homotopy type of { }G and

HG (S 0 ) HGk ({ }) HGk ({+ }) HGk (X ) HGk (X + ).

By Lemma 7.1.4, X G0 has the homotopy type of RP n1 . By Proposition 3.1.53, the

+
Mayer-Vietoris data (X G , X G , XG , XG
0 ) gives rise to the long exact sequence

M V M V
H k1 (RP n ) / H k (S n ) r / H (S 0 ) J / H k (RP n ) /
G 0 G
(7.1.14)
with J = HG j + + HG j , where j : X 0  X denotes the inclusion. The
map j is G-homotopy equivalent to the constant map S1 n  . As noted

before Example 7.1.2, the induced map HG ( ) HG (X ) is the GrA-morphism
0

H p : H (RP ) H (RP n1 ). By Lemma 7.1.4, p is the characteristic map for

n1 n1
the covering S1 (S1 )G = RP n1 . As noticed in Example 4.3.5, this map
is just the inclusion RP n1  RP and, then, HG j is surjective by Proposi-
tion 4.3.10. Hence, J is surjective and the exact sequence (7.1.14) splits. This proves
that r : HG (S0n ) HG (S 0 ) is injective. For a more precise analysis of HG (S0n ) see
Examples 7.1.16 or 7.6.9.
Suppose, by induction on p 1, that r : HGi (S m p1 ) HG (S
i p1 ) is injective

for all i N and all m p 1. We have to prove that r : HG (S np ) HGk (S p )

k

is injective for all k N and all n p. As (S np )G and (S p )G are path-connected,

the required assertion is true for k = 0 by Lemma 3.1.9. Thus, we may suppose
that k 1. As n p 1, the G-sphere S n1 n
p1 exists and S p is the suspension of
S n1 n G n1 G
p1 , with (S p ) being the suspension of (S p1 ) . Let Y = Y
G = { , } be
+
the suspension points. As above, we decompose the X = S p = X X + with
n

X X + = X Y . The maps r sit in a commutative diagram

7.1 Spaces with Involution 271

M V
HGk1 (Y ) / H k1 (S n1 ) / H k (S n ) / H k (Y )
G p1 G p G

r r r r (7.1.15)
   
M V
HGk1 (Y G ) / H k1 (S p1 ) / H k (S p ) / H k (Y G )
G G G

where the horizontal line are the Mayer-Vietoris sequences for the data (X, X + , X ,
X Y ) and (X G , (X + )G , (X )G , (X Y )G ). Hence, r : HGk (S np ) HGk (S p ) is
injective by the proof of the five lemma (see [82, p. 129]).
The last assertion is now obvious, since r 1 is injective and, as H G (S p )
H (S p )[u], one has H Gn (S p ) = Z2 u n p .


Example 7.1.16 We use the notations of the proof of Proposition 7.1.15 in the case
p = 0, with (S0n )G = S 0 = { }. The isomorphism : H n (S0n ) HGn (S0n )
defined by the commutative diagram

1 r
H n (S0n ) / H n (S n ) / / H n (S 0 )
G 0 G
O O O

HGn (S0n , ) / r / H n (S 0 , )
G

 

H n (S0n ) / H n (S n ) / r / H n (S 0 ) o / H n ( ) H n (+ )
G 0 G G G

is an extension of the fiber for the bundle S0n (S0n )G RP . Another one,
+ , is obtained using + (there are two of them by the exact sequence (7.1.11) and
+ (a) = (a) + u n ). Then, r (a) = (u n , 0) and r + (a) = (0, u n ). Hence,
neither nor + is multiplicative. We see the relation r (a)2 = u n r (a).
Hence, as r is a monomorphism of Z2 [u]-module, the relation (a)2 = u n (a)
holds in HG (S0n ). By the Leray-Hirsch Theorem 4.7.17, HG (S0n ) is a free Z2 [u]-
module generated by A = + (a) (or, by B = (a)). By dimension counting, we
check that HG (S0n ) admits, as a Z2 [u]-algebra, the presentation

HG (S0n ) Z2 [u][A] (A2 + u n A) . (7.1.16)

As (a)+ (a) = 0, a more symmetric presentation is obtained using the two

generators A and B:

HG (S0n ) Z2 [u][A, B] I
272 7 Equivariant Cohomology

where I is the ideal generated by

A + B + u n , and A2 + u n A.

Note that AB and B 2 + u n B are in I. Indeed, mod I, one has

AB = A(A + u n ) = A2 + u n A = 0

and

B 2 = (A + u n )2 = A2 + u 2n = u n A + u 2n = u n (A + u n ) = u n B.

Corollary 7.1.17 If p 0, then S np is equivariantly formal. There is a section

: H n (S np ) HGn (S np ) of such that r : H n (S np ) HGn ((S np )G ) H (S p )[u]
satisfies

r (a) = b u n p . (7.1.17)

Proof By Proposition 7.1.15 : H Gn (S np ) H n (S np ) is an isomorphism, so the

commutative diagram

HGn (S np ) / H n (S np )

 

H Gn (S np ) / H n (S n )
p

shows that is surjective and thus S np is equivariantly formal. Choose a section


of . By Proposition 7.1.15, Eq. (7.1.17) holds true modulo ker HGn (S np )

H Gn (S np ) = Z2 u n . By changing (a) by (a)+u n if necessary, (7.1.17) will hold true
strictly. 

As another example, we consider CP n as a G-space with the involution being

the complex conjugation. Thus, (CP n )G = RP n . Let 0
= a H 2 (CP n ) and
0
= b H 1 (RP n ).

Proposition 7.1.18 For n , CP n is equivariantly formal. Moreover, there is a

section : H (CP n ) HG (CP n ) which is multiplicative and satisfies r (a) =
bu + b2 .

Proof As H i (CP n ) = 0 for i 1, Proposition 7.1.12 implies that : H G2 (CP n )

H 2 (CP n ) is an isomorphism. As in the proof of Corollary 7.1.17, this implies that
: HG2 (CP n ) H 2 (CP n ) is surjective. As H (CP n ) is generated by a as an
algebra and is a GrA-morphism, we deduce that : HG (CP n ) H (CP n ) is
surjective. Thus CP n is equivariantly formal.
7.1 Spaces with Involution 273

Choose a section 2 : H 2 (CP n ) HG2 (CP n ) of . As HG (RP n ) H (RP n )

[u], there exists , and in Z2 such that

r 2 = u 2 + bu + b2 .

By changing 2 (a) by 2 (a) = (a) + u 2 , we may assume that = 0. We must

prove that = = 1. The inclusions i
RP n CP n and j : CP 1 CP n provide commutative diagrams

r / H 2 (RP n ) HG2 (CP n ) / H 2 (CP n )
HG2 (CP n ) G m

HG j H j
G
   
1
H i / H 2 (RP n ) HG (CP 1 )
2 / H 2 (CP 1 )
H (CP n )
2
m
1

with r = HG i. By Proposition 6.1.11, H i(a) = b2 , so = 1. Note that 1 =

HG j (H j)1 is a section of 1 . As CP 1 with the complex conjugation is G-
diffeomorphic to S12 (via the stereographic projection of S 2 onto C {} CP 1 ),
Corollary 7.1.17 implies that r1 1 (a) = au, which proves that = 1.
For n , we now define [n] : H (CP n ) HG (CP n ) by [n](a k ) =
2 (a)k . This is a section of and [] is clearly multiplicative. As [n] is the
composition of [] with the morphism HG (RP ) HG (RP n ) induced by the
inclusion, the section = n of is multiplicative and satisfies the requirements of
Proposition 7.1.18. 

Corollary 7.1.19 For n , let : H (CP n ) HG (CP n ) be the section

of Proposition 7.1.18. Then, the correspondence a  (a) provides a GrA[u]-
isomorphism

HG (CP n ).
Z2 [u, a] (a n+1 )

Corollary 7.1.20 For n , the restriction morphismr : HG (CP n ) HG (RP n )

is injective.

Proof Let x H m (CP n ) with x ker r . Write x under the form x = (a k )u r + ltr
(k + r = m), where is given by Corollary 7.1.19 and ltr denotes some polynomial
in the variable u of degree less than r . Then, the equation

0 = r (x) = bk u r +k + ltr +k (7.1.18)

holds in HG (RP n ) = H (RP n )[u]. This first proves that k > 0. Choose x so that
k is minimal. Then, (7.1.18) again implies that bk = 0. Hence, n < and k > n.
As is multiplicative, one has (a k ) = 0 and x = ltr , contradicting the minimality
of k. 
274 7 Equivariant Cohomology

The proof of Corollary 7.1.20 generalizes for conjugation spaces (see Lemma
10.2.8). For n < , Corollary 7.1.20 is a consequence of the equivariant formality
of CP n (see Proposition 7.3.9).
Remark 7.1.21 As an exercise, the reader may develop the analogous of Proposi-
tion 7.1.18 and Corollaries 7.1.19 and 7.1.20 for the G-space X = HP n , where G
acts via the involutions on H defined by (x + i y + j z + kt) = x + i y j z kt
(thus X G CP n ), or (x + i y + j z + kt) = x i y j z kt (thus X G RP n ).
The same work may be done with X = OP 2 with various R-linear involutions on O
so that X G HP 2 , CP 2 or RP 2 .
t

7.2 The General Case

Let  be a topological group. Let p : E B be the universal principal -

bundle constructed by Milnor [144]. The space E is contractible, being the join
of infinitely many copies of  with a convenient topology. An element of E is
represented by a sequence (ti i ) (i N) with (ti )  and i ; two such
sequences (ti i ) and (ti i ) represent the same class in E if ti = ti and i = i
whenever ti
= 0. There is a free right action of  on E given by (ti i ) g = (ti i g).
One defines B = E/ . The quotient map p : E B enjoys local triviality, in
other words is a principal -bundle. These constructions are functorial: a continuous
homomorphism :    induces a continuous map E : E  E, defined
by E(ti i ) = (ti (i )), which descends to a continuous map B : B  B.
Example 7.2.1 Consider the case  = G = {I, }. Then, E G BG is homotopy
equivalent to S RP . This is because the join of a space Y with G S 0 is
homeomorphic to the suspension of Y . In the same way, E S 1 B S 1 is homotopy
equivalent to S CP .
Lemma 7.2.2 Let  be a finite group of odd order. Then H (B) H ( pt).
Proof By Kronecker duality, it is equivalent to prove that H (B) H ( pt). When
 is a discrete group, the principal -bundle p : E B is the universal covering
of B. One has the transfer chain map tr : Cm (B) Cm (E) as in Sect. 4.3.3,
sending a singular simplex : m B to the set of its liftings in E. If the
number of sheets is odd, the composition

tr H p
H (B)
H (E) H (B)

is the identity. Since E is contractible, this proves the lemma. 

When  is a discrete group, the cohomology of B is isomorphic to the coho-
mology H (; Z2 ) of the group  in the sense of [3, 26]. The isomorphism
7.2 The General Case 275

H (B) H (; Z2 ) is proven in e.g. [3, II.2]. The following proposition,

proven in [26, Proposition III.8.3] will be useful.

Proposition 7.2.3 Let  be a discrete group. Let be an inner automorphism of ,

i.e. (g) = g0 gg01 for some g0 . Then H B(a) = a for all a H (B).

Let X be a left -space. The Borel construction X  or homotopy quotient, is the

quotient space 
X  = E  X = (E X )

where is the equivalence relation (z, x) (z, x) for all x X , z E and

. A map p : X  B is then given by p(z, x) = p(z). If  = G = {I, },
this Borel construction coincides with that defined in (7.1.1).
As in Sect. 7.1, one proves the following statements.
(1) The Borel construction is a covariant functor from the category Top to Top,
where Top is the category of -spaces and -equivariant maps. The map p :
X  B coincides with the map X  pt = B induced by the constant
map X pt.
(2) X  and Y have the same homotopy type if X and Y have the same -homotopy
type.
(3) The map p : X  B is a locally trivial fiber bundle with fiber homeomorphic
to X .
(4) If f : Y X is a G-equivariant map, then the diagram

f
Y / X
DD z
DDp p zz
DD zz
D! z| z
B

is commutative.
(5) If the  action on X has a fixed point, then p admits a section. More precisely,
the choice of a point v X  provides a section sv : B X  of p.
(6) If  acts trivially on X , then X  has the homotopy type of B X (see the
proof of Point (3) of Lemma 7.1.3).
(7) The projection E X X  is a -principal bundle induced from the universal
bundle by p.

Example 7.2.4 Let i : 0  denote the inclusion of a closed subgroup 0 of .

We consider the homogeneous -space X = / 0 . Then, the map h : E X

E/ 0 given by h(z, []) = z descends to a homeomorphism h : E  X
E/ 0 . Consider the commutative diagram
276 7 Equivariant Cohomology

/ E0 / E0 / 0 o = / B0

0
ttt
= Ei ttt
zt g
Ei
   t
0 / E / E/ 0 Bi
O
=
  
/ E/  o = / B
E

The two upper lines are 0 -principal bundles. As both E0 and E have vanishing
homotopy groups, the map Ei is a weak homotopy equivalence, and so is g. Hence
X  has the weak homotopy type of B0 . In addition, the map Bi : B0 B is
weakly homotopy equivalent to the locally trivial bundle X  B with fiber X .
More generally, let Y is a -space and consider / 0 Y endowed with the diagonal
-action; then H (/ 0 Y ) H0 (Y ) (see the proof of Theorem 7.4.3).

For a general -space Y , the quotient map q : Y \Y descends to a surjective

map q : Y \Y such that q 1 ([y]) has the weak homotopy type of B y for all
y Y , where  y is the stabilizer of y.
Let (X, Y ) be a -pair, i.e. a -space X with an -invariant subspace Y . The
-equivariant cohomologyH (X, Y ) is the cohomology algebra

H (X, Y ) = H (X  , Y ) and H (X ) = H (X  ) = H (X, ).

In particular, H ( pt) = H (B). The map p : X  B induces an GrA-

homomorphism H p : H (B) H (X ), endowing the latter with a structure of
H ( pt)-algebra.

7.2.5 Changing spaces and groups. Let :    be a continuous homomorphism.

Let X be a -space and X  be a   -space. A continuous map f : X  X satisfying

f (x) = () f (x)

E f
is called equivariant with respect to . The continuous map E  X  E X
then descends to a continuous map f   , : X   X  (depending on ). There is a
commutative diagram
f   ,
X  / X
. (7.2.1)
 
B 
B / B

The map f   , induces a GrA-homomorphism

f  , : H (X ) H (X  ). (7.2.2)
7.2 The General Case 277

By commutativity of the Diagram (7.2.1), one has

f  , (av) = f  , (a)(B) (v) a H (X ) and v H ( pt)

for all a H (X ) and v H ( pt). We say that f  , preserves the module structures
via . More simply, the -space X becomes a   -space via , thus H (X ) becomes a
H (B  )-algebra and f  , is a morphism of H (B  )-algebras. An important case
is given by f = id : X X , Setting id,  = , we get a map : H (X )
H (X ) which is a morphism H (B  )-algebras.

7.2.6 Free actions. Let 0 be a closed normal subgroup of  and let X be a -space.
For x X ,  and 0 , 0 0 , the equation

(0 ) (0 x) = (0 0 1 ) x

shows that the -action on X descends to a (/ 0 )-action on 0 \X . By the functo-

riality of equivariant cohomology (see 7.2.5), we get a map

H/ 0 (0 \X ) H (X ) (7.2.3)

which is a homomorphism of H (B(/ 0 ))-algebras. The following lemma gener-

alizes Lemma 7.1.4. To avoid point-set topology complications, we restrict ourselves
to the smooth action of a Lie group.

Lemma 7.2.7 Let 0 be a compact normal subgroup of a Lie group . Let X be a

smooth -manifold on which 0 acts freely. Then, the map (7.2.3) is an isomorphism
of H (B(/ 0 ))-algebras.

Proof Let Y = E(/ 0 ) / 0 (0 \X ). Consider the commutative diagram

q
E X / E  X
p p .
 q

E(/ 0 ) (0 \X ) / Y

Let a Y represented by ((ti i ), x) in E X . Then,

 
(q p)1 (a) = { (ti i i ), x | i , 0 }.

Therefore,
 
p 1 (a) = q((q p)1 (a)) { (ti i i ), x | i 0 }
E0 , (7.2.4)
 
the last homeomorphism being given by (ti i i ), x  (ti i ).
278 7 Equivariant Cohomology

As 0 acts smoothly and freely on X , the quotient map X 0 \X is a locally

trivial bundle (this follows from the slice theorem: see [12, Theorem 2.2.1]). Hence,
p is homotopy equivalent to a locally trivial bundle, which is numerable ([179, p. 94])
since 0 \X is paracompact. The map q is also a numerable locally trivial bundle
(see (7) on p. 275). Therefore, q p is a fibration (i.e. satisfies the homotopy covering
property for any space: see [179, Theorem 12, p. 95]), and so does p. As 0 \X is
a manifold and E(/ 0 ) is contractible, the space Y admits a numerable covering
{V } such that each inclusion V Y is null-homotopic. As each fiber of p is
contractible by (7.2.4), [42, Theorem 6.3] implies that p is a homotopy equivalence,
which proves the lemma. 

7.2.8 The forgetful homomorphism. Choosing a point E provides a map i :

X X  defined by i (x) = [, x]. As E is path-connected, the homotopy class
of i does not depend on . For instance, we can take = 0 = (1e, 0, 0, . . . ) where
e  is the unit element. Therefore, we get a well defined GrA-homomorphism

: H (X ) H (X ) (7.2.5)

given by = H i for some E. As in (7.1.5), one proves that is functorial.

In fact, using 7.2.5, the homomorphism coincides with the homomorphism id{e},
induced by inclusion of the trivial group {e} into :

= id{e}, : H (X ) H{e}

(X ) = H (X ). (7.2.6)

Indeed, i 0 factors through X X { e} X . Hence, may be seen as a forgetful

homomorphism (one forgets the -action).
A consequence of (7.2.5) is that is functorial for the changing of groups: if
:    is a continuous homomorphism and X a -space, the diagram

/ H  (X )
H (X ) 
JJJ t
JJJ tt
JJ tt
$ yttt
H (X )

is commutative.

7.2.9 Equivariant formality. A -space X is called equivariantly formal if :

H (X ) H (X ) is surjective. For instance, X is equivariantly formal if the -
action is trivial. For relationships with other kind of formal spaces, see [173]. If X
is equivariantly formal, one can choose, as in the proof of Proposition 7.1.6, a GrV-
section : H (X ) H (X ) of . Then is a cohomology extension of the fiber
p
for the fiber bundle X 
B. If X is of finite cohomology type, the Leray-Hirsch
7.2 The General Case 279

Theorem 4.7.17 then gives an map (depending on )

H ( pt) H (X )
H (X ) (7.2.7)

which is an isomorphism of H ( pt)-modules.

7.2.10 Reduced cohomology. Let X be a -space. The reduced equivariant coho-

mology H  (X ) is the H ( pt)-algebra defined by
 
H  (X ) = coker H p : H ( pt) H (X ) (7.2.8)

where p : X pt denotes the constant map to a point (which is -equivariant).

Warning: H (X )
= H (X  ). Examples:
(1) H  ( pt) = 0.
(2) If Y is a space with trivial -action, there is an isomorphism of H ( pt)-algebras
 
H  (Y ) = H (Y ) H ( pt) /(1 H ( pt)) H (Y ) H ( pt). (7.2.9)

(3) If X is equivariantly formal and is of finite cohomology type, one uses (7.2.7)
to provides an isomorphism of H ( pt)-modules between H  (X ) and H (X )
H ( pt). This isomorphism depends on the choice of a section of : H (X )
H (X ) and is not, in-general, an isomorphism of algebras.
Any -equivariant map f : Y X satisfies p f = p, so H  is a contravariant
functor from Top to the category of H ( pt)-algebra.
Let v X  . As for 3.1.14, one has the following diagram.

H ( pt)
III
p III
II
 i 
$
j
H (X  , {v} ) / H (X )
/ H (v) (7.2.10)
 
MMM
MM
MMM
M& 
H  (X )

where the line and the column are exact. This proves that

H  (X ).
H (X  , {v} ) (7.2.11)

Observe that, in (7.2.10), i  coincides with the section sv of p . We see that the choice
of v X  produces a supplementary vector subspace to p (H ( pt)) in H (X ).
280 7 Equivariant Cohomology

A pair (X, A) of -spaces is called equivariantly well cofibrant if it admits a

presentation (u, h) as a well cofibrant pair which is -equivariant, i.e. u(x) = u(x)
and h(x, t) = h(x, t) for all , x X and t I .

Lemma 7.2.11 Let (X, A) be a pair of -spaces which is equivariantly well cofi-
brant. Then (X  , A ) is well cofibrant.

Proof Let (u, h) be a presentation of (X, A) as an equivariantly well cofibrant pair.

Define u : E X I and h : E X I E X by u(z, x) = u(x)
x, t) = (z, h(x, t)). We check that these maps descend to u  : X  I
and h(z,
and h  : X  I X  and that (u  , h  ) is a presentation of (X  , A ) as a well
cofibrant pair. 

Lemma 7.2.12 Let (X, A) be a 1 -equivariantly well cofibrant pair of 1 -spaces.

Let (Y, B) be a 2 -equivariantly well cofibrant pair of 2 -spaces. Then, (X Y, A
Y X B) is a 12 -equivariantly well cofibrant pair of 12 -spaces, where 12 =
1 2 .

Proof One checks that the proof of Lemma 3.1.40 works equivariantly. 

If (X, A) is a pair of -spaces, the quotient space X/A inherits a -action, with
[A] (X/A) , where [A] denotes the set A as a class in X/A. The proof of the
following lemma is the same as that of Lemma 3.1.43.

Lemma 7.2.13 If (X, A) is an equivariantly well cofibrant pair of -spaces, so is

the pair (X/A, A/A).

Example 7.2.14 Let (X, x) be a pointed space. The group G = {I, } acts on X X
by exchanging the coordinates and this action descends to X X . If (X, x) is well
pointed, the proof of Lemma 3.1.40 shows that the pair (X X, X X ) is G-
equivariantly well cofibrant. By Lemma 7.2.13, (X X, x x) is G-equivariantly
well pointed.

The quotient map : (X, A) (X/A, A/A) is a -equivariant map of pairs

which induces  : (X  , A ) ((X/A) , (A/A) ).

Proposition 7.2.15 Let (X, A) be a pair of -spaces which is equivariantly well

cofibrant. Then,

 : H ((X/A) , (A/A) )
H (X  , A )

is an isomorphism.

Proof Let (K , L) = (X/A, A/A). Let (u, h) be a presentation of (X, A) as an

equivariant well cofibrant pair and let (u, be the induced presentation of (K , L).
h)
1 1
Let V = u ([0, 1/2]) and W = u ([0, 1/2]) = (V ). As noticed in the proof
of Lemma 3.1.41, the condition u(h(x)) u(x) implies that h and h restrict to -
equivariant deformation retractions from V to A and from W to L. The tautological
7.2 The General Case 281

homeomorphism from E (V I ) onto (E V ) I descends to a homeomor-

phism (V I ) V I . Using this, h and h descend to a deformation retractions
h  : V I V , and h  : W I W onto A and L, making (X  , A ) and
(K  , L  ) good pairs.
The inclusion (K , L) (K , W ) gives rise to a morphism of exact sequences

H k1 (K  ) / H k1 (L  ) / H k (K  , L  ) / H k (K  ) / H k (L  )

= =
    
H k1 (K  ) / H k1 (W ) / H k (K  , W ) / H k (K  ) / H k (W )

which, by the five lemma, implies that H k (K  , L  )
H k (K  , W ) is an iso-

morphism. The same proof gives the isomorphism H k (X  , A ) H k (X  , V ).
Proposition 7.2.15 then comes from the commutativity of the following diagram
(where the vertical arrows are induced by inclusions)


H (K  , L  ) / H (X  , A )
O O


H (K  , W ) / H (X  , V )

excision excision
  
H (K  L  , W L  ) / H (X  A , V A ).

The bottom horizontal arrow is indeed an isomorphism since : (X A, V A)

(K L , W L) is a -equivariant homeomorphism. 

Corollary 7.2.16 Let (X, A) be a pair of -spaces which is equivariantly well cofi-
brant. If A is non-empty, there is a functorial isomorphism of H ( pt)-algebras

H (X  /A )
H  (X/A).

The hypothesis A
= is necessary since H (X  ) is not isomorphic to H  (X ).

Proof This follows from the following diagram

/ H (X/A)
H ((X/A) , (A/A) ) 

 .

/ H (X /A )
H (X  , A )  
282 7 Equivariant Cohomology

The bijectivities come

from (7.2.11) for the top horizontal arrow, since, as A

= , A/A is a point.
from Lemma 7.2.11 and Proposition 3.1.45 for the bottom horizontal arrow.
from Proposition 7.2.15 for the vertical arrow. 

As in 7.2.5, the reduced equivariant cohomology is functorial for changing groups.

In particular, as in 7.2.8, the inclusion of the trivial group {e} into  provides the
forgetful homomorphism

: H  (X ) H {e}

(X ) H (X )

which is functorial.
As in Lemma 7.1.10, one proves that (H (X )) H (X ) . The following
statement generalizes Proposition 7.1.12.

Proposition 7.2.17 Let X be a -space with X 

= . Suppose that H j (X ) = 0 for
0 j < r . Then, : H r (X ) H r (X ) is an isomorphism.

Proof (Using a spectral sequence). In the E 2 -term of the Serre spectral sequence of
the bundle X X  B, the lines from 1 to n 1 vanish:
H r (X)
0
..
. 0
0
H 0 (B) H 1 (B) H 2 (B)

Therefore, it gives rise to the edge exact sequence:

H p H p
H r +1 (B) Hr +1 (X ).
H r (X )
0 H r (B) Hr (X )
(7.2.12)

The choice of a fixed point v X  provides a section sv : B X  , as seen

in Lemma 7.1.3. Therefore, H p is injective and Hr (X )
H r (X ) is surjective.
Proposition 7.2.17 follows form this, as in the proof of Proposition 7.1.12. 

Remark 7.2.18 When  is discrete, a proof of Proposition 7.2.17 without spectral

sequence is possible, following the pattern of that of Proposition 7.1.12. The role of
RP k is played by Bk , the quotient by  of the join E k  =   (k + 1 times)
(see [144, Sect. 3]). The space Yk is defined to be E k   X . We leave this proof as
an exercise to the reader.
7.3 Localization Theorems and Smith Theory 283

7.3 Localization Theorems and Smith Theory

As in Sect. 7.1, we consider in this section the group G = {I, } of order 2, so

BG RP and HG ( pt) = Z2 [u] with u of degree one. Strong results come
out if we invert u, namely if we tensor the Z2 [u]-modules with the ring of Laurent
polynomials Z2 [u, u 1 ]. For a pair (X, Y ) of G-spaces, we thus define

h G (X, Y ) = Z2 [u, u 1 ] Z2 [u] HG (X, Y ),

with the notation h G (X ) = h G (X, ). Note that Z2 [u, u 1 ] is Z-graded, with

Z2 [u, u 1 ]k = Z2 u k and we use the graded tensor product. Hence, h G (X, Y ) is
a Z-graded Z2 [u, u 1 ]-algebra, with
 j

h kG (X, Y ) = Z2 u i HG (X, Y ) HGk (X, Y ). (7.3.1)
i+ j=k Z

The theorem below is an example of the so called localization theorems. For more
general statements, see e.g. [37, Chap. 3] or [9, Chap. 3].
Theorem 7.3.1 Let X be a finite dimensional G-complex. Then, the inclusion X G
X induces an isomorphism

h G (X )
h G (X G )

of Z-graded Z2 [u, u 1 ]-algebras.

Before proving Theorem 7.3.1, we discuss a few examples.
Example 7.3.2 Suppose, in Theorem 7.3.1, that X is a free G-complex. By Lemma
7.1.4, HG (X ) H (X/G). As X/G is a finite dimensional CW-complex, there
exists an integer m such that u m HG (X ) = 0. As u is invertible in Z2 [u, u 1 ], this
proves that h G (X ) = 0, as predicted by Theorem 7.3.1, since X G = . We see here
that the finite dimensional hypothesis is necessary in Theorem 7.3.1. Indeed, the
free G-complex S = E G satisfies HG (E G) = H (BG) = Z2 [u], so h G (E G) =
Z2 [u, u 1 ].
Example 7.3.3 Consider the G-space S np of Example 7.1.14, i.e. the sphere S n
endowed with a linear G-action with (S np )G S p . We assume that 0 p n. Using
Corollary 7.1.17, there are elements a HGn (S np ) and b HG ((S np )G ) generating
p

respectively HG (S np ) and HG ((S np )G ) as free Z2 [u]-modules and r (a) = bu n p .

Then, as predicted by Theorem 7.3.1, r : h G (S np ) h G ((S np )G ) admits an inverse,
sending b to au pn .
Example 7.3.4 Consider the G-space CP n (n 1), where G acts via the complex
conjugation, with (CP n )G = RP n . By Proposition 7.1.18 and Corollary 7.1.19,
284 7 Equivariant Cohomology

there is a commutative diagram

 r 
Z2 [u, u 1 , a] (a n+1 ) / Z2 [u, u 1 , b] (bn+1 )

 
h G (CP n ) r / h (RP n )
G

with a of degree 2, b of degree 1 and r (a) = bu + b2 . If n < , the correspondence

 i i+1
b  au 1 + a 2 u 3 + a 4 u 7 + = a 2 u 2 1 (7.3.2)
i0
 
extends to a GrA[u]-isomorphism r 1 : Z2 [u, u 1 , b] (bn+1 ) Z2 [u, u 1 , a]
(a n+1 ) which is the inverse of r .
Example 7.3.5 If n = , the right hand member of (7.3.2) is not a polynomial and
no inverse of r may be defined this way. In fact, r (and then r ) is not an isomorphism.
Indeed, the composition of r with the epimorphism Z2 [u, u 1 , b] Z2 sending
both b and u to 1 is the zero map. Of course, CP violates the finite dimensional
hypothesis in Theorem 7.3.1.
Proof of Thorem 7.3.1 The proof is by induction on the dimension of X , which
starts trivially with X = (dimension 1). The induction step reduces to proving
that, if the theorem is true for X , it is then true for Z = X C where C is a finite
family of G-cells. We consider the commutative diagram.

h k1 / h k (Z , X ) / h k (Z ) / h k (X ) / h k+1 (Z , X )
G (X ) G G G G

rX r Z ,X rZ rX r Z ,X
    
h k1 / h k (Z G , X G ) / h k (Z G ) / h k (X G ) / h k+1 (Z G , X G )
G (X )
G
G G G G

The two lines are exact sequences, obtained by tensoring with Z2 [u, u 1 ] the exact
sequence of (Z , X ) for HG (as in the proof of Lemma 4.6.9, we use that a direct
sum of exact sequences is exact and that, over a field, tensoring with a vector space
preserves exactness). If r X is an isomorphism by induction hypothesis, it is enough,
using the five lemma, to prove that r Z ,X is an isomorphism. Note that C is a disjoint
union of free G-cells C f and of isotropic G-cells G i . By excision, one has the
commutative diagram

/ h (C, BdC) / h (C f , BdCf ) h (Ci , BdCi )

h G (Z , X ) G G G

r Z ,X r
 
/ h (Ci , BdCi )
h G (Z G , X G ) G
7.3 Localization Theorems and Smith Theory 285

where r (a, b) = b. It is then enough to prove that h G (C f , BdCf ) = 0. But this

follows from the exact sequence of (C f , BdCf ) for h G and, since C f and BdCf are
free G-space, from Example 7.3.2.
We are now leading toward the Smith inequalities. Let us extend our ground ring
Z2 [u, u 1 ] to the fraction field Z2 (u) of Z2 [u] (this is just a field of characteristic 2,
the grading is lost). For a space X , the total Betti number b(X ) of X is defined by


b(X ) = dim H (X ) N {}.
i=0

Lemma 7.3.6 Let X be a finite dimensional G-complex with b(X ) < . Then, as
a vector space over Z2 (u),

dim Z2 (u) Z2 [u] HG (X ) b(X )

with equality if and only if X is equivariantly formal.

Proof From the transfer exact sequence (7.1.7), we extract the exact sequence

u u
HGk1 (X ) HGk (X ) HGk (X ) HGk+1 (X ).
H k (X )

We deduce that HGk (X ) is generated by u HGk1 (X ) and a number of elements

dim H k (X ), which proves the first assertion. Moreover, dim Z2 (u) Z2 [u] HG (X ) =
b(X ) if and only if : HG (X ) H (X ) is surjective, that is X is equivariantly
formal. 

Proposition 7.3.7 Let X be a finite dimensional G-complex with b(X ) < . Then

b(X G ) b(X ) (7.3.3)

with equality if and only if X is equivariantly formal.

Proof

b(X ) dim Z2 (u) Z2 [u] HG (X ) by Lemma 7.3.6

= dim Z2 (u) Z2 [u] HG (X G ) by Theorem 7.3.1
= b(X G ),

the last equality coming from Lemma 7.3.6, since X G is equivariantly formal. From
Lemma 7.3.6 again, the above inequality is an equality if and only if X is equivariantly
formal. 

Formula (7.3.3) is an example of Smith inequalities, a development of the work

of P. Smith started in 1938 [178]. The following corollary is a classical result in the
theory.
286 7 Equivariant Cohomology

Corollary 7.3.8 Let X be a finite dimensional G-complex. Then,

(1) If H (X ) H ( pt), then H (X G ) H ( pt).
(2) If X has the cohomology of a sphere, so does X G .

Proof If X has the cohomology of a point, it is equivariantly formal and, by Proposi-

tion 7.3.7, b(X G ) = 1 which proves (1). For Point (2), Proposition 7.3.7 implies that
b(X G ) 2. Statement (2) is true if b(X G ) = 2 or if b(X G ) = 0 (since = S 1 ). It
remains to prove that b(X G ) = 1 is impossible if H (X ) H (S n ).
If b(X G ) = 1, then X is not equivariantly formal. Using Exact sequence (7.1.7),
this implies that HG (X ) H (RP n ). As in Example 7.3.2, we deduce that h G (X ) =
0, contradicting Theorem 7.3.1 (h G (X G ) = Z2 [u, u 1 ] if b(X G ) = 1). 

Here is another consequence of Theorem 7.3.1.

Proposition 7.3.9 Let X be a finite dimensional G-complex. Then, the following

statements are equivalent.
(1) X is equivariantly formal.
(2) r : HG (X ) HG (X G ) is injective.

Proof From Corollary 7.1.8, we already know that (2) implies (1). For the converse,
suppose that X is equivariantly formal. Then HG (X ) is a free Z2 [u]-module by
Proposition 7.1.6 and thus j : HG (X ) Z2 [u, u 1 ] Z2 [u] HG (X ) = h G (X ) is
injective. Therefore, in the commutative diagram

HG (X ) r / H (X G )
G
j jG
 
r / h (X G )
h G (X ) G

the left vertical arrow is injective. When X is finite dimensional G-complex, the
bottom arrow is an isomorphism by Theorem 7.3.1. Hence j is injective. 

We shall now prove a localization theorem analogous to Theorem 7.3.1 for S 1 -

spaces. Since we are working with Z2 -cohomology, an important role is played by
the subgroup {1} = S 0 of S 1 . We also need the notion of a -CW-complex for  a
topological group. If 0 is a closed subgroup of , the -space / 0 D n is called
a -cell of dimension n (of type 0 ), with boundary / 0 S n1 (the group  acts
on the left on / 0 and trivially on D n ). One can attach a -cell to a -space Y via
a G-equivariant map : / 0 S n1 Y . A -CW-structure on a -space X is
a filtration

= X 1 X 0 X 1 X = Xn (7.3.4)
nN
7.3 Localization Theorems and Smith Theory 287

by -subspaces, such that, for each n, the space X n (the n-skeleton) is -homeo-
morphic to a -space obtained from X n1 by attachment of a family -cells of
dimension n (of various type). A -space endowed with a -CW-structure is a -
CW-complex (or just a -complex). The topology of X is the weak topology with
respect to the filtration (7.3.4).
If X is a -complex, then X/  admits a CW-structure so that the projection
X X/  is cellular. For  = G of order 2, the above definition is easily made
equivalent to that of p. 260 (compare also [37, pp. 101102]). If  is a compact Lie
group acting smoothly on a smooth manifold X , then X admits a -CW-structure
(see [107]).
The Milnor classifying space B S 1 for principal S 1 -bundles is homotopy equivalent
to CP . Then, by Proposition 6.1.5, HS1 ( pt) = Z2 [v] with v of degree 2. For a pair
(X, Y ) of S 1 -spaces, we thus define

h S 1 (X, Y ) = Z2 [v, v 1 ] Z2 [v] HS1 (X, Y ),

with the notation h S 1 (X ) = h S 1 (X, ). As in (7.3.1), h S 1 (X, Y ) is a Z-graded

Z2 [v, v 1 ]-algebra.
1 0
Theorem 7.3.10 Let X be a finite dimensional S 1 -complex such that X S = X S .
1
Then, the inclusion X S X induces an isomorphism

h S 1 (X )
h S 1 (X S )
1

of Z-graded Z2 [v, v 1 ]-algebras.

1 0
The hypothesis X S = X S is necessary in the above localization theorem
Theorem 7.3.10. For example, let X = S 1 with S 1 -action g z = g 2 z. Then
X S 1 B S 0 RP by Example 7.2.4, so h S 1 (X ) Z2 [v, v 1 , u] while X S = .
1

Proof The proof follows the plan of that of Theorem 7.3.1, by induction on the
skeleton of X , starting trivially with the (1)-skeleton which is the empty set. The
induction step reduces to proving that, if the theorem is true for X , it is then true
for Z = X C where C is a family of S 1 -cells. As for Theorem 7.3.1, this even-
tually reduces to proving that h G (C, BdC) = 0 when C is not an isotropy cell. As
1 0
X S = X S , the isotropy group  of C is then a finite group of odd order. The pair
(C, BdC) is of the form (S 1 /  D n , S 1 /  S n1 ) and, as seen in Example 7.2.4,
C S 1 B and (BdC) S 1 B S n1 . By Lemma 7.2.2, HS1 (C) = H ( pt) and
HS1 (BdC) = H (S n1 ). In particular, the multiplication by u is the zero map and
thus h S 1 (C) = h S 1 (BdC) = 0. From the exact sequence of (C, BdC) for h G , it
follows that h G (C, BdC) = 0. 
1 0
The Smith theory for S 1 -complexes with X S = X S is very similar to that of
Let Z2 (v) be the fraction field of Z2 [v].
S 0 -spaces.
288 7 Equivariant Cohomology

Lemma 7.3.11 Let X be a finite dimensional S 1 -complex with b(X ) < and
1 0
X S = X S . Then, as a vector space over Z2 (v),

dim Z2 (v) Z2 [v] HS1 (X ) b(X )

with equality if and only if X is equivariantly formal.

Proof The proof is the same as that of Lemma 7.3.6. The transfer exact sequence is
replaced by the Gysin exact sequence of the S 1 -bundle X S X S 1 which, as
indicated in (7) p. 275, is induced from the universal bundle by p : X S 1 B S 1
CP . Therefore, this Gysin sequence looks like

v v
HSk1 HSk+1
1 (X ) 1 (X ) HSk1 (X ) HSk+2
H k+1 (X ) 1 (X )

and permits us the same arguments as for Lemma 7.3.6. 

The proofs of [Link].14 below are then the same as those of [Link].9,
replacing Theorem 7.3.1 by Theorem 7.3.10.

Proposition 7.3.12 Let X be a finite dimensional S 1 -complex with b(X ) < and
1 0
X S = X S . Then
1
b(X S ) b(X ) (7.3.5)

with equality if and only if X is equivariantly formal.

1 0
Corollary 7.3.13 Let X be a finite dimensional S 1 -complex with X S = X S . Then,
(1) If H (X ) H ( pt), then H (X S ) H ( pt).
1

1
(2) If X has the cohomology of a sphere, so does X S .
1 0
Proposition 7.3.14 Let X be finite dimensional S 1 -complex with X S = X S . Then,
the following statements are equivalent.
(1) X is equivariantly formal.
(2) r : HS1 (X ) HS1 (X S ) is injective.
1

7.4 Equivariant Cross Products and Knneth Theorems

Let 1 and 2 be two topological groups; we set 12 = 1 2 . Let X be a

1 -space and Y be a 2 -space. Then X Y is a 12 -space by the product action
(1 , 2 )(x, y) = (1 x, 2 y). The projections P1 : X Y X and P2 : X Y Y
are equivariant with respect to the projection homomorphisms 12 i . Passing
7.4 Equivariant Cross Products and Knneth Theorems 289

to the Borel construction gives a map

P
(X Y )12
X 1 Y2 (7.4.1)

The map P is a homotopy equivalence, being induced by the homotopy equivalence

P : E12 (X Y ) (E1 X ) (E2 Y ) (7.4.2)

given by
   
P (ti (ai , bi ), (x, y) = (ti ai , x), (ti bi , y) ,

where (ti )  , (ai , bi ) 12 and (x, y) X Y . The case X = Y = pt

provides a homotopy equivalence P0 : B(12 ) B1 B2 and a commutative
diagram
P /
(X Y )12 X 1 Y2
. (7.4.3)
 
B(12 )
P0
/ B1 B2

The cross product H (B1 ) H (B2 )
H (B1 B2 ) post-composed with

H P0 gives a ring homomorphism

h : H1 ( pt) H2 ( pt)

H12 ( pt).

Note that, if B1 or B2 is of finite cohomology type, the Knneth theorem implies
that h is an isomorphism. The homotopy equivalence (7.4.1) together with (7.4.3)
and the Knneth theorem gives the following lemma.

Lemma 7.4.1 The composed map

/ H (X 1 Y2 ) H P / H (X Y ).
12 : H1 (X ) H2 (Y )
12

is an homomorphism of algebras. The (H1 ( pt) H2 ( pt))-module structure on

H1 (X ) H2 (Y ) and the H12 ( pt)-module structure on H12 (X Y ) are preserved
via h. If Y2 is of finite cohomology type, then 12 is an isomorphism.

Example 7.4.2 Let 1 = 2 = G = {1}. We let 1 act on the linear sphere

X = S0m with X 1 = { 1 }, and let  act on Y = S n with Y 2 = { 2 } (see
2 0
Example 7.1.16). Set H1 ( pt) = Z2 [u 1 ] and H2 ( pt) = Z2 [u 2 ] (u i of degree 1).

As seen in Example 7.1.16, H1 (X ) and H2 (Y ) admit the following presentations

H1 (X ) Z2 [u 1 , A1 , B1 ] (A1 + B1 + u m
1 , A1 + u 1 A1 )
2 m
290 7 Equivariant Cohomology

and

H2 (Y ) Z2 [u 2 , A2 , B2 ] (A2 + B2 + u n2 , A22 + u n2 A2 ),

where A1 , B1 are of degree m and A2 , B2 are of degree n. To shorten the formulae,

we also denote by A1 the element A1 12 1 H12 (X Y ), by A2 the element
1 12 A2 H12 (X Y ), etc. By Lemma 7.4.1. we thus get the presentation

H12 (X Y ) Z2 [u 1 , u 2 , A1 , B1 , A2 , B2 ] I, (7.4.4)

where I is the ideal generated by

A1 + B1 + u m
1 , A1 + u 1 A1 , A2 + B2 + u 2 and A2 + u 2 A2 .
2 m n 2 n

One can of course eliminate the Bi s and get the shorter presentation

H12 (X Y ) Z2 [u 1 , u 2 , A1 , A2 ] (A21 + u m
1 A1 , A2 + u 2 A2 ).
2 n

The commutative diagram

12
H1 (X ) H2 (Y ) / H (X Y )
12

r X rY r (7.4.5)
 
12
H1 (X ) H2 (Y 2 )
1 / H ((X Y )12 )
12

permits us to compute the image under r of the various classes of H12 (X Y ). Set

H1 (X 1 ) = Z2 [u 1 ]
1
Z2 [u 1 ]+
1
and H2 (Y 2 ) = Z2 [u 2 ]
2
Z2 [u 2 ]+
2
.

Denote the four points of (X Y )12 = X 1 Y 2 by = (1 , 2 ),

+ =
( , + ), etc. With the notation R = Z2 [u 1 , u 2 ], one has
1 2

H12 ((X Y )12 ) R R + R + R ++ (7.4.6)

One has

r (A1 ) = r X (A1 ) 12 rY (1) = u m

1 + 12 (1 + 1 + ) = u 1 + + u 1 ++ .
1 2 2 m m

Hence, the coordinates of r (A1 ) using (7.4.6) are (0, u m

1 , 0, u 1 ). Similar computa-
m

tions provide the following table.

We now concentrate our interest to the case where 1 = 2 =  and see X

Y as a -space using the diagonal homomorphism  :   . We get a
7.4 Equivariant Cross Products and Knneth Theorems 291

x Coord. of r (x) in (7.2.6)

1 1 1 1 1
ui ui ui ui ui
A1 0 um 1 0 um1
A2 0 0 u2 n u n2
A1 A2 0 0 0 um
1 u2
n

B1 um1 0 u m
1 0
B2 u n2 u n2 0 0
B1 B2 u m n
1 u2 0 0 0

homomorphism  : H (X Y ) H (X Y ). The composed map

 H P 
H (X ) H (Y ) / H (X  Y ) / H
(X Y ) / H (X Y )
2 


(7.4.7)

is called the equivariant cross product. For X = Y = pt, one has the commutative
diagram


H ( pt) H ( pt) / H ( pt)


. (7.4.8)
 
 / H (B)
H (B) H (B)

Indeed, one has

B / P / B B
B B( ) 5 (7.4.9)
 B

and H  B (a b) = a  b by (4.6.5).
The equivariant cross product  will be useful in Sect. 8.3 but one may wish
to get some Knneth theorem. As this is not even the case for X = Y = pt, some
adaptation is needed. Lemma 7.4.1 together with diagram (7.4.8) implies that

(w a)  b = a  (w b) = w (a  b). (7.4.10)

for all a H (X ), b H (Y ) and w H ( pt). Therefore,  descend to the

strong equivariant cross product

 : H (X ) H ( pt) H (Y ) H (X Y ).

292 7 Equivariant Cohomology

The tensor product H (X ) H ( pt) H (Y ) still carries an H ( pt)-action, defined by

w (a b) = (w a) b = a (w b). Lemma 7.4.1 together with (7.4.10) implies
 is a morphism of H ( pt)-algebras.
that

Theorem 7.4.3 (Equivariant Knneth theorem) Let  be a topological group such

that B0 is of finite cohomology type for any closed subgroup 0 of . Let X and Y
be -spaces, where X is a finite dimensional -CW-complex. Suppose that Y is of
finite cohomology type and is equivariantly formal. Then, the strong equivariant

cross product is an isomorphism of H ( pt)-algebras.

Proof As  is a morphism of H ( pt)-algebra, it suffices to prove that it is a GrV-

isomorphism. We follow the idea of the proof of the ordinary Knneth Theorem 4.6.7,
fixing the -space Y and comparing the equivariant cohomology theories

h (X, A) = H (X, A) H ( pt) H (Y ) and k (X, A) = H (X Y, A Y )

defined for a -pair (X, A). The definition of the strong equivariant cross product
extends to pairs and we get a morphism of H ( pt)-algebras

 : h (X, A) k (X, A)

One gets a commutative diagram

h (X ) / h (A) / h +1 (X, A) / h +1 (X ) / h +1 (A)







    

h (X ) / k (A) / k +1 (X, A) / k +1 (X ) / k +1 (A)
(7.4.11)
where the lines are exact. That the square diagram with the s commutes comes
from the definition of  , using the commutativity of Diagram (4.6.13).
The theorem is proven by induction on the dimension of X . If X is 0-dimensional,
it is a disjoint union of homogeneous -spaces. As the disjoint union axiom holds for
our theories, the induction starts by proving the theorem for X = / 0 , where 0
is a closed subgroup of . As Y is -equivariantly formal, it is also 0 -equivariantly
formal and one has

h (X ) = H (/ 0 ) H ( pt) H (Y ) 

H (B0 ) H (B) H (B) H (Y )
H (B0 ) H (Y )
H0 (Y ).

On the other hand, consider the map : E  ( Y ) E  Y given by

(, y)) = (z, 1 y).

(z,
7.4 Equivariant Cross Products and Knneth Theorems 293

It satisfies (z,
(, y)) = (z,
(0 , y)) = (z0 , 01 1 y);
(, y)) and (z,
it thus descends to a map

: E   (/ 0 Y )
E 0 Y

which is a homeomorphism: its inverse is induced by the map (z, y) = (z, ([e], y)),
where e  is the unit element. Hence, k (X ) is also isomorphic, as an H ( pt)-
algebra, to H0 (Y ). It remains to show that
 is a GrV-isomorphism. As Y and B0
are both of finite cohomology type, the graded vector space H0 (Y ) H (B0 )
H (Y ) is finite dimensional in each degree. Therefore, it suffices to prove that
 is
surjective.
If Z is a -space, we denote by i : Z Z  the inclusion i(z) = [(1e, 0, ...), z]
(it induces the forgetful homomorphism H i = : H (Z ) H (Z )). One has a
commutative diagram

/ 0 Y o
s
Y
i i
 

E   (/ 0 Y ) o E 0 Y

where s is the slice inclusion s(y) = ([e], y]). We thus get a commutative diagram

/ H (X Y ) H s / H (Y )
H (/ 0 ) H (Y )
O O O


H
H (/ 0 )  H (Y ) / H (X Y ) / H (Y )
 0

Let B be a GrV-basis of H (Y ). Let : H k (Y ) Hk (Y ) be a section of . For

b B, one has

H s((1) ((b)) = H s(1

 b) = b,

the last equality coming from Lemma 4.7.2. Therefore, H s () is surjective

and the formula

(a)
= H (1
 (a))

defines a section : H (Y ) H0 (Y ) of . The Leray-Hirsch theorem then implies

that H0 (Y ) is generated, as a H ( pt)-module, by (B).
Hence,  is surjective.
The induction step reduces to proving that, if the theorem is true for A, it is then true
for X = AC where C is a family of -cells. By the five lemma in Diagram (7.4.11),
it suffices to prove that  : h (X, A) k (X, A) is an isomorphism. By excision
294 7 Equivariant Cohomology

and the disjoint union axiom one can restrict ourselves to the case of a pair (X 0 , A0 ) =
/ 0 (D n , S n1 ) (a -cell). By the five lemma in Diagram (7.4.11) for the pair
(X 0 , A0 ), it suffices to prove the theorem for X 0 and for A0 . The former is covered by
the 0-dimensional case (sinceX 0 is -homotopy equivalent to / 0 ) and the latter
is (n 1)-dimensional, thus covered by the induction hypothesis. 
Remark 7.4.4 If  = {e}, Theorem 7.4.3 reduces to the ordinary Knneth Theo-
rem 4.6.7. Therefore, the hypotheses that Y is of finite cohomology type is essential.
Theorem 7.4.3 is also wrong if Y is not equivariantly formal. For example, set
 = {1}, X = S 1 and Y = S 2 , with the antipodal involution. These are free
-spaces and, by Lemma 7.1.4,

H (X ) H (S 1 / 1) Z2 [u]/(u 2 ) and H (Y ) H (S 2 / 1) Z2 [u]/(u 3 ).

Moreover, H ( pt) = H (RP ) Z2 [u] and, using Lemma 7.1.4 again together
with Proposition 4.3.10, the Z[u]-morphisms H ( pt) H (X ) and H ( pt)
H (Y ) are surjective. Therefore,

H (X ) H ( pt) H (Y ) Z2 [u]/(u 2 ) Z2 [u] Z2 [u]/(u 3 ) Z2 [u]/(u 2 ).

In particular, H (X ) H ( pt) H (Y ) vanishes in degree 3, while H3 (X Y ) = Z2 .

Indeed H (X Y ) H ((X Y )/ 1) and (X Y )/ 1 is a closed manifold of
dimension 3.
The hypothesis that B0 is of finite cohomology type is fulfilled if 0 is a compact
Lie group. Note that it is only used in the proof for the stabilizers of points of X . For
other kind of equivariant Knneth theorems (see [176]).
Example 7.4.5 Consider the diagonal action of the group G = {1} on the product
of linear spheres S0m S0n . Set HG ( pt) = Z2 [u], with u of degree 1. By Example 7.4.2
and Theorem 7.4.3, one has

HG (S0m S0n ) Z2 [u, A1 , B1 , A2 , B2 ] I,

where I is the ideal generated by

A1 + B1 + u m , A21 + u m A1 , A2 + B2 + u n and A22 + u n A2 .

Using the notations of Example 7.4.2 for the fixed points, one has

HG ((S0m S0n )G ) Z2 [u] Z2 [u] + Z2 [u] + Z2 [u] ++ (7.4.12)

and one has the following table for r : HG (S0m S0n ) HG ((S0m S0n )G )
For a generalization of this example, see Proposition 10.3.5.
We now define the equivariant reduced cross product, related to the equivariant
cohomology of a smash product. Let X be a 1 -space and Y be a 2 -space, pointed
7.4 Equivariant Cross Products and Knneth Theorems 295

x coord. of r (x) in (7.4.12)

1 1 1 1 1
u u u u u
A1 0 um 0 um
A2 0 0 un un
A1 A2 0 0 0 u m+n

B1 um 0 um 0
B2 un un 0 0
B1 B2 u m+n 0 0 0

by x X 1 and y Y 2 . Then, X Y is a 12 -invariant subspace of X Y . Consider

the space

Y2 = (X 1 {y}2 ) ({x}1 Y2 ) X 1 Y2 .

X 1

If the pairs (X, {x}) and (Y, {y}) are equivariant well cofibrant pairs, we say that
(X, x) and (Y, y) are equivariantly well pointed.
Lemma 7.4.6 Let (X, x) be an equivariantly well pointed 1 -space and (Y, y) be
an equivariantly well pointed 2 -space. Then, the map P : (X Y )12 X 1 Y2
Y2 and induces an isomorphism
of (7.4.1) sends (X Y )12 onto X 1

H P : H (X 1 H12 (X Y )
Y2 )

Proof That P((X Y )12 ) = X 1 Y2 follows directly from the definition of P,
using (7.4.2). This gives a commutative diagram

{x}1 {y}2 / {x}1 Y2

hQQQ o7
QQQP P oooo
QQ oo
h.e. QQ ooo h.e.
({x} {y})12 / ({x} Y )12

 
(X {y})12 / (X Y )12
n OOO
P nnnn OOOP
nn OOO
 vnnn h.e. O' 
X 1 {y}2 / X Y2
1

(7.4.13)
where the unlabeled arrows are inclusions and h.e. means homotopy equivalence.
Our hypotheses and Lemma 7.2.11 imply that pairs like ((X {y})12 , ({x}{y})12 ),
etc., are good. Hence, the hypotheses of Proposition 3.1.54 to get Mayer-Vietoris
sequences are fulfilled. We thus get a morphism from the Mayer-Vietoris sequence
for the outer square of (7.4.13) to that of the inner square, and the proposition follows
from the five-lemma. 
296 7 Equivariant Cohomology

Remark 7.4.7 The map P : (X Y )12 X 1 Y2 of Lemma 7.4.6 is actually a

weak homotopy equivalence, since the squares in (7.4.13) are homotopy co-Cartesian
diagrams (see [38, Prop. 5.3.3]).

As X Y is a 12 -invariant subspace of X Y , the wedge product X Y inherits

a 12 -action.

Lemma 7.4.8 Let (X, x) be an equivariantly well pointed 1 -space and (Y, y) be
an equivariantly well pointed 2 -space. Then, there is a natural isomorphism

H (X 1 Y2 , X 1 H 12 (X Y )).
Y2 )

Proof By Lemma 7.4.6, the map P produces a morphism from  the cohomology
sequence
 of the pair (X 1 Y2 , X 1
Y2 ) to that of the pair (X Y )12 , (X
Y )12 . By Lemma 7.4.6 again and the fact that the map P of (7.4.1) is a homotopy
equivalence, the five lemma implies that

H P : H (X 1 Y2 , X 1
Y2 ) H ((X Y )12 , (X Y )12

is an isomorphism. By Lemma 7.2.12, the pair (X Y, X Y ) is 12 -equivariantly

well cofibrant. As X Y ) is not empty, Corollary
 7.2.16 provides a natural isomor-
phism between H ((X Y )12 , (X Y )12 and H 12 (X Y ).

Using the isomorphism of Lemma 7.4.8 as well as those of (7.2.11). one constructs
the commutative diagram

/ H (X ) H (Y )
H (X 1 , {x}1 ) H (Y2 , {y}2 ) 1 2


(7.4.14)
12
 
/ H (X Y )
H (X 1 Y2 , X 1 Y
2 ) 12

which defines the equivariant reduced cross product 12 . The relative cross product
(left vertical arrow) is indeed defined as in (4.6.6), since, as (Y, y) is equivariantly
well pointed, the couple (Y2 , {y}2 ) is a good pair by Lemma 7.2.11.
In the case where 1 = 2 = , one can see X Y as a -space via the diagonal
homomorphism  :   . Composing 12 with  : H (X Y )

H  (X Y ), we get the equivariant reduced cross product


H  (X ) H  (Y ) H  (X Y ). (7.4.15)

Lemma 7.4.9 Let (X, x) and (Y, y) be equivariantly well pointed -spaces. Then,
(1) there is an equivariant reduced cross product
7.4 Equivariant Cross Products and Knneth Theorems 297


H  (X ) H  (Y ) H  (X Y )

which is a bilinear map.

(2) the diagram


H  (X ) H  (Y ) / H (X Y )


(7.4.16)
 


H (X ) H (Y )
/ H (X Y )

is commutative, where is the forgetful homomorphism.

(3) the hypotheses on X , Y are inherited by X  , Y  and there is a commutative
diagram



H  (X ) H  (Y ) / H (X Y )


r r r
 


H  (X  ) H  (Y  ) / H (X  Y  )

O O

 


[ H (X  ) H (B)] [ H (Y  ) H (B)] / H (X  Y  ) H (B)
(7.4.17)

Proof The equivariant reduced cross product of (1) is obtained by post-composing

12 of (7.4.14) (with 1 = 2 = ) with  : H (X Y ) H (X Y ).

 

Let :   is a continuous homomorphism. Then (X, x) and (Y, y) are
  -equivariantly well cofibrant. Our constructions are natural enough so that there is
a commutative diagram



H  (X ) H  (Y ) / H (X Y )


. (7.4.18)
 


H  (X ) H  (Y )
/ H  (X Y )


For   = {I }, the homomorphism coincides with the forgetful homomorphism

(see 7.2.8), which proves (2).
To prove (3), we note that the upper square of (7.4.16) commutes by obvious
naturality of the equivariant reduced cross product with respect to equivariant maps.
The commutativity of the lower square is obtained using the considerations of (7.4.8)
and (7.4.9). 
298 7 Equivariant Cohomology

Example 7.4.10 Let (Z , z) be a well pointed space, considered with the trivial action
of G = {I, }. Then, H G (Z ) H (Z )[u] and the bottom square in (7.4.17) becomes

G

H G (Z ) H G (Z ) / H (Z Z )
G
O O
,
 [u]


H (Z )[u] H (Z )[u] / H (Z Z )[u]

where, for a, b H (Z ),
[u] is defined by

au m
[u] bu n = (a
b) u m+n .

7.5 Equivariant Bundles and Euler Classes

Although we are mostly interested in equivariant vector bundles, passing through

equivariant principal bundles is easier and more powerful. Let A be a topologi-
cal group. A principal A-bundle over a space Y consists of a continuous map
p : P Y , a continuous right action of A on P such that p(u) = p(u) for all
u P and all A; in addition, the following local triviality should hold: for each
x X there is a neighbourhood U of x and a homeomorphism : U A p 1 (U )
such that p (x, ) = x and (x, ) = (x, ). In consequence, A acts freely
on P and transitively on each fiber. Also, p is a surjective open map, descending to a

homeomorphism P/A X (use [44, Sect. I Chap. VI]). Two principal A-bundles
p p
= (P X ) and = ( P
X ) are isomorphic if there exists an A-equivariant
homeomorphism h : P P such that p h = p.
Let  be a topological group and let X be a (left) -space. An A-principal bundle
p
: P X is called a -equivariant principal A-bundle if it is given a left action
 P P commuting with the free right action of A and such that the projection p
is -equivariant (a more general setting is considered in e.g. [37, 126, 139]). Two -
p p
equivariant principal A-bundles = (P X ) and  = ( P X ) are isomorphic if
there exists an (, A)-equivariant homeomorphism h : P P such that p h = p.

Example 7.5.1 Let p : P pt be a -equivariant principal A-bundle over a point.

The A-action on P is free and transitive. Hence, choosing a point s P provides a
continuous map :  A by the equation s = s(). For ,  , one has

s(  ) = (  )s = (  s) = (s)(  ) = (s())(  ) = s(()(  )),

which proves that is a homomorphism. Another point s P is of the form s = s

for some A. The map obtained from s is related to by
7.5 Equivariant Bundles and Euler Classes 299

s()
= s ()
= s = s = s()

and hence ()
= 1 (). If p : P pt is another -equivariant princi-
pal A-bundle and if h : P P is a (, A)-equivariant homeomorphism, then
h(s) = h(s)(). This provides a map from the isomorphism classes of -
equivariant principal A-bundles over a point and the set hom(, A)/A of the conju-
gation classes of continuous homomorphisms from  to A. This map is a bijection.
A homomorphism :  A is realized by the bundle A pt with the -action
= () (hence, if e is the unit element of A, one has indeed e = e()).
This proves the surjectivity. The proof of the injectivity is left to the reader.
p
Let : P
X be a -equivariant principal A-bundle. Being -equivariant, the
map p induces a map p : P X  . Let i : X X  be an inclusion as in (7.2.8).

Lemma 7.5.2 The map p : P X  is a principal A-bundle, denoted by  .

Moreover, is isomorphic to the induced principal A-bundle i  .

Example 7.5.3 Let be a -equivariant principal A-bundle over a point, correspond-

ing to [] hom(, A)/A (see Example 7.5.1). It is then isomorphic to A pt
with the -action = (). Then  is the principal A-bundle over B induced
by the map B : B B A. Indeed, the map f : E  A E A given by
f ([(ti i ), ]) = (ti (i )) is A-equivariant and covers the map B .

Before proving Lemma 7.5.2, let us recall the standard local cross-sections for
the Milnor construction of the universal -bundle p : E B. For i N, let
(E)i = {(t j j ) E | ti
= 0} and let (B)i = p((E)
i ). There is a cross-
section si of p over (B)i sending b (B)i to the unique element in (t j  j )

p 1 (b) with i = 1. If Z is a -space, the map i : (B)i Z (E)i  Z
given by i (b, u) = [si (b), u] is a homeomorphism: its inverse is induced by the
correspondence [(t j j ), u]  ( p(t
j j ), i z).
Proof of Lemma 7.5.2 The right A-action on P = E  P is defined by
[z, u] = [z, u]. For i N, one has the commutative diagram

i
(B)i P / (E)i  P

id p p .
 i

(B)i X / (E)i  X

The upper homeomorphism is A-equivariant and id p is a principal A-bundle. As

{(E)i  X }iN is an open covering of X  , the map p admits local trivializations
of a principal A-bundle.
We have proven that p is a principal A-bundle. If z E, let i z : X X 
be the inclusion defined in (7.2.8). Then, iz : P P is an A-equivariant map
covering i z , inducing an isomorphism of principal A-bundles i z  .
300 7 Equivariant Cohomology

Remark 7.5.4 When A is abelian, the last assertion of Lemma 7.5.2 may be strength-
ened: a principal A-bundle over a -space X admits a structure of a -equivariant
bundle if and only if it is induced from a principal A-bundle over X  (see [127]).

The construction   enjoys some functorial properties. Let :    be

a continuous homomorphism between topological groups. Let X  be a   -space, X
a -space and let f : X  X be a continuous map which is   -equivariant with
respect to . Recall from (7.2.1) that f induces a map f   , : X   X  . If is
a -equivariant principal A-bundle over X , then f is a   -equivariant principal
A-bundle over X  .

Lemma 7.5.5 ( f ) f  ,  .

Proof The map f is covered by a   -equivariant map of principal A-bundle f :

P( f ) P, where P( f ) denotes the total space of f . By functoriality of the
Borel construction (see 7.2.5), there is a commutative diagram

f  , =
P( f )  / P o / P( )

 f   ,  
/ X o = / X
X  

Thanks to the description of the A-actions (see the proof of Lemma 7.5.2), the
map f  , A-equivariant. Hence, f  , factor through an isomorphism ( f )
f  ,  . 

For another functoriality of  let : A A be a continuous homomorphism

between topological groups. It makes A a left A-space ( = () ). If : P
X is a -equivariant A-principal bundle, P A A is, in an obvious way, the total
space of a -equivariant A -principal bundle . The tautological homeomorphism
E  (P A A ) (E  P) A A gives an isomorphism of A -principal
bundles

( )  . (7.5.1)

By a -equivariant K-vector bundle over X (K = R or C), we mean a -

equivariant map p : E = E() X which is a K-vector bundle, such that, for each
 and each x X , the map y  y is a K-linear map from p 1 (x) to p 1 (x).
The tangent bundle to a smooth -manifold is an example for K = R.
It is convenient here to see a K-vector bundle of rank r as associated to a principal
G L(Kr )-bundle, the bundle Fra() of frames of . Its total space is

Fra() = { : Kr E() | is a K linear isomorphism onto some fiber of }.

7.5 Equivariant Bundles and Euler Classes 301

with the map pFra : Fra() X given by pFra () = p (0). The right G L(Kr )-
action on Fra() is by precomposition (we use the same notation for the bundle Fra()
and for its total space). The evaluation map sending [, t] Fra() G L(Kr ) Kr to
(t) E() defines an isomorphism of K-vector bundles

Fra() G L(Kr ) Kr
E(). (7.5.2)

For more details and developments, see 9.1.9 If is a -equivariant K-vector bundle,
then  acts on Fra() by ()(t) = (t). Hence, Fra() is a -equivariant principal
G L(Kr )-bundle and (7.5.2) is an isomorphism of -equivariant K-vector bundles.
The tautological homeomorphism

E  (Fra() G L(Kr ) Kr ) (E  Fra()) G L(Kr ) Kr

implies that

E() Fra() G L(Kr ) Kr . (7.5.3)

Using Lemma 7.5.2, this proves the following lemma.

Lemma 7.5.6 Let = ( p : E() X ) be a -equivariant K-vector bundle of

rank r . Then, the map p : E() X  is a K-vector bundle of rank r , denoted by
 . Moreover, is isomorphic to the induced vector bundle i  .

Let :    and f : X  X as for Lemma 7.5.5. If is a -equivariant

K-vector bundle of rank r over X , then f is a   -equivariant K-vector bundle over
X  of the same rank. One has an isomorphism of   -equivariant G L(Kr )-principal
bundles f  , Fra() Fra( f ). Therefore, Lemma 7.5.5 gives an isomorphism of
K-vector bundles

( f ) f  ,  . (7.5.4)

The correspondence   commutes with some operations on vector bundles,

like the Whitney sum or the tensor product. We first define these operations in the
category of -equivariant K-vector bundles. Let (respectively  ) be two such
bundles over a -space X , of ranks r (respectively r  ). Set F = Fra(), F  = Fra(  ),

G = G L(Kr ) and G  = G L(Kr ). The diagonal inclusion  X : X X X is
-equivariant with respect to the diagonal homomorphism  :   . Hence,
X (F F  ) is a -equivariant principal (G G  )-bundle. The linear (G G  )-

action on Kr Kr given by (R, R  ) (v, v  ) = (Rv, R  v  ) defines a continuous

homomorphism : G G  G L(Kr Kr ). This permits us to define the

Whitney sum

 = 
 X (F F ) G L(Kr Kr  ) (K K )
r r
302 7 Equivariant Cohomology

as a -equivariant K-vector space. The -equivariant tensor product is defined

accordingly

 = 
 X (F F ) G L(Kr Kr  ) (K K ),
r r
(7.5.5)


using the homomorphism : G G  G L(Kr Kr ) induced by the unique

linear action of G G on Kr Kr satisfying (R, R  ) (v v  ) = (Rv R  v  ).
Lemma 7.5.7 (  )   and (  )   .
Proof One has
 
(  ) = X (F F  ) G L(Kr Kr  ) (Kr Kr ) 

r

 X (F F )  G L(Kr Kr  ) (K K ) by (7.5.3)
r

 r
 X (F F )  G L(Kr Kr  ) (K K ) by (7.5.1)
r

while

  = 
 X  (F F ) G L(Kr Kr  ) (K K ).
r r

Therefore, it is enough to construct an isomorphism of principal (G G  )-bundles


X  (F F ) X (F F  )  . (7.5.6)

This will prove the lemma for Whitney sum, and also for the tensor product, using
instead of .
As  X : X X X is -equivariant with respect to the diagonal homomorphism
 :   , it induces a map = ( X ), : X  (X X ) . By
Lemma 7.5.5, one has

 X (F F  )  (F F  ) . (7.5.7)


Z  Z 
For -space Z and Z  , a natural homotopy equivalence P : (Z Z  )
 
was constructed in (7.4.1). For Z = F and Z = F , we thus get a homotopy

equivalence P : (F F  ) F F which is (G G  )-equivariant. The
diagram

/ (F F  ) P / F F 
(F F  )  

  
/ (X X ) P / X X
X 

is commutative, thus each square is a morphism of principal (G G  )-bundles.

By the definition of P in Sect. 7.4, one has P =  X  . Therefore, (F
7.5 Equivariant Bundles and Euler Classes 303

F  ) X  (F F ). This together with (7.5.7) gives the required isomorphism
of (7.5.6).

Let be a -vector bundle of rank r over X . The equivariant Euler class e () is

the Euler class of  :

e () = e( ) Hr (X ).

Example 7.5.8 Let :  G L(V ) be a representation of  on a vector space V

of dimension r . This makes V a -space. One can also see V as a vector bundle with
basis a point. This gives a -vector bundle of rank r over a point and then a vector
bundle  = (V B) of rank r over B. Its equivariant Euler class e () is an
element of Hr ( pt) = H r (B). Its vanishing is related to the existence of a non-zero
fixed vector in V , as seen in the following lemma.

Lemma 7.5.9 If V 
= {0}, then e ()
= 0.

Proof A non-zero fixed vector 0

= v V  determines a nowhere-zero section of
p : V B (see (5) p. 275). This implies that e ()
= 0 by Lemma 4.7.39. 

We now give a few recipes to compute an equivariant Euler class. A -equivariant

vector bundle p : E X is called rigid if the -action on X is trivial.

Lemma 7.5.10 Let = ( p : E X ) be a rigid -vector bundle of rank r . Let

:  G L(E x ) be the representation of  on the fiber E x over x X . Suppose
that H k (X ) = 0 for k < r . Then, the equation

e () = 1 e() + e ()
1

holds in Hr (X ) = H r (B X ).

Proof The inclusion j : {x} X satisfies j = . It induces j : B = {x}

X  satisfying j  =  . Hence, H j (e ()) = e ().
By construction of  , one
has i  = , where i : X X  denotes the inclusion. Hence, H i(e ()) = e().
Using the homeomorphism X  B X , the maps j and i are slice inclusions.
The lemma then follows from Corollary 4.7.3. 

Let :  O(1) be a continuous homomorphism, permitting  to act on

R. This gives a -line bundle over a point (see Example 7.5.8). Its equivariant
Euler class lives in H1 ( pt) = H 1 (B). As is continuous, it factors through the
homomorphism 0 : 0 () O(1) Z2 . As E is contractible, the homotopy
exact sequence of  E B identifies 0 () with 1 (B). One thus gets
(using Lemma 4.3.1) the isomorphism

homcont (, O(1)) hom(0 (), Z2 ) hom(1 (B), Z2 ) H 1 (B) = H1 ( pt).
(7.5.8)
304 7 Equivariant Cohomology

Lemma 7.5.11 Under the isomorphism of (7.5.8), one has e ()

= .

Proof Note that E  O(1) B is a 2-fold covering and that

= E  R = (E  O(1)) O(1) R.

E()

Then, E  O(1) B is the sphere bundle S(  ) for the Euclidean structure on

 given by the standard Euclidean structure on R. By Proposition 4.7.36, the Euler
class e ()
coincides with the characteristic class w(S(  )) of the two-covering
S(  ) B. But E  O(1) B ker and thus

1 (S(  )) = 1 (B ker ) = 0 (ker ) = ker 0 = ker 1 B.

Therefore, S(  ) B is the 2-covering with fundamental group to ker 1 B

1 (B). Diagram (4.3.4) then implies that w(S(  )) = . 

A discrete group  is a 2-torus if it is finitely generated and if every element has

order 2. It follows that  is isomorphic to {1}m , the integer m being called the rank
of . As seen in Sect. 7.4,

B  B({1}m )  (B{1})m  (RP )m . (7.5.9)

Hence, H ( pt) is isomorphic to a polynomial algebra

H ( pt) H ((RP )m ) H (RP ) H (RP ) Z2 [u 1 , . . . , u m ],

(7.5.10)
where degree(u i ) = 1. Under the identifications of (7.5.8), u i H 1 ( pt) corre-
sponds to the homomorphism {1}n {1} which is the projection onto the ith
factor.

Example 7.5.12 Let  be the 2-torus formed by the diagonal matrices of O(n). Then
H (B) Z2 [u 1 , . . . , u n ], where u i H 1 (B) corresponds to the homomorphism
i :  {1} given by the i-th diagonal entry. The inclusion :  O(n)
provides a -equivariant vector bundle of rank n over a point. Note that is a direct
sum of 1-dimensional representations i . Using Lemmas 7.5.7 and 7.5.11, we get that

n
w()
= (1 + u i ).
i=1
Lemma 7.5.13 Let :  G L(V ) be a representation of a 2-torus  on a finite
dimensional vector space V . Then the following two conditions are equivalent.
(1) V  = {0}.
(2) e ()

= 0.

Proof That (2) implies (1) is given by Lemma 7.5.9. To prove the converse, we use
the fact that is diagonalizable, with eigenvalues 1: indeed, this is true for a linear
7.5 Equivariant Bundles and Euler Classes 305

involution (see Example 7.1.14) and, if a, b G L(V ) commute, then b preserves the
eigenspaces of a. Thus, V = V1 Vr and  acts on V j through a homomorphism
j :  {1} = O(1). Hence, is the Whitney sum 1 r . Therefore,

e ()
= e(  )
= e(( 1 ) ( r ) ) by Lemma 7.5.7
= e(( 1 ) )   e(( r ) ) by Proposition 4.7.40
= 1   r by Lemma 7.5.11.

The condition V  = {0} implies that none of the j vanishes. As H (B) is a

polynomial algebra, this implies that e ()

= 0. 
p
Proposition 7.5.14 Let = (E X ) be a rigid -vector bundle of rank r , where
 is a 2-torus. Suppose that E  consists only of the image of the zero section. Then,
the cup-product with the equivariant Euler class

e ()
H (X ) H+r (X )

is injective.

Proof Without loss of generality, we may suppose that X is path-connected and

non-empty. Let x X . Consider the slice inclusion s : B B X with image
B {x}. Then, H s(e ()) = e (x ), where x = (E x {x}) is the restriction of
over the point x. As X is path-connected, Lemma 4.7.2 implies that the component
of e () H r (B X ) in H r (B) H 0 (X ) is equal to e (x ) 1 (as B is of
finite cohomology type, we identify H r (B X ) with H (B) H (X ) by the
Knneth theorem).
Now let 0
= a H k (B X ). We isolate its minimal component amin by the
formula

a = amin + A

with

0
= amin H k p (B) H p (X ) and A H kq (B) H q (X ).
q> p

Then

a  e () = amin  (e (x ) 1) + A

with

amin  (e (x ) 1) H k p+r (B) H p (X ) and A H kq+r (B) H q (X ).
q> p
306 7 Equivariant Cohomology

Therefore it suffices to prove that amin  (e (x ) 1)

= 0. The condition on
E  implies that E x = {0} and thus, by Lemma 7.5.13, e (x )
= 0. As H (B)
is a polynomial algebra, this implies that a  e ()
= 0. Let B be a basis of
H k p (B) and C be a basis of H p (X ). Then, {b c | (b, c) B C} is a
basis of H k p (B) H p (X ). As H (B) is a polynomial algebra, the family
{b  e (x ) | b B} is free in H k p+r (B). Hence, if

0
= amin = bc (b c) (bc Z2 ),
(b,c)BC

then
  
amin  (e (x ) 1) = bc (b  e (x )) c
= 0 . 
(b,c)BC
Statements 7.5.11, 7.5.13 and 7.5.14 have analogues, replacing O(1) by S O(2)
and 2-tori by tori. Let  be a topological group and let :  S O(2) be a
continuous homomorphism, making  to act on R2 . This gives a -vector bundle of
rank 2 over a point (see Example 7.5.8). Its equivariant Euler class lives in H2 ( pt) =
H 2 (B). As S O(2) S 1 , one has B S O(2) CP (see Example 7.2.1). Define

() = H B() H2 ( pt) = H 2 (B), (7.5.11)

where is the non-zero element of H 2 (B S O(2)) = Z2 .

Lemma 7.5.15 e ()
= ().
Proof The homomorphism :  S O(2) makes S O(2) a -space. The map
E : E E S O(2) descends to continuous maps E  S O(2) E S O(2) and
there is a commutative diagram

/ E S O(2) o / S
E  S O(2)
(7.5.12)
  
B
/ B S O(2) o / CP
B

Note that E  S O(2) B is an S O(2)-principal bundle and that

= E  R2 = (E  S O(2)) S O(2) R2 .

E()

As S O(2) S 1 , E S O(2) is the sphere bundle S(  ) for the Euclidean structure
on  given by the standard Euclidean structure on R2 . Diagram (7.5.12) implies
that S(  ) is induced by B from the Hopf bundle S CP , whose Euler class
in H 2 (CP ) (see Proposition 6.1.6). Hence,

= e(S(  )) = H B() = ().

e () 
7.5 Equivariant Bundles and Euler Classes 307

A torus  is a Lie group isomorphic to (S 1 )m , the integer m being called the rank
of . For instance, S O(2) is a torus of rank 1. As seen in Sect. 7.4,

B  B((S 1 )m )  (B S 1 )m  (CP )m . (7.5.13)

Hence, H ( pt) is isomorphic to a polynomial algebra

H ( pt) H ((CP )m ) H (CP ) H (CP ) Z2 [v1 , . . . , vm ],

(7.5.14)

where degree(vi ) = 2. One has vi = (i ) where i : (S 1 )n S 1 S O(2) is the

projection onto the ith factor. If  is a torus, its associated 2-torus 2 is the subgroup
of elements of order 2 in .
Lemma 7.5.16 Let  be a torus and 2 be its associated 2-torus. Let : 
G L(V ) be a representation of  on a finite dimensional vector space V . Then the
following two conditions are equivalent.
(1) V 2 = {0}.
(2) e ()

= 0.
Moreover, if (1) or (2) holds true, then dim V is even.
Proof As  is a torus, () is contained in a maximal torus T of G L(V ). Those
are all conjugate (see [21, Sect. IV.1]). If dim V = 2s + 1, there is an isomorphism
V R2 R2 R intertwining T with S O(2) S O(2)1 (see [21, Chap. IV,
Theorem 3.4]). This contradicts the condition V 
= {0}. We can then suppose that
dim V = 2s, in which case there is an isomorphism V R2 R2 conjugating
T with S O(2) S O(2). The homomorphism takes the form = (1 , . . . , s )
where j :  S O(2). Hence = 1 s and, using Lemma 7.5.7 and
Proposition 4.7.40,

e ()
= e ( 1 )   e ( s ) = (1 )   (s ),

the last equality coming from Lemma 7.5.15. Since H (B) is a polynomial algebra,
the condition e ()
= 0 is equivalent to ( j )
= 0 for all j. The condition V  =
{0} = V 2 is equivalent to V j = {0} = V j2 for all j, where V j is the 2-dimensional
vector space corresponding to the jth factor R2 in the decomposition of V .
We are thus reduced to the case dim V = 2 and :  S O(2). We start with
preliminaries. Choose isomorphisms  (S 1 )m , S O(2) S 1 and S 1 R/Z. We
get a commutative diagram

Zm / / Rm / / (S 1 )m

1 (7.5.15)
  
Z / / R / / S1
308 7 Equivariant Cohomology


where the vertical arrows are homomorphisms. Therefore, (x
1 , . . . , xm ) = i bi xi
with bi Z and

(1 , . . . , m ) = 1b1 mbm . (7.5.16)

We deduce that

V  = {0} 1 non-trivial surjective b j

= 0 j. (7.5.17)

If is surjective, one gets a fiber bundle ker  S 1 and, using its homotopy
exact sequence and (7.5.16), we get

V  = {0} = coker 1 0 (ker ) Z gcd(b1 , . . . , bm )Z. (7.5.18)

Since 0 V  V 2 , the condition V 2 = {0} implies that V  = {0}. Hence,

using (7.5.15)(7.5.18),

V 2 = {0} 2 ker
21 Z ker 1
. (7.5.19)
2 | gcd(b1 , . . . , bm )
hom(0 (ker ); Z2 ) = 0

As in the proof of Lemma 7.5.15, we consider S = E  S 1 , which is the total

space for the sphere bundle S().
One has a commutative diagram

 / E / B

=
  
S1 / S / B

whose rows are fiber bundles. Passing to the homotopy exact sequences, we get a
commutative diagram
/ 1 () / 0
2 (B)
= 1
  
/ 1 (S 1 ) / 1 (S) / 0
2 (B)

whose rows are exact sequences. Hence,

1 (S) coker 1 (7.5.20)

Now, the Gysin sequence for S B gives

7.5 Equivariant Bundles and Euler Classes 309

e ()

H 1 (B) H 1 (S) H 0 (B) H 2 (B). (7.5.21)
     
0 Z2

= 0 implies V  = {0}. Therefore,
By Lemma 7.5.9, one knows that e ()
using (7.5.18)(7.5.21),

e ()

= 0 H 1 (S) = 0
hom(1 (S); Z2 ) = 0
hom(coker 1 ; Z2 ) = 0 . (7.5.22)
hom(0 (ker ); Z2 ) = 0
V 2 = {0} 

Proposition 7.5.17 Let  be a torus and 2 be its associated 2-torus. Let = ( p :

E X ) be a rigid -vector bundle of rank r . Suppose that E 2 consists only of the
image of the zero section. Then r is even and the cup-product with the equivariant
Euler class
e ()
H (X ) H+r (X )

is injective.

Proof The proof is the same as that of Proposition 7.5.14, using Lemma 7.5.16
instead of Lemma 7.5.13. 

7.6 Equivariant Morse-Bott Theory

Let f : M R be a smooth function defined on a smooth manifold M. A point

x M is critical for f if d f (x) = 0. Let Crit f M be the subspace of critical
points for f . Then, f (Crit f ) R is the set of critical values of f . We say that f is
Morse-Bott if the following two conditions hold
Crit f is a disjoint union of submanifolds. A connected component of Crit f is
called a critical manifold of f .
the kernel of the Hessian Hx at a critical point x equals the tangent space to the
critical manifold N containing x.
This definition coincides with that of a Morse function when Crit f is a discrete set.
See e.g. [13, 18, 95, 149, 151] for presentations of Morse and Morse-Bott theory.
The index of x Crit f is the number of negative eigenvalues of Hx . This number is
constant over a critical manifold and thus defines a function ind : 0 (Crit f ) N.
Also, the normal bundle N to a critical manifold N decomposes into a Whitney
sum N N + N of the negative and positive normal bundles, i.e. the bundles
spanned at each x N respectively by the negative and positive eigenspaces of Hx .
Note that rank (N ) = ind N .
310 7 Equivariant Cohomology

A (continuous) map g : X Y is called proper if the pre-image of any compact

set is compact. For instance, if X is compact, then any map g is proper. Let f :
M R be a proper Morse-Bott function and let a < b be two regular values.
Define Ma,b = f 1 ([a, b]), a compact manifold whose boundary is the union of
Ma = f 1 (a) and Mb = f 1 (b). Denote by f a,b the restriction of f to Ma,b . The
Morse-Bott polynomial Mt (Ma,b ) is defined by

Mt ( f a,b ) = t indN Pt (N )
N 0 (Crit f a,b )

(the sum is finite since, as f is proper, Crit f a,b is compact).

Proposition 7.6.1 (Morse-Bott inequalities) There is a polynomial Rt , with positive
coefficients, such that

Mt ( f a,b ) = Pt (Ma,b , Ma ) + (1 + t)Rt . (7.6.1)

Equation (7.6.1) implies that the coefficients of Mt ( f a,b ) are greater or equal
to those of Pt (Ma,b , Ma ) (whence the name of Morse-Bott inequalities). For the
equivalence of (7.6.1) with other classical and more subtle forms of the Morse-Bott
inequalities (see [13, Sect. 3.4]).
Proof The map f a,b has a finite number of critical values, all in the interior of [a, b].
Let a = a0 < a1 < < ar = b be regular values such that [ai , ai + 1] contains
a single critical value. We shall prove by induction on i that (7.6.1) holds true for
f a,ai . The induction starts trivially for i = 0, with the three terms of (7.6.1) being
zero.
As [ai , ai+1 ] contains a single critical level, there is a homotopy equivalence

Mai ,ai+1  Mai Si Di (7.6.2)

where (Di , Si ) is the disjoint union over N 0 (Crit f ai ,ai +1 ) of the pairs formed
by the disk and sphere bundles of (N ) (see [18, pp. 339344]). By excision and
the Thom isomorphism,

H (Mai ,ai+1 , Ma ) H (Di , Si ) H ind N (N ). (7.6.3)

N 0 (Crit f ai ,ai +1 )

Therefore,

Mt ( f ai ,ai +1 ) = Pt (Mai ,ai +1 , Mai ). (7.6.4)

On the other hand, Corollary 3.1.27 applies to the triple (Ma,ai+1 , Ma,ai , Ma ) gives
the equality

Pt (Ma,ai+1 , Ma ) + (1 + t) Q t = Pt (Ma,ai+1 , Ma,ai ) + Pt (Ma,ai , Ma ) , (7.6.5)

7.6 Equivariant Morse-Bott Theory 311

for some Q t N[t]. By excision and (7.6.4), one gets

Pt (Ma,ai+1 , Ma,ai ) = Pt (Mai ,ai +1 , Mai ) = Mt ( f ai ,ai +1 ). (7.6.6)

Thus, (7.6.5) and (7.6.6) provide the induction step. 

A proper Morse-Bott function f : M R is called perfect if for any two regular

values a < b, Eq. (7.6.1) reduces to

Mt ( f a,b ) = Pt (Ma,b , Ma ). (7.6.7)

The easiest occurrence of perfectness is the following lacunary principle.

Lemma 7.6.2 Suppose that no consecutive powers of f occur in Mt ( f ). Then, f

is perfect.

Proof Suppose that Rt

= 0 in (7.6.1). Then, two successive powers of t occur in
(1 + t)Rr . The same happens then in Mt ( f a,b ), and then in Mt ( f ). 

Other simple criteria for perfectness are given by the following three results. For
a regular value x of f : M R, set Wx = f 1 (, x].

Lemma 7.6.3 Let f : M R be a proper Morse-Bott function. Then, the following

two conditions are equivalent.
(1) f is perfect.
(2) For any regular values a < b < c of f , the cohomology sequence of the triple
(Wc , Wb , Wa ) cuts into a global short exact sequence

0 H (Wc , Wb ) H (Wc , Wa ) H (Wb , Wa ) 0.

Proof Suppose that f is perfect. Then, by excision,

Pt (Wc , Wa ) = Pt (Ma,c , Ma ) = Mt ( f a,c )

and analogously for Pt (Wb , Wa ) and Pt (Wc , Wb ). As

Mt ( f c,a ) = Mt ( f c,b ) + Mt ( f b,a ),

one has

Pt (Wc , Wa ) = Pt (Wc , Wb ) + Pt (Wb , Wa ).

By Corollary 3.1.27 and its proof, this implies that H (Wb , Wa ) H (Wc , Wb ) is
surjective, whence (2).
Conversely, suppose that (2) holds true. For two regular values a < c, we prove
that Pt (Ma,c , Ma ) = Pt (Wc , Wa ) = Mt ( f a,c ), by induction on the number n a,c of
312 7 Equivariant Cohomology

critical values in the segment [a, c]. This is trivial for n a,c = 0, since then Ma,c is
then diffeomorphic to Ma [a, c] (see [95, Chap. 6, Theorem 2.2]). When n a,c = 1,
one uses (7.6.4). For the induction step, when n a,c 2, choose a regular value
b (a, c) such that n a,b = n a,c 1. Then,

Pt (Wc , Wa ) = Pt (Wc , Wb ) + Pt (Wb , Wa ) by (2)

= Mt ( f c,b ) + Mt ( f b,a ) by induction hypothesis
= Mt ( f a,c ),

which proves the induction step. 

Lemma 7.6.4 Let f : M R be a proper Morse-Bott function. Then if for any

two regular values a < b, one has

dim H (Ma,b , Ma ) dim H (Crit f a,b ) (7.6.8)

and f is perfect if and only if (7.6.8) is an equality.

Proof The evaluation of (7.6.1) at t = 1 implies (7.6.8) and the equality is equivalent
to Rt = 0. 

In the case where M is a closed manifold, one has the following result.

Proposition 7.6.5 Let f : M R be a Morse-Bott function, where M is a closed

manifold. Then f is perfect if and only if

dim H (M) = dim H (Crit f ). (7.6.9)

Proof Equation (7.6.8) implies (7.6.9) when f (M) (a, b). Conversely, let a < b
be two regular values of f . Let a  < a and b > b such that f (M) (a  , b ). Using
Corollary 3.1.28 and excision, we get

dim H (M) = dim H (Ma  ,b , Ma  )

dim H (Ma  ,b , Ma ) + dim H (Ma  ,b , Ma  ,b ) (7.6.10)
= dim H (Ma  ,b , Ma ) + dim H (Mb,b , Mb ).

Doing the same for dim H (Ma  ,b , Ma ) and using (7.6.1) gives

dim H (M) dim H (Ma  ,a , Ma ) + dim H (Ma,b , Ma ) + dim H (Mb,b , Mb )

dim H (Crit f a  ,a ) + dim H (Crit f a,b ) + dim H (Crit f b,b )
= dim H (Crit f ).
(7.6.11)

Now, if (7.6.9) holds true, then all the inequalities occurring in (7.6.10) are equalities,
including dim H (Ma,b , Ma ) = dim H (Crit f a,b ). 
7.6 Equivariant Morse-Bott Theory 313

Theorem 7.6.6 Let M be a smooth -manifold, where  is a 2-torus. Let f : M

R be a proper -invariant Morse-Bott function which is bounded below. Suppose
that Crit f = M  . Then
(1) f is perfect.
(2) M is -equivariantly formal.
(3) the restriction morphism H (M) H (M  ) is injective.

Remark 7.6.7 When  = {1}, Theorem 7.6.6 follows from Smith theory. Indeed,
for any regular values a < b of f ,

 ) by hypothesis
dim Crit f a,b = dim H (Ma,b

dim H (Ma,b ) by Proposition 7.3.7
dim Crit f a,b by Lemma 7.6.4.

Therefore, the above inequalities are equalities and f is perfect by Lemma 7.6.4
and equivariantly formal by Proposition 7.3.7. Point (3) then follows from Proposi-
tion 7.3.9.

Remark 7.6.8 Under the hypotheses of Theorem 7.6.6, when  = {1} and f is a
Morse function, M. Farber and D. Schtz have proven that each integral homology
group Hi (M; Z) is free abelian with rank equal to the number of critical points of
index i [60, Theorem 4]. By the universal coefficients theorem [82, Theorem 3.2],
such a function is perfect.

Proof of Theorem 7.6.6 For a regular value x of f , set Wx = f 1 (, x]. We

first prove that

H (Wx ) H (Wx ) is injective for all regular value x. (7.6.12)

This is proven by induction on the number n x of critical values in the interval

(, x], following the argument of [198, proof of Proposition 2.1]. If n x = 0,
then Wx = and H (Wx ) = 0, which starts the induction as f is bounded below.
Suppose that n 1 and that (7.6.12) holds true when n x < n. Let y be a regular
value of f with n y = n. Choose z < y such that n z = n 1 (this is possible since
the set of critical values is discrete Morse-Bott function is discrete). As in (7.6.2),
one has a homotopy equivalence

Mz,y  M y S D,

where (D, S) is the disjoint union over N 0 (Crit f z,x ) of the pairs formed by the
disk and sphere bundles of (N ). Using (7.6.3) and the proof of Proposition 4.7.32,
we get the commutative diagram
314 7 Equivariant Cohomology


H (W y , Wz ) o H (Mz,y , Mz ) o H (D, S) o Hind N (N )
excision  excision Thom N
 
 e ( (N ))
   
/ H (Mz,y ) / H (Mz,y /
H (W y )  
 )
N H (N )
(7.6.13)
 ). As M  = Crit f , the linear -action
where N runs over 0 (Crit f z,y ) = 0 (Mz,y
of (N ) has fixed point set consisting only of the image of the zero section. By
Lemma 7.5.13, the right vertical arrow of (7.6.13) is injective. Thus, we deduce
from (7.6.13) that H (W y , Wz ) H (W y ) is injective. This cuts the -equivariant
cohomology sequence of (W y , Wz ) into short exact sequences. The same cutting
occurs for the pair (W y , Wz ) using Proposition 3.1.21, and one has a commutative
diagram

0 / H (W y , Wz ) / H (W y ) / H (Wz ) / 0
  

r z,y ry rz (7.6.14)
  
0 / H (Mz,y
 ) / H (W y ) / H (W  ) / 0
   z

where the vertical arrows are induced by the inclusions. The left vertical arrow is
injective by (7.6.13). Since n z = n 1, the right one is injective by induction
hypothesis. By diagram-chasing, we deduce that the middle vertical arrow is injective,
which proves (7.6.12).
Warning: As W y is the disjoint union of Mz,y and W  , one has H (W  )
z  y
 
H (Mz,y ) H (Wz ). Consider the image Im r y of r y under this decomposition. The
above arguments imply that Im r z,y 0 Im r y . But, in general 0 Im r z
Im r y
(see Example 7.6.9). 
We deduce Point (3) from (7.6.12). Indeed as M = x Wx Corollary 3.1.16
provides a commutative diagram

/ lim H (Wx )
H (M)
x

. (7.6.15)


/ lim H (Wx )
H (M  )
x

As the right vertical arrow is injective by (7.6.12), so is the left one.

For Point (2), we first prove that x : H (Wx ) H (Wx ) is surjective for all
regular value x. This is also done by induction on n x , starting trivially when n x = 0.
For the induction step, consider as above two regular values z < y such that n y = n z +
1. The cohomology sequences of the pair (W y , Wz ) give the commutative diagram
7.6 Equivariant Morse-Bott Theory 315

0 / H k (W y , Wz ) / H k (W y ) / H k (Wz ) / 0
  

y,z y z ,
  
H k1 (Wz ) / H k (W y , Wz ) / H k (W y ) / H k (Wz ) / H k+1 (W y , Wz )
(7.6.16)

the top sequence being cut as seen above. Similarly to (7.6.13), we get a commutative
diagram


H (W y , Wz ) o H (Mz,y , Mz ) o H (D, S) o Hind N (N )
excision  excision Thom N

y,z Crit .
   

H (W y , Wz ) o H (Mz,y , Mz ) o H (D, S) o N Hind N (N )
excision excision Thom

(7.6.17)

Since Crit f M  , the map Crit is surjective and so is y,z . If z is surjective by

induction hypothesis, a diagram-chase proves that y is surjective. Now, recall that,
for a -space X , : H (X ) H (X ) is equal to H i for some fiber inclusion
i : X X  . One can thus consider the Kronecker dual : H (X ) H (X  ).
One has a commutative diagram

/ H (M)
lim

H (Wx )
x

lim x, . (7.6.18)



/ H (M )
lim

H ((Wx ) )
x

As x is surjective, x, is injective and thus lim x, is injective. By Diagram (7.6.18)

and Kronecker duality, : H (M) H (M) is surjective and M is equivariantly
formal.
Let us finally prove Point (1). Consider two regular values z < y such that
n y = n z +1. As in (7.6.16) the vertical maps are surjective, the cohomology sequence
of (W y , Wz ) cuts into a global short exact sequence. By the proof that (2) implies (1)
in Lemma 7.6.3, this implies that f is perfect.

Example 7.6.9 Consider the action of  = {1} on M = S n R Rn given by

(t, x) = (t, x), with fixed points p = (1, 0). Note that M is a sphere with
linear involution S0n in the sense of Example 7.1.14. The formula f (t, x) = t defines
a Morse function M satisfying the hypotheses of Theorem 7.6.6. Taking z = 0 and
y = 2 as regular values of f , one has W y = M and Diagram (7.6.14) becomes
316 7 Equivariant Cohomology

j
0 / H (M, W0 ) / H (M) / H (W0 ) / 0
  

r+ r r .
  
0 / H ( p+ ) / H ( p+ ) H ( p ) / H ( p ) / 0
   

Set H (B) = Z2 [u]. By Lemma 7.5.13, e ( ( p+ )) = u n . Together with

Diagram (7.6.13), this shows that r j (U ) = (u n , 0), where

U Hn (M, W0 ) Hn (M0,2 , M0 ) H0 ( p+ )

is the Thom class of ( p+ ). Let B = j (U ). Using the diagram

Thom / j
H0 ( p+ ) Hn (M, W0 ) / H n (M)


  
Thom / n (M, W ) / H n (M)
H ( p+ )
0

H 0

one sees that (B) is the generator of H n (S n ) = Z2 . Hence, is surjective, as

expected by Theorem 7.6.6. By the Leray-Hirsch theorem, H (M) is the free Z[u]-
module generated by B. Now, r (B) = (u n , 0) and r (u) = (u, u). As is r injective by
Theorem 7.6.6, the relation B 2 = u n B holds true in H (M). Given the dimension
of Hk (M), this establishes the GrA[u]-isomorphism

Z2 [B, u] (B 2 + u n B) H (M)

Takingthe image by adds the relation u = 0 and we recover that H (S n )

Z2 [B] (B 2 ). Note that (0, u) is not in the image of r , confirming the warning in the
proof of Theorem 7.6.6. But (0, u n ) = r (B + u n ) is in the image of r , corresponding
to the generator A = B + u n of H (M) (see Example 7.1.16). Had we considered
f instead of f , the above discussion would have selected the generator A first.
The relation A2 = u n A also holds true and recall from Example 7.1.16 that H (M)
admits the presentation

Z2 [u, A, B] (A2 + u n A, A + B + u n ) H (M).

Example 7.6.10 Let  be a 2-torus and let :  {1} be a homomorphism,

identified with H 1 (B) under the bijection (7.5.8). Consider the -action on
M = S 1 R2 given by (t, x) = (t, () x), with fixed points p = (1, 0). We
call M a -circle. As in Example 7.6.9, one sees that the image of

r : H (M) H ( p ) H ( p ) H ( p+ ) H (B) H (B)

7.6 Equivariant Morse-Bott Theory 317

is the H (B)-module generated by B = (, 0) and that H (M) admits the presen-

tation

H (M) H (B)[B] (B 2 + B).

Moreover, the image of r is the set of classes (a, b) such that b a is a multiple
of .

Theorem 7.6.6 admits the following analogue for torus actions.

Theorem 7.6.11 Let  be a torus and 2 be its associated 2-torus. Let M be a

smooth -manifold. Let f : M R be a proper -invariant Morse-Bott function
which is bounded below. Suppose that Crit f = M  = M 2 . Then
(1) f has only critical manifolds of even index. In particular, if f is a Morse function,
then M is of even dimension.
(2) f is perfect.
(3) M is -equivariantly formal.
(4) the restriction morphism H (M) H (M  ) is injective.

Proof The proof is the same as that of Theorem 7.6.6. The hypothesis Crit
f = M  implies that the negative normal bundles are -vector bundles and the
hypothesis Crit f = M 2 permits us to use Proposition 7.5.17 instead of Proposition
7.5.14. 

In Theorem 7.6.11, note that the perfectness of f is implied by (1). When  = S 1 ,

Points (3) and (4) follows from Smith theory, in the same way as in Remark 7.6.7.

Example 7.6.12 Let  be a torus with associated 2-torus 2 . Let :  S 1 be a

continuous homomorphism. Consider the -action on M = S 2 R C given by
(t, x) = (t, () x), with fixed points p = (1, 0). We call M a -sphere. Let
us assume that the restriction of to the associated 2-torus 2 of  is not trivial. This
implies that M  = M 2 , so we can apply Theorem 7.6.11 and, as in Example 7.6.9,
one sees that the image of

r : H (M) H ( p ) H ( p ) H ( p+ ) H (B) H (B)

is the H (B)-module generated by B = ((), 0), where () H 2 (BT ) is

defined in (7.5.11). Also, H (M) admits the presentation

H (M) H (B)[B] (B 2 + () B).

Moreover, the image of r is the set of classes (a, b) such that b a is a multiple
of ().

A consequences of Theorem 7.6.6 and 7.6.11 are the surjectivity theorems la

Kirwan (see Remark 7.6.16 below). For f : M R is a continuous map, we set
318 7 Equivariant Cohomology

M = f 1 ((, 0]), M+ = f 1 ([0 )) and M0 = M M+ = f 1 (0). The

inclusions form a commutative diagram.

i+
M0 / M+
DD
DDi
i DD j+ . (7.6.19)
 D" 
j
M / M

Proposition 7.6.13 Let M be a closed smooth -manifold, where  is a 2-torus. Let

f : M R be a -invariant Morse-Bott function satisfying Crit f = M  . Suppose
that 0 is a regular value of f . Then

H i : H (M) H (M0 )

is surjective and its kernel is the ideal ker H j +ker H j+ , generated by ker H j
and ker H j+ .

For applications of this proposition, see Sect. 10.3.2.

Proof We use the abbreviations i = H i , j = H j , etc. As M is compact, f is

 
proper and bounded. By Theorem 7.6.6, the restriction homomorphism H (M)
H (M  ) = H (Crit f ) is injective. The commutative diagram

, j )
( j
H (M)
+
/ H (M ) H (M+ )
 


 
/ H (M  ) H (M  )
H (M  )   +

shows that the Mayer-Vietoris sequence in equivariant cohomology for Diagram

(7.6.19) splits into a global short exact sequence
, j )
( j +i
i
+ +
0 H (M) H (M ) H (M+ ) H (M0 ) 0. (7.6.20)

Suppose that x1 < x2 < x3 < . . . are regular values of f such that f xi ,xi+1 has
only one critical level (we use the notations of Theorem 7.6.6 and its proof). Then,
by (7.6.14), H (Wi+1 ) H (Wi ) is surjective. As W0 = M and M is compact,
this argument shows that j is surjective. By symmetry, replacing f by f (using
again that M is compact), one also has that j+ is surjective.
Let a H (M0 ). Using (7.6.20), choose a H (M ) such that a = i
(a )+


i (a+ ). As j is surjective, there exist b H (M) with i (b ) = a . Then

i (b +b+ ) = i (b )+i (b+ ) = i

j (b )+i + j+ (b+ ) = i (a )+i + (a+ ) = a,
7.6 Equivariant Morse-Bott Theory 319

which proves that i = H i is surjective. As i = i

j , we have also proven that


i is surjective. Therefore, one has a commutative diagram

j+
H (M, M+ ) / / H (M)

/ / H (M+ )

LLL
j
L LLL
i
i (7.6.21)
 
L% %  +
i
H (M , M0 ) / / H (M ) / / H (M0 )
 

where the horizontal and vertical sequences are exact and the left hand vertical arrow
is an isomorphism by excision. Hence, ker i = ( j )1 (ker i
), which yields to an

exact sequence

0 ker j ker i ker i

0.

But Diagram (7.6.21) provides a section of ker i ker i , whose image is equal

to ker j+ . This proves the assertion on ker i (which is actually GrV-isomorphic to
ker j ker j+ ). 

Example 7.6.14 Consider the action of  = {1} on M = RP n given by

[x0 , . . . , xn ] = [x0 , . . . , xn1 , xn ].

The Morse-Bott function defined on M by f ([x0 , . . . , xn ]) = 1 2xn2 satisfies the

hypotheses of Theorem 7.6.6 and Proposition 7.6.13. Set Mc = f 1 (c). The critical
submanifolds are M1 and one has M  M1 = pt and M+  M1 = RP n1 .
Let u H 1 (B) = Z2 be the generator.
The bundle projection p : M B and its restriction pc to (Mc ) give elements
H pc (u) = u c H (Mc ) and H p(u) = v H (M). One has

H (M ) Z2 [u 1 ] and H (M+ ) H (B M1 ) Z2 [b, u 1 ]/(bn ),

with the degree of b equal to 1. Consider the commutative diagram

H0 (M1 )
Thom / H 1 (M, M ) / 1 / H 1 (M)
 3 H (M)
r1 .
 
1
H1 (M1 ) / H 1 (M1 )

By Diagram (7.6.13), r1 (1) = e ( ), the equivariant Euler class of the (negative)

normal bundle (M1 ). By Lemma 7.5.10,

e ( ) = 1 e( (M1 )) + e()
1 H 1 (B M1 ),
320 7 Equivariant Cohomology

where is the representation of  on a fiber of (M1 ). Note that (M1 ) = (M1 )

is the canonical line bundle over RP n1 (since M = RP n is obtained by attaching
an n-cell to M1 = RP n1 by the Hopf map (two fold covering) S n1 RP n1 ).
Then, by Proposition 4.7.36 and its proof, e( (M1 )) = b. As has no non-zero
fixed vector, e ()
= u by Lemma 7.5.13. Therefore,

e ( ) = 1 b + u 1 = b + u 1 ,

the last formula making sense in the presentation H (M1 ) Z2 [b, u 1 ]/(bn ). As
1 (b + u 1 ) = b, this proves that is surjective, as already known by Theorem 7.6.6.
Let a = (1) + v. One also has 1 r1 (a) = b so, by the Leray-Hirsch theorem,
H (M) is the free Z2 [v]-module generated by a, a 2 , . . . , a n1 and the Poincar
series of M is

1 t n+1
Pt (M ) = Pt (M) Pt (B) = . (7.6.22)
(1 t)2

By Diagram (7.6.14), r1 (1) = 0 in H1 (M ) H1 (M1 ). Therefore, the

homomorphism r : H (M) H (M  ) H (M1 ) H (M1 ) satisfies
r (a) = (u 1 , b) and r (v) = (u 1 , u 1 ). Finally, we claim that there is a GrA-
isomorphism

H (M).
Z2 [v, a] (a n+1 + va) (7.6.23)

Indeed, one already knows that H (M) is GrA-generated by v and a, and, using the
injective homomorphism r , one checks that the relation a n+1 = va holds true. This
gives the GrA-morphism of (7.6.23) which is surjective, and hence bijective since
both sides of (7.6.23) have the same Poincar series, computed in (7.6.22).

Replacing Theorem 7.6.6 by Theorem 7.6.11 in the proof of Proposition 7.6.13

gives the following result.

Proposition 7.6.15 Let M be a closed smooth -manifold, where  is a torus with

associated 2-torus 2 . Let f : M R be a -invariant Morse-Bott function
satisfying Crit f = M  = M 2 . Suppose that 0 is a regular value of f . Then

H i : H (M) H (M0 )

is surjective and its kernel is the ideal ker H j +ker H j+ , generated by ker H j
and ker H j+ .

As an example, one can take the complex analogue of Example 7.6.14, i.e.  = S 1
acting on M = CP n given by [x0 , . . . , xn ] = [x0 , . . . , xn1 , xn ] and the Morse-
Bott function f ([x0 , . . . , xn ]) = 1 2xn2 . All the formulae of Example 7.6.14 hold
true, with the degrees of all the classes multiplied by 2.
7.6 Equivariant Morse-Bott Theory 321

Remark 7.6.16 For  = S 1 , the hypotheses of Proposition 7.6.15 are realized when
f is the moment map of a Hamiltonian circle action (see [12]). In this case, it follows
from F. Kirwans thesis [117, Sect. 5] that H (M; Q) H (M0 ; Q) is surjective
(see e.g. [198, Theorem 2]). This justifies the terminology of surjectivity theorem
la Kirwan used above to introduce Propositions 7.6.13 and 7.6.15. For the assertion
on ker H i in these propositions, compare [199, Theorem 2]; our proofs followed
the hint of [199, Remark 3.5].

7.7 Exercises for Chapter 7

Notations. As in Sect. 7.1, G denotes the group with 2 elements G = {id, }. A

G-space is thus a space endowed with an involution . The notation S np stands for
the G-linear sphere S n with (S np )G = S p , as in Example 7.1.14.

7.1. Let G acting on X = S 1 with (z) = z . Prove that X G is homeomorphic to

the double mapping cylinder CCq , where q : S RP is the covering
projection. Prove that X G has the homotopy type of RP RP (use [82,
Proposition 0.17] and that S is contractible).
7.2. If p 1, prove that HG (S np ) admits, as a Z2 [u]-algebra, the presentation

HG (S np ) Z2 [u][A] (A2 ), where A is of degree n.
7.3. Let X be an equivariantly formal G-space which is of finite cohomology type.
Find a formula giving Pt (X G ) and Pt (u H (X G )) in terms of Pt (X ).
7.4. Write the details for Remark 7.1.21.
7.5. What is H G (S np )?
7.6. Let X = S d S d with d 1, endowed with the G action intertwining
the two  spheres. Prove that HG (X ) is, as a GrA[u]-algebra, isomorphic to
Z2 [u, a] (ua), with a of degree d. Prove that a 2 = 0.
7.7. Let  be the subgroup of SU (2) formed by the diagonal matrices. Prove that
the map B B SU (2) induced on the Milnor classifying spaces by the
inclusion is, up to homotopy type, equivalent to the inclusion CP HP .
7.8. Let  be a topological group acting on a space X . Let Y be a space of finite
cohomology type, considered as a -space with trivial -action (Y = Y  ).
Prove that H (X Y ) H (X ) H (Y ) (tensor product over Z2 ).
7.9. Let  be a compact Lie group. Let X and Y be two -space which are equivar-
iantly formal. We suppose that X is a finite dimensional -complex and that
Y is of finite cohomology type. Prove that X Y (with the diagonal -action)
is equivariantly formal.
7.10. Let Y be a G-space with Y G = Y . We suppose that Y is of finite cohomology
type. Let X = S01 Y , with the diagonal G-action. Give a presentation of
HG (X ) as a Z2 [u]-algebra. Describe, for HG , the Mayer-Vietoris sequence
analogous to that of Exercise 4.13. Describe the injective restriction homo-
morphism r : HG (X ) HG (X G ).
322 7 Equivariant Cohomology

7.11. Let X = S01 S01 , with the diagonal G-action. Give a presentation of HG (X )
as a Z2 [u]-algebra. Prove that the map f : X R given by f (ei , ei ) =
cos + 2 cos is an equivariant Morse function satisfying the hypotheses
of Theorem 7.6.6 and, with the help of this theorem, describe the injective
restriction homomorphism HG (X ) HG (X G ).
7.12. Find a connected equivariantly formal G-space X such that X G is the disjoint
union of a point and of a sphere.
7.13. For 0 p n, let Ppn denote the projective space RP n endowed with the
involution

(x0 : x1 : : xn ) = (x0 : : x p : x p+1 : : xn ).

(a) Prove that Ppn is G-equivariantly formal.

(b) Describe the restriction homomorphism r : HG (Ppn ) HG ((Ppn )G ) (it is
injective by (a) and Proposition 7.3.9).
(c) Prove that HG (P0n ) admits, as Z2 [u]-algebra, the presentations
 
HG (P0n ) Z2 [u][A] (An+1 +u An ) or Z2 [u][B] ((B+u)n+1 +u(B+u)n )

(d) Prove the GrA[u]-isomorphisms


HG (P1n ) and
(d.1) Z2 [u][A] (An+1 + u 2 An1 )
 12
(d.2) Z2 [u][A] (A + u A11 + u 4 A8 + u 5 A7 ) HG (P411 ).
7.14. Prove that the algebras
 
R = Z2 [u, A] (A3 + u A2 ) and S = Z2 [u, B] (B 3 + u 2 B)

are GrA[u]-isomorphic (A and B of the same degree). Find a G-space X such

that HG (X ) is GrA[u]-isomorphic to R, with A of degree 4.
7.15. Let Ppn be the G-space of Exercise 7.13. One checks that the function f :
Ppn R given by

f (x0 : x1 : : xn ) = x02 + x 2p

is a Morse-Bott function. Prove that it satisfies the hypotheses of Theo-

rem 7.6.6. With the help of the proof of this theorem (as in Example 7.6.9),
describe the restriction homomorphism r : HG (Ppn ) HG ((Ppn )G ). Com-
pare Exercise 7.13 (b).
7.16. What would be the analogue of Exercises 7.13 and 7.15 for S 1 -actions?
7.17. Let  be a topological group. Prove that any functor J from Top to Top extends
to a functor J from Top to Top . What is here the meaning of extends?
7.18. We apply Exercise 7.17 to the functor suspension X 
X . Let X be a
-space.
7.7 Exercises for Chapter 7 323

(a) Prove that there exists a suspension homomorphism

 : H  (X ) H +1 (
 X )

which is a morphism of H ( pt)-module and which is injective. Discuss its

functoriality.
(b) Find an example where
 is not surjective.
(c) Suppose that X is -equivariantly formal. Prove that
 X is -equivariantly
formal and that
 is an isomorphism.
7.19. Let X be a G-space. The G-action on X may be extended to a G-action on
X (note that

X , permuting the suspension points, giving rise to a G-space

X X G). Suppose that X is a connected finite dimensional G-complex
satisfying b(X ) = b(X G ) < . Let i : X
X denote the inclusion. Prove
that the sequence

H G i
0 / H (

X) / H (X ) / H (X ) / 0
G G

is exact.
7.20. Let  = SU (2) acting on X = SU (2) by conjugation.
(a) Show that X  has the homotopy type of the double mapping cylinder CC j
where j is the inclusion of CP HP (see Exercise 6.4). [Hint: use
that X/  is homeomorphic to a segment]
(b) Deduce from Exercise 6.4 that X is -equivariantly formal.
(c) Prove that there is a (unique) isomorphism of H ( pt)-algebras

H (X ) H ( pt)[b]/(b2 ) Z2 [a, b]/(b2 ),

where a is of degree 4 and b of degree 3.

7.21. Let (X, X 1 , X 2 , X 0 ) be a Mayer-Vietoris data. Suppose that X is a -space ( a
topological group) and that X i are closed -invariant subspaces of X . Suppose
that X = X 1 X 2 and that (X i , X 0 ) is a -equivariantly well cofibrant pair for
i = 1, 2. Prove that there is a Mayer-Vietoris sequence for the -equivariant
cohomology.
7.22. Let X and Y be -spaces, equivariantly well pointed by x X  and y Y  .
Thus, X Y (using these base points) is a -space. Prove that there is a
H -algebra isomorphism H  (X Y ) H  (X ) H  (Y ).
Chapter 8
Steenrod Squares

In Chap. 4, the power of cohomology was much increased by the introduction of the
cup product, making H (X) a graded algebra. Another rich structure on H (X) comes
from cohomology operations, i.e. the natural self-transformations of the mod 2 coho-
mology functor (see Sect. 8.1). The basic examples of such operations, the Steenrod
squares Sqi : H (X) H +i (X), were discovered by Norman Steenrod and Henri
Cartan in the late 1940s (see, e.g. [40, pp. 510523] for historical details). The GrA-
morphism induced by any continuous map must then commute with all the Steenrod
squares, which imposes strong restrictions. For instance, the spaces Y = S 2 S 3
and Y  = RP2 do not have the same homotopy type, although their cohomology
are GrA-isomorphic. Indeed, Sq1 vanishes on H (Y ) but not on H (Y  ). In the same
way, we show that all suspensions of Hopf maps are essential, i.e. not homotopic to
a constant map (see Sect. 8.6).
After an introductory section on cohomology operations, we state in Sect. 8.2
the basic properties and make some computations of Steenrod squares. Their con-
structions and the proof of Adem relations are given in Sects. 8.3 and 8.4. Based on
equivariant cohomology, these two technical sections may be skipped on first read-
ing, since the applications of Steenrod squares are consequences of the properties
presented in Sect. 8.2.
The last two sections of this chapter treat applications of Steenrod squares. Promi-
nent among them are Adams theorem on the Hopf invariant one problem and Serres
computation of the cohomology algebra of Eilenberg-MacLane spaces K(Z2 , n).
The latter implies that mod 2 cohomology operations are, in some sense, generated
by sums, cup products and iterations of Steenrod squares (see Remark 8.5.7). More
applications will appear in Chap. 9, for instance Thoms definition of Stiefel-Whitney
classes and Wus formula.

Springer International Publishing Switzerland 2014 325

J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_8
326 8 Steenrod Squares

8.1 Cohomology Operations

This section contains some generalities on mod 2 cohomology operations, in order

to present Steenrod squares. We take a global approach which may shed a new light
with respect to existing texts on the subject.
A cohomology operation is a map

Q = Q(X,Y ) : H (X, Y ) H (X, Y )

defined for any topological pair (X, Y ), satisfying the following two conditions:
(1) Q is functorial, i.e. if g : (X  , Y  ) (X, Y ) is a continuous map of pairs, then

H g Q = Q H g . (8.1.1)

(2) Q(X,Y ) = Q[i](X,Y ) where Q[i](X,Y ) is the restriction of Q(X,Y )
to H i (X, Y ).
We may restrict the definition to some classes of pairs, like CW-pairs, etc. For
instance, restricting to pairs (X, ) gives operations on absolute cohomology, since

H (X, )
H (X). Point (2) is a partial linearity (Q is not supposed to be linear)
and permits us to define Q via its restrictions Q[i] .
Examples of cohomology operations are given by Q = 0 or Q = id . A less
trivial example is the cohomology operation Q such that by Q[n] (a) = an for all
n N, where an = a   a (n times). Cohomology operations may be added,
multiplied by cup products and composed, giving rise to more examples.
Here are a few remarks about cohomology operations. They are used throughout
this section, without always an explicit mention.

8.1.1 By Theorem 3.7.1, a topological pair has, in a functorial way, the same coho-
mology as a CW-pair. Hence, when studying cohomology operations, we do not lose
generality by restricting to CW-pairs. For instance, a cohomology operation defined
for CW-pairs extends in a unique way to a cohomology operation defined for all
topological pairs.

8.1.2 Let (X, Y ) be a CW-pair with Y non-empty. The quotient map (X, Y )

(X/Y , [Y ]) induces an isomorphism H (X/Y , [Y ]) H (X, Y ) (Proposition
3.1.45). Most questions on cohomology operations may thus be settled by consider-
ing the CW-pairs of type (X, ) and (X, pt). In particular, a cohomology operation
defined for these pairs extends to a unique cohomology operation. Moreover, a coho-
mology operation Q defined on absolute cohomology for CW-complexes extends to
a unique cohomology operation for CW-pairs, using the commutative diagram
8.1 Cohomology Operations 327

0 / H (X, pt) / H (X) / H (pt) / 0

Q Q Q , (8.1.2)
  
0 / H (X, pt) / H (X) / H (pt) / 0

where (X, pt) is a CW-pair.

8.1.3 Let (X, Y ) be a CW-pair. By Corollary 3.1.12, the family of inclusions iA : A

X for A 0 (X) gives rise to the commutative diagram

(H iA ) 
H (X, ) / A0 (X) H
(A, )


Q Q
 
(H i A) 
H (X, ) / A0 (X) H
(A, emptyset)

or, if pt A0 0 (X),

(H iA ) 
H (X, pt) / H (A0 , pt) A0 (X)A0 H (A, )


Q Q

(H i A)  
H (X, pt) / H (A0 , pt) A0 (X)A0 H (A, )

In other words, a cohomology operation preserves the connected components.

Together with 8.1.2, this permits us often to restrict, without loss of generality, a
cohomology operation to pairs (X, Y ) where X is path-connected.

Lemma 8.1.4 If Q be a cohomology operation, then Q(0) = 0.

Proof The class 0 H (X, Y ) is in the image of H (X, X) H (X, Y ). As,

H (X, X) = 0, the lemma follows from functoriality. 

An important property of cohomology operations is that it does not decrease

dimensions.

Lemma 8.1.5 Let Q be a cohomology operation. Then, there is a function N : N

N, satisfying N(0) = 0 and N(m) m, such that


N(m)
Q(H m (X, Y )) H k (X, Y ) (8.1.3)
k=m

for all topological pairs (X, Y ).

328 8 Steenrod Squares

Proof By [Link].3 above, it is enough to prove the lemma for CW-pairs (X, Y )
with X connected and Y = pt or . As H 0 (X, pt) = 0, Q(H 0 (X, pt)) = 0 by Lemma
8.1.4. The constant map X pt induces an isomorphism H 0 (pt, ) H 0 (X, ).
As H >0 (pt, ) = 0, the functoriality implies (8.1.3) for m = 0, with N(0) = 0.

If m > 0, then H m (X, pt)
H m (X, ) H m (X), so it is enough to prove (8.1.3)
in the absolute case. By functoriality, the following diagram is commutative.

H m (X, X m1 ) / / H m (X)

Q Q
 
H (X, X m1 ) / H (X)

As H k (X, X m1 ) = 0 for k < m, this proves that the direct sum in (8.1.3) starts
at k = m. Also, any class a H m (X) is of the form a = H f () for some map
f : X Km . Thus, N(m) is the maximal degree of Q() H (Km ). 

A cohomology operation Q restricts to a cohomology operation Q on absolute

cohomology by QX = Q(X,) . Not every absolute cohomology operation Q is such
restriction because it may not satisfy Q (0) = 0, contradicting Lemma 8.1.4. As
an example, we can, for X path-connected and non-empty, define Q by Q (a) = 0
for a H m (X) (m = 1) and Q (H 1 (X)) = 1 (the functoriality of Q coming from
Lemma 2.5.4). In fact, we have the following lemma.

Lemma 8.1.6 Let Q be a cohomology operation defined on absolute cohomology

for connected CW-complexes. Then, there exists a cohomology operation Q such that
QX = Q(X,) if and only if Q (0) = 0. The cohomology operation Q is unique.

Proof The condition is necessary by Lemma 8.1.4. For the converse, using 8.1.1
8.1.3 above, it suffices to define Q(X,pt) for an path-connected non-empty CW-
complex X. On H 0 (X, pt) = 0, Q is defined by Q(0) = 0 (this is compulsory
by Lemma 8.1.4). The functoriality for the inclusion (X, )  (X, pt) on H 0 is
guaranteed by the condition Q (0) = 0. Let P : H (X) H 0 (X) be the projection
onto the component of degree 0. Let j : pt X be an inclusion of a point in X. As
Q (0) = 0, the commutative diagram

Q
H >0 (X) / H (X) P / H 0 (X)

Hj Hj Hj
  
Q =0
H >0 (pt) / H (pt) P / H 0 (pt)

shows that Q (H >0 (X)) H >0 (X). Hence, the commutative diagram
8.1 Cohomology Operations 329

/ H >0 (X)
H >0 (X, pt)
Q Q
 
/ H >0 (X)
H >0 (X, pt)

defines Q and shows its uniqueness. 

The notion of cohomology operation makes sense for the reduced cohomology,
with the same definition.

Lemma 8.1.7 A cohomology operation Q descends to a unique cohomology oper-

ation on reduced cohomology, also called Q.

Proof Let p : X pt be the unique map from X to a point. Consider the diagram

Hp
H (pt) / H (X) / / H (X)

Q Q Q
  
Hp
H (pt) / H (X) / / H (X)

where the line are exact. As Q is a cohomology operation, the left square is commuta-
tive, so there is a unique Q : H (X) H (X) so that the right square commutes and
this construction is functorial. (Recall that, if X is path-connected and Y is non-empty,
then H (X, Y ) = H (X, Y )). 

We now study the multiplicativity of a cohomology operation Q. Note that, by

Lemma 4.1.14, the relative cup product H (X, Y )H (X, Y ) H (X, Y ) is defined
for all topological pairs (X, Y ). Consider the following four statements.
(a) Q(a  b) = Q(a)  Q(b) for all a, b H (X) and all spaces X.
(b) Q(a  b) = Q(a)  Q(b) for all a, b H (X, Y ) and all topological pairs
(X, Y ).
(c) Q(a b) = Q(a) Q(b) for all a H (X1 ), b H (X2 ) and all spaces X1 and
X2 .
(d) Q(a Q(b) for all a H (X1 ), b H (X2 ) and all pointed spaces
b) = Q(a)
X1 and X2 .

Proposition 8.1.8 For a cohomology operation Q, Conditions (a), (b) and (c) are
equivalent and (a) implies (d). If Q(1) = 1, then (d) implies (a).

Proof Without loss of generality, we may suppose that the spaces X and Xi are con-
nected CW-complexes. Statement (b) is stronger than (a) since H (X) = H (X, ).
To prove that (a) implies (b), it suffices to consider the case Y = pt, which is obvious.
Using the functoriality of Q, (a) (c) follows from the definition of the cross
product and (c) (a) from the formula a  b =  (a b) (see Remark 4.6.1).
330 8 Steenrod Squares

That (c) (d) is obvious, so (a) (d). Now, (d) implies (c) for classes of positive
degree. As (c) (a), property (d) implies that Q(a  b) = Q(a)  Q(b) except
possibly for a or b equal to 1. If, say a = 1, then

Q(1  b) = Q(b) = 1  Q(b) = Q(1)  Q(b) ,

since Q(1) = 1. Thus, (d) (a) if Q(1)) = 1. 

Corollary 8.1.9 Let Q be a cohomology operation with Q(1) = 1. Then, (a) is
equivalent to
(d) the diagram

H m (Km ) H n (Kn ) / H m+n (K K )
m n

QQ Q
 


H (Km ) H (Kn )
/ H (K K )
m n

is commutative for all positive integers m and n.

Proof It is clear that (d) (d). The corollary will then follow from Proposition
8.1.8 if we prove that (d) (d).
It suffices to prove (d) for a H m (X1 ) and b H n (X2 ) where X1 and X2 are
connected CW-complexes, so m, n > 0. Then a = fa (m ) and a = fb (n ) for maps
fa : X1 Km and fb : X2 Kn . Condition (d) says that Q(m n ) = Q(m ) Q(
n)
and Q(a b) = Q(a) Q(b) from this special case, using the functoriality of and
of Q. 

8.2 Properties of Steenrod Squares

One of the most remarkable feature of mod 2 cohomology is the existence of coho-
mology operations, introduced by N. Steenrod and H. Cartan in the late 1940s (see,
e.g. [40, pp. 510523]), called the Steenrod squares Sqi : H (X, Y ) H (X, Y )
(i N). For a H (X, Y ),
m one has Sqi (a) = 0 for i > m (see 2a in Theorem 8.2.1).
Hence, the sum iN Sq (a) has only a finite number of non-zero terms and thus
i

defines the total Steenrod square


Sq : H (X, Y ) H (X, Y ) , Sq(a) = Sqi (a) . (8.2.1)
iN

Here is the main theorem of this section.

Theorem 8.2.1 There exists a cohomology operation Sq and Sqi as in (8.2.1), which
enjoys the following properties:
8.2 Properties of Steenrod Squares 331

(1) Sq is Z2 -linear.
(2) if a H n (X, Y ) then Sqi (a) H n+i (X, Y ) and
(a) Sqi (a) = 0 for i < 0 and i > n.
(b) Sq0 (a) = a.
(c) Sqn (a) = a  a.

(3) Sq(a  b) = Sq(a)  Sq(b). This is equivalent to the formula


Sqk (a  b) = Sqi (a)  Sqj (b) (Cartans formula).
i+j=k

(4)  Sq = Sq  , where  : H (X)
H +1 (X) is the suspension isomor-
phism of Proposition 3.1.49.
(5) The Adem relations:
[i/2]
 jk1
Sqi Sqj = i2k Sqi+jk Sqk (0 < i < 2j) .
k=0

The Steenrod squares are characterized amongst cohomology operations by some

of these properties (see Proposition 8.5.12). Also, the Adem relations generate all
the polynomial relations amongst the compositions of Sqi s which hold true for any
space (see Corollary 8.5.11).

Example 8.2.2 Theorem 8.2.1 permits us to compute Sqi easily for the projective
spaces RPn , CPn and HPn . Indeed, one has the following results.

(a) Let a H 1 (X). Then (2) implies that Sq(a) = a + a2 (we write an for the cup
product of n copies of a). Then, (3) implies that

n
n i
Sq(an ) = (a + a2 )n = an  (1 + a)n = an  i a .
i=1

Therefore
n n+i
Sqi (an ) = i a . (8.2.2)

(b) If a H 2 (X) satisfies Sq1 (a) = 0, then Sq(a) = a + a2 and, as in (a), one has

Sq2i (an ) = ni an+i and Sq2i+1 (an ) = 0 . (8.2.3)

(c) If a H 4 (X) satisfies Sq(a) = a + a2 , then

 
n n+i
a if k 0 mod 4.
Sq4i+k (an ) = i (8.2.4)
0 otherwise.
332 8 Steenrod Squares

Besides the trivial case of OP2 , there are no more such examples. Indeed, by Corollary
8.6.3 and Theorem 8.6.6, if a H m (X) satisfies Sq(a) = a + a2 with a2 = 0, then
m = 1, 2, 4 or 8.

We finish by two more properties of Steenrod squares. As Sq(a  b) = Sq(a) 

Sq(b), Proposition 8.1.8 implies the following result.

Proposition 8.2.3 Let X1 and X2 be topological spaces [pointed for (b)]. Then,
(1) Sq(a b) = Sq(a) Sq(b) for all a H (X1 ), b H (X2 ).
(2) Sq(a Sq(b) for all a H (X1 ), b H (X2 ).
b) = Sq(a)

Proposition 8.2.4 Let (X, Y ) be a topological pair. Then

Sq = Sq ,

where is the connecting homomorphism : H (Y ) H +1 (X, Y ).

Proof By 8.1.1 , we may suppose that (X, Y ) is a CW-pairs. Let Z = X [2, 1]

Y and Z  = X [2, 1] Y {1} Y . The projection p : [2, 1] Y
{2} Y extends to a homotopy equivalence of pairs p : (Z, {1} Y ) (X, Y ).
The commutative diagram

p
H ({1} Y ) o o H (S 0 Y )

 
H +1 (Z, {1} Y ) o o / H +1 (D1 , S 0 Y )
H +1 (Z, Z  ) excision

shows that it is enough to prove that Sq = Sq for a CW-pair of the type

(D1 Y , S 0 Y ). In the proof of Proposition 4.7.44, we showed that
 
(a) = e H i+ (a) + H i (a) ,

where i : {1}F S 0 F denote the inclusions and 0 = e H 1 (D1 , S 0 ) = Z2 .

One has Sq(e) = Sq0 (e) = e. Using the linearity of Sq and Proposition 8.2.3, we get
  
Sq (a) = Sq e H i+ (a) + H i (a)
= Sq(e) Sq(H i+ (a) + H i (a))
 
= e Sq H i+ (a) + Sq H i (a)
 
= e H i+ Sq(a) + H i Sq(a)
= Sq(a) . 
8.2 Properties of Steenrod Squares 333

Sections 8.3 and 8.4 contain the proof of Theorem 8.2.1. They are based on the
ideas of Steenrod (see [184, VII.1 and VIII.1]) using the equivariant cohomology.
Other treatments of similar ideas are developed in [3, VI.7] and [82, Sect. 4.L].

8.3 Construction of Steenrod Squares

The involution on X X given by (x, y) = (y, x) makes X X a G-space

for G = {id, }. We consider the cross-square map : H n (X) H 2n (X X)
defined by (a) = a a. Its image is obviously contained in H 2n (X X)G . By
Lemma 7.1.10, the image of : HG (X X) H (X X) is also contained in
H (X X)G . The same considerations are valid for the reduced cross-square map
: H n (X) H 2n (X X)G , defined for a space X which is well pointed by x X. The
maps and are not linear but they are functorial: if f : X  X is a continuous maps,
then H (f f ) = H f and H (f f ) = H f . Using Diagram (4.7.7),
the use of base points x X and (x, x) X X provide a commutative diagram

H (X X) / / H (X X)
O O
(8.3.1)

H (X) / / H (X)

Lemma 8.3.1 Let X be a connected CW-complex, pointed by x X 0 . Then, the

cross-square maps and admit liftings G and G so that the diagram

H (X X) / / H (X X)
O fNNN O fNNN
NNN NNN
N N
NNN
NN

H G (X X) / H (X X)
G
8 8
G G

H (X) / / H (X)

is commutative. These liftings are functorial and satisfy G (0) = 0 and G (0) = 0.
Such liftings are unique.

Proof The maps , , and preserving the connected components, one may sup-
pose that X is connected. Lemma 8.3.1 is obvious when n = 0, giving G (0) = 0
and G (1) = 1. We can then assume n > 0. In this case, H n (X) H n (X) is an
isomorphism, so it is enough to define G .
334 8 Steenrod Squares

We first define G when X = Kn = K(Z2 , n) with its CW-structure given in

the proof of Proposition 3.8.1, whose 0-skeleton consists in a single point x. By
Proposition 4.7.11, the G-space Kn Kn satisfies H k (Kn Kn ) = 0 for k < 2n.
Hence, Proposition 7.1.12 implies that : H 2n 2n
G (Kn Kn ) H (Kn Kn ) is an
G

isomorphism. We define G = 1 .

Now, a cohomology class a H n (X) is of the form a = H fa () for a map

fa : X Kn , well defined up to homotopy. We define

G (a) = H G (fa fa ) G () . (8.3.2)

This definition makes G functorial. Indeed, let g : Y X be a continuous map

and a = H fa () H n (X). If b = H g(a), then b = H fb () with fb = fa g. By
definition of G , one has

H G (g g) G (a) = H G (g g) H G (fa fa ) G () = H G (fb fb )G () = G (b) .

The functoriality of G follows from that of G . The uniqueness of G (then, that

of G ) is obvious, since there was no choice for X = Kn and Definition (8.3.2) is
compulsory by the required functoriality. 

Remark 8.3.2 When H i (X) = 0 for i < n, then H j (X X) = 0 for j < 2n by

Proposition 4.7.11. The homomorphism : H 2n (X X) H 2n (X X)G is an
isomorphism which is natural. Then, by functoriality, the formula G = 1
holds true.

The map G : H n (X) HG2n (X X) extends to G : H (X) HG (X X) by

 
G ( aj ) = G (aj ) , where aj H j (X) . (8.3.3)
jN jN

The inclusion of (X X)G into X X induces a GrA[u]-morphism r : HG (X X)

HG ((X X)G ). Observe that (X X)G is the diagonal subspace {(x, x)} of X X,
hence homeomorphic to X. We thus write r : HG (X X) HG (X), considering X
as a G-space with trivial G-action. Using (7.1.2), we thus get a GrA[u]-isomorphism
HG (X) H (X)[u]. We also consider the (non-graded) ring homomorphism
ev1 : H (X)[u] H (X) which extends the identity on H (X) by sending u to
1 (evaluation of a polynomial at 1).
Let X be a CW -complex. By definition, the Steenrod square Sq : H (X) H (X)
is the composition

Sq = ev1 r G (8.3.4)

making the following diagram commutative.

8.3 Construction of Steenrod Squares 335

/ H (X) o / H (X)[u]
HG (X X) r
G
O
G ev1 . (8.3.5)
Sq

H (X) / H (X)

By (8.3.3) and Lemma 8.3.1, the map Sq is a cohomology operation, defined so far
on absolute cohomology for connected CW-complexes. As G (0) = 0, Lemma 8.1.6
implies that this partial definition of Sq extends to a unique cohomology operation
Sq : H (X, Y ) H (X, Y ) defined for all topological pairs (X, Y ).
For a H n (X, Y ) let Sqi (a) be the component of Sq(a) in H n+i (X, Y ). Again,
defining Sqi : H (X, Y ) H (X, Y ) by
 
Sqi ( aj ) = Sqi (aj ) , where aj H j (X) (8.3.6)
jN jN

provides a family of cohomology operation Sqi . A priori, i Z

but, by Lemma 8.1.5,
Sqi = 0 if i < 0. It follows from these definitions that Sq = iN Sqi .
By Lemma 8.1.7, the Steenrod squares Sq and Sqi are also cohomology operations
on reduced cohomology. In this case, the following diagram is commutative.

/ H (X) o / H (X)[u]
H G (X X)
r
G
O
G ev1 (8.3.7)

Sq
H (X) / H (X)

In the important case where H j (X) = 0 for j < n (e.g. X = Kn ), one ca use
Remark 8.3.2 to get the following commutative diagram

1
H 2n (X X)G / H 2n (X X) r / H (X) o / H (X)[u]
O G G
o ooo7
ooo ev1 (8.3.8)
ooo G Sq

H n (X) / H (X)

We now prove Properties (1)(4) of Theorem 8.2.1, for the absolute cohomology
H (X) of a connected CW-complex X.
Proof of (2). That Sqi sends H n (X) to H n+i (X) is by definition and we already
noticed that Sqi = 0 for i < 0. If a H n (X), then r G (a) HG2n (X), which implies
that Sqi = 0 for i > n. Note that, from (8.3.4) and the definition of the Sqi , one has,
for a H n (X):
336 8 Steenrod Squares


n
r G (a) = Sqi (a) uni . (8.3.9)
i=0

Let us prove that Sqn (a) = a  a. From the above equation, we deduce that
Sqn (a)= ev0 r G (a). Using Diagrams (7.1.5) and (7.1.6), we get the diagram

G
H (X) / H (X X) r / H (X) o / H (X)[u]
G G
O
= ev0

 
/ H (X X)  / H (X)
H (X)

where  is induced by the diagonal map  : X X X. By (4.6.5) p. 157, one

has

Sqn (a) = ev0 r G (a) =  (a) =  (a a) = a  a .

It remains to prove that Sq0 (a) = a. By naturality and since H n (Kn ) = Z2 , it

suffices to find, for each integer n, some space X with a class a H n (X) such that
Sq0 (a) = 0. The space X will be the sphere S n . Indeed, there is a homeomorphism

h : Sn Sn S 2n (see Example 4.7.12). We leave as an exercise to the reader to
construct such a homeomorphism h which conjugates the G-action on S n S n to a
linear involution on S 2n . Therefore, S n S n is G-homeomorphic to the sphere Sn2n
of Example 7.1.14. We now use Diagram (8.3.8). By the reduced Knneth theorem,

(a) = a , the generator of H 2n (S n S n ) = Z2 . By Proposition 7.1.15, r 1 (a) =
a u , whence Sq0 (a) = a.
n

Linearity of Sq. We have to prove that for each n N, the map Sq : H n (X)

H (X) is linear. By Point (2) already proven, the restriction of Sq to H 0 is Sq0 = id,
so we may assume that n 1. By functoriality of Sq, the diagram

H n (X/X n1 ) / / H n (X)

Sq Sq
 
H (X/X n1 ) / H (X)

is commutative, where the horizontal maps are induced by the projection X

X/X n1 . Therefore, it is enough to prove the linearity of Sq : H n (X/X n1 )
H (X/X n1 ). We may thus assume that H k (X) = 0 for k < n and use
Diagram (8.3.8) to define Sq. It is then enough to show that r 1 : H n (X)
HG2n (X) is linear. One has

+ b) = (a)
(a
+ (b) +a b+b a

b + (a
= (a) + (b) + a b) .
8.3 Construction of Steenrod Squares 337

Using Proposition (7.1.12), we get

+ b) = (a)
1 (a
+ (b) + tr (a
b) .

As r tr = 0 by Proposition 7.1.9, this prove that r 1 is linear.

From the already proven properties of Sq, we now deduce a structure result about
HG (X X) (Proposition 8.3.3) which will be used to prove the multiplicativity of
Sq. Let N : H (X X) H (X X) be the GrV-morphism defined by N(x y) =
x y + y x and let N be the image of N. Note that

ker N = H (X X)G = D N

where D is the subgroup generated by {x x | x H (X)}. By definition of the

transfer map tr : H (X X) HG (X X) [see (7.1.7)], one has

tr(x y) = x y + (x y) = N(x y) .

The correspondence (x x, N(y z))  G (x) + tr (y z) produces a section

: D N HG (X X) of : HG (X X) H (X X)G . We identify D N
with (D N ) and thus see D N as a subgroup of HG (X X). Let D be the
Z[u]-module generated in HG (X X) by D. The following result is due to Steenrod
(unpublished, but compare [78, Sect. 2]).

Proposition 8.3.3 With the above identifications, the following properties hold true.
id
(a) The restriction of to D N coincides with the identity D N H (X X)G .
(b) N = Ann (u).
(c) D is a free Z2 [u]-module with basis D {0}. In particular, D is isomorphic to
Z2 [u] D.
(d) As a Z2 [u]-module, HG (X X) = D N .

Proof Point (a) is obvious from the identification via . Hence, H (X X)G is the
image of and one has the commutative diagram

/ H (X X)/(u) / H (X X) tr / Ann (u) / 0

0 G
O
=
  
0 / H (X X)G / H (X X) / N / 0

where the lines are exact (the upper line is the transfer exact sequence (7.1.8)). By the
techniques of the five lemma, we deduce that : Ann (u) N is an isomorphism.
As (N) is contained in the image Ann (u) of the transfer map, this proves (b). Also,
we get the isomorphism
338 8 Steenrod Squares

: HG (X X)/(u)
H (X X)G . (8.3.10)

Let B be Z2 -basis H (X) formed by homogeneous classes. To prove (c), one has
to show that

G (a)uk(a) = 0 . (8.3.11)
aB0

for any non-empty finite subset B 0 of B and any function k : B 0 N. Let Bmin
0 be

the subset formed by the elements in B which are of minimal degree. Then
0

   
ev1 r G (a)uk(a) = Sq(a) = a + terms of higher degrees = 0 .
aB0 aB0 aBmin
0

To prove (d), let A = HG (X X) and B = D N . If B = A, let a A B be

of minimal degree. By the above and Sequence (7.1.8), one has B/uB = A/uA
H (X X)G . Hence, there exists b B and c A such that a = b + uc. By the
minimality hypotheses on a, one has c B and thus a B (contradiction). 

Proof that Sq(a  b) = Sq(a)  Sq(b) (multiplicativity). One has

(a  b) = (a  b) (a  b) = (a  a) (b  b) = (a)  (b) .

Hence,

G (a  b) = G (a)  G (b) + x

with x ker = (u) [the last equality was established in (8.3.10)]. We may suppose
that a H m (X) and b H n (X). Let V be Z2 -basis H <m+n (X) formed by homoge-
neous classes. By Proposition 8.3.3, x = vV 0 G (a)uk(v) for some finite subset
V 0 of V, with k(v) > 0. Let Vmin
0 be the subset formed by the elements in V 0 which

are of minimal degree. Then,


Sq(a  b) = ev1 r G (a  b) = a + terms of higher degrees .
vVmin
0

Since Sqi (a  b) = 0 for i < 0, one has Vmin0 = and therefore V 0 = . This

implies that G (a  b) = G (a)  G (b) and thus Sq(a  b) = Sq(a)  Sq(b).

Proof that  Sq = Sq  . By Lemma 4.7.13, using the reduced suspension
S X, the relation  Sq = Sq  is equivalent to the following equation in
1

H (S 1 X):

b c) c H n (X) ,
Sq(c) = Sq(b (8.3.12)
8.3 Construction of Steenrod Squares 339

where b is the generator of H 1 (S 1 ) = Z2 . As noticed in Proposition 8.2.3, the formula

Sq(b  c) = Sq(b)  Sq(c) already proven implies that Sq(b c) = Sq(b) Sq(c).

Also, by (2) already established, Sq(b) = b. Hence, Eq. (8.3.12) holds true.
The proof of Points (1)(4) in Theorem 8.2.1 is now complete for pairs (X, )
with X a connected CW-complex. We check easily that the extension of Sq and Sqi
to topological pairs given by Lemma 8.1.6 satisfies the same properties.

8.4 Adem Relations

The Adem relations are relations amongst the compositions Sqi Sqj . They were con-
jectured by Wu wen-Tsn around 1950 and first proved in 1952 by Adem in his
thesis at Princeton University (summary in [4] and full proofs in [5]). We present
below a proof based on the idea of Steenrod [184, Chap. VIII], using the equivariant
cohomology for the symmetric group Sym4 . The proof in [82, Sect. 4.L] is another
adaptation of the same idea. For different proofs, see [33] and Remark 8.5.10.
Let X be a topological space. Consider the map

G
: H (X)
Sq
r
HG (X X)
HG ((X X)G ) H (BG X)

(recall that BG RP : see Example 7.2.1). The map Sq would sit diagonally
is
in (8.3.5), the diagram used to define Steenrod squares. By Sect. 8.3, the map Sq
functorial, Z2 -linear and multiplicative.
Consider now the iterated map

Sq
Sq : H (X) H (BG BG X) H (X)[u, v] ,

using the Knneth theorem and that H (BG BG) Z2 [u, v] with u and v in
degree 1.
Sq(a)
Proposition 8.4.1 For any a H (X), the polynomial Sq is symmetric in
the variables u and v.

Before proving Proposition 8.4.1, we do some preliminaries. Let Y be a K-space

for a topological group K. The equivariant cross product K of Sect. 7.4 gives rise to
an equivariant cross square map K : HKn (Y ) HK2n (Y Y )G , defined by K (a) =
a K a, where G = {I, } acting on Y Y by exchanging the factors. A map
K : H nK (Y ) H 2nK (Y Y ) is similarly defined, using the reduced equivariant
G

cross product K . The following is a generalization of Lemma 8.3.1.

340 8 Steenrod Squares

Lemma 8.4.2 Let Y be a K-space, equivariantly well pointed by y Y K . Then,

(1) The cross-square maps K and K admit liftings GK and GK so that the diagram

H 2n G / / H 2n (Y Y )G
K (Y Y )
gNNN
K
O gOOO O
OOO NNN
OOO K NNN
H 2n / HGK
2n (Y Y )
K GK (Y Y )
G
K 7 G
K 7

H nK (Y ) / / HKn (Y )

is commutative, where and are induced by the homomorphism K G K.

When K is the trivial group, these lifting coincide with those of Lemma 8.3.1.
(2) The lifting GK is functorial in Y and G, i.e. if Y  is a K  -space, f : Y Y  is
equivariant with respect to a continuous homomorphism : K K  , then the
diagram
f f
2n (Y  Y  )
HGK 
/ H 2n (Y Y )
GK
O O

G
K G
K

f
HKn  (Y  ) / H n (Y )
K

is commutative. The analogous property holds for G .

(3) Suppose that the K-action on Y is trivial. Then, the following diagram is com-
mutative
2n (Y Y ) r / H 2n (Y ) / H 2n (B(G K) Y )
HGK GK
O O
KG P

Sq
HKn (Y ) o / H n (BK Y ) / H 2n (BG BK Y )

Proof The lifting GK is defined using the lifting G of Lemma 8.3.1 and the com-
mutative diagram
8.4 Adem Relations 341

H 2n (Y Y )
r / HGK
2n (Y )
f GK O
ffffff
fffff O
ffff
fffff

ff ff
rff
HK2n (Y Y )G o / H 2n ((Y Y )K )G o HG2n ((Y Y )K )
r / H 2n (YK )
=
O O O
G
O
K K

H 2n ((Y Y )KK )G o / H 2n ((Y )K )
r
HG2n ((Y Y )KK )
O O
G
O
K P P

H 2n (YK YK )G o / H 2n ((Y )K ))
r
HG2n (YK YK )
O l6 fff 2 G
G lll
l ffff ffff
l f

lll ffff f
lllffffffffff Sq
HKn (Y ) o
= / H n (YK ) ff
(8.4.1)

(we do not need the last column for the definition of GK but it will be used later). The
commutativity of the right rectangle is the definition of K , using the formula (7.4.7)
for the equivariant cross product. The top vertical isomorphisms come from the
following fact: if Z is a (G K)-space,
 the homotopy
 equivalence E(G K)
Z EG EK Z given by (ti (gi , ki ), z  (ti gi , z), (ti ki , z) descends to
a homotopy equivalence E(G K)GK Z EG G (EK K Z). The same
homotopy equivalence is used to get the map

ZK = EK K Z EG G (EK K Z) E(G K) GK Z = ZGK (8.4.2)

For GK , we use the commutative

giving rise (for Z = Y Y ) to the homomorphism .
diagram

ee H 2n
ee eeeeeeeee GK (Y Y )
O
e e e e

e e e ee eeeee (2)
eee
reeeeee
H 2n
K (Y Y )
G o / H 2n ((Y Y )K , (Y Y )K )G o 2n ((Y Y ) , (Y Y ) )
HG K K
O (1) O O

K 
K

H 2n ((Y Y )KK , (Y Y )KK )G o 2n ((Y Y )
HG KK , (Y Y )KK )
O O
K P

Y K )G
H 2n (YK YK , YK o HG K K K YK )
2n (Y Y , Y
O
iii4

iiiiiii
i G
o / iiii
H nK (Y ) H n (YK , {y}K )

(8.4.3)
342 8 Steenrod Squares

As Y is equivariantly well pointed by y, the pair (YK , {y}K ) is well cofibrant, so the
cross-square map is defined. Also, the pair (Y Y , Y Y ) is K-equivariantly
well cofibrant (the proof is the same as for Lemma 7.2.12). Hence, Identification
(1) then comes from Corollary 7.2.16. The same argument gives Identification (2),
using (8.4.2) for Z = Y Y and Z = Y Y . The lifting G is defined using the
following diagram.

H K (Y K Y K )G / YK )G
HK2n (YK YK , YK
O fNNN O iRRR
NNN RRR
NNN RRR
RR
H 2n
G (YK YK )
/ HGK K YK )
2n (Y Y , Y
K K
p8 5
ppppp
ppp G G

H n (Y K )

/ H (YK , {y}K )

With these definitions, Diagram (8.4.3) is mapped into Diagram (8.4.1), giving rise
to the diagram of Point (1) of our lemma. That GK and GK coincide with G and G
when K is the trivial group follows from the definitions, as well as Point (2). Point
(3) comes from (8.4.1), where the occurrence of Sq is noticed. 

Proof of Proposition 8.4.1 By naturality, it is enough to prove the proposition for

X = Kn and a the generator of H n (Kn ) = Z2 (n 1). We consider X as pointed
by x X 0 , using the CW-structure given in the proof of Proposition 3.8.1, whose
0-skeleton consists in the single point x.
Let the symmetric group  = Sym4 act on X 4 = X X X X and X 4 =
X X X X by permutation of the factors. This action may be restricted to the
subgroup of  of  generated by s = (1, 2)(3, 4) and t = (1, 3)(2, 4). As for (8.3.1),
the use the base points x X and (x, x, x, x) X X provide a commutative diagram

H (X 4 ) / / H (X 4 )
O O
. (8.4.4)

H (X) / / H (X)

By Proposition 4.7.11, H k (X 4 ) = 0 for k < 4n and H 4n (X 4 ) = H 4n (X 4 ) = Z2 .

As in Lemma 8.3.1, using Proposition 7.2.17, we get liftings

 : H n (X) H 4n (X 4 ) and ( ) : H n (X) H 4n (X 4 )

( )  

of and . Consider the composite map.

( ) r
 : H n (X) H (X )
4n 4
H ((X 4 ) ) H (X) H (B) .
8.4 Adem Relations 343

Let  be the subgroup of  generated by s = (1, 2)(3, 4) and t = (1, 3)(2, 4). As s
and t commute,  is isomorphic to G G. Let i :   denote the inclusion. We
shall prove that the diagram

H (X)
9 H (B)
idH Bi
 
H (X) H (B0 ) . (8.4.5)
O

Sq
Sq 
H n (X) / H (X)[u, v]

is commutative. For the moment, let us show that this property implies Proposition
8.4.1. It is enough to prove that the image of H (B) Z2 [u, v] consists of sym-
metric polynomials. Under the isomorphism Z2 [u, v] H (B) the automorphism
exchanging u and v corresponds to that induced by exchanging s and t. But in ,
exchanging s and t is achieved by the conjugation by the transposition (2, 4). Such
an inner automorphism of  induces the identity on H (B) by Proposition 7.2.3.
It remains to prove that Diagram (8.4.5) is commutative. Let G1 and G2 be the
subgroups of  generated by s and t respectively, so  G1 G2 . The commutativity
of Diagram (8.4.5) comes from that of the diagram

H4n (X 4 )
r / H 4n (X)

O LLL
LLL
LL
% 
4n 4 r / H 4n
HG G (X ) G2 G1 (X)
2
O 1
NNN O
NNrr
NNN =
( )
N' 
G 4n 2 r / H 4n
I G 1 HG G (X ) G2 G1 (X)
O O
2 II 2 1

G
G 1
2
G1
H n (X) VV / H 2n (X 2 ) r / H 2n (X) III
VV VVVV G1 G1
O
VVVV Sq
VVV V
VVVV
VV+ 
Sq
H 2n (BG1 X) / H 4n (BG BG X)
2 1

(8.4.6)

together with the obvious commutativity of the diagram

344 8 Steenrod Squares

4n (X 4 ) r / H 4n (X) / H 4n (B X)
H 
O KKK
KKj K

2
KK j (Bjid) .
%  
2
H (X)
n / H4n (X 4 )
r / H 4n (X) / H 4n (B X)


The commutativity of Diagram (8.4.6) comes from that of its subdiagram I, II and
III (the commutativity of the other diagrams is obvious). Diagrams II and III com-
mute because of Points (2) and (3) of Lemma 8.4.2. To verify the commutativity of
Diagram (I), we put it as the inner square of the following diagram.

/ H 4n
H 4n 4
 (X )
4
G2 G1 (X )
O KK O
KK oo
KK
KK
(D)
ooooo
% ow o
/ H 4n
G2 G1 (X )
4
H4n (X 4 )
O O
(A)
G 1
G (C) G
( )
 ( ) G 1
2 2
G1
H n (X) / HG2n1 (X 2 )
9 gOOO
sss
ss (B) OOO
ss OOO
ss G1
H n (X) / H 2n (X 2 )
G1

Diagrams (A), (B) and (C) commute because of Lemmas 8.3.1 and 8.4.2, and Dia-
gram (A) is obviously commutative. As H n (X) H n (X) is an isomorphism, the
inner inner square will commute if the outer does. But the outer square commutes by
default, since, using Proposition 7.2.17, all the groups are equal to Z2 and the maps
are non trivial (for G1 and GG21 , this is checked using the restrictions to the trivial
group and Point(2) of Lemma 8.4.2).
From Proposition 8.4.1, we deduce relations between Sqi Sqj called the Adem
relations.

Theorem 8.4.3 (Adem relations) Let X be a topological space. For each i, j N

with i < 2j, the relation

[i/2]
jk1

Sqi Sqj = Sqi+jk Sqk (0 < i < 2j)
i 2k
k=0

holds amongst the non-graded endomorphisms of H (X).

Example
  8.4.4 (1) When i = 1, the right hand member in the Adem relation reduces
to j1
1 Sq j+1 . Hence
8.4 Adem Relations 345

Sqj+1 if j is even
Sq Sq =
1 j
.
0 if j is odd

For instance, Sq1 Sq1 = 0 and Sq1 Sq2 = Sq3 .

jk1 
(2) In the limit case i = 2j 1 with j > 0, the binomial coefficient 2j12k vanish
if k j 1, since 2j 1 2k > j k 1. Then

Sq2j1 Sqj = 0 if j > 0 .

: H n (X)
Proof of the Adem Relations Let a H n (X). The homomorphism Sq

H (RP X)
H (X)[u] satisfies


n 

Sq(a) = Sq (a)un = Sq (a)un .
=0 Z

The range extension Z is possible since the other summands vanish. We shall
do that repeatedly in the computations below and, without other specification, the
summations will be over the integers (with only a finitely many non-zero terms).
This permits us to exchange the summation symbols.

Observe that Sq(u) one has
= u2 + uv = u(u + v) thus, by multiplicativity of Sq,
) = Sq(u)
Sq(u k k = u (u + v)k . Therefore
k


Sq(a)
Sq n )Sq(Sq
= Sq(u (a))

 
= un (u + v)n Sq Sq (a)vn+

= , un vn+ (u + v)n Sq Sq (a)
 n n+ 2n
= ,, u v Sq Sq (a)

Setting + = i yields
 n n+ 2ni i
Sq(a)
Sq = Sq Sq (a) .
,,i u v

Setting n + = 2n q yields
 2nq  2nq 2ni i q+n
Sq(a)
Sq = ,q,i u v Sq Sq (a) . (8.4.7)

By Proposition 8.4.1, for each i and q, the coefficient of u2nq v2ni in (8.4.7) must
be equal to that of u2ni v2nq . This leads to the equation
 2nq  i q+n    q i+n
Sq Sq (a) = 2ni
Sq Sq (a) .
346 8 Steenrod Squares

In the left hand member,  xj =

  set  q n. In the right hand member, set k = i + n
and use the relation xy = xy . In both side, restrict the range of summation so that
the summands are not zero for obvious reasons. This gives


i [i/2]
  nk 
nj  i j+
Sq Sq (a) = i2k Sq
j+ik
Sqk (a) . (8.4.8)
=0 k=0

Now, i and j being fixed, suppose that n = 2r 1+j for r large. Then, by Lemma 6.2.6,
nj  2r 1 
= = 0 if = 0

for, the dyadic expansion of 2r 1 has a zero where that of has a one. Hence,
the left hand member of (8.4.8) reduces to the single term Sqi Sqj (a). Also,
 nk  2r +jk1 jk1
i2k = i2k = i2k ifi < 2j

since the length of the dyadic expansions of i 2k is not more than that of j 1 and
adding 2r to j k 1 only puts a single 1 far to the left.
Thus, (8.4.8) proves the Adem relations classes of degree 2r 1 + j with r large
(2 > max{i, j}). As  Sq = Sq  , Eq. (8.4.8) holds for a if and only if it holds
r

for  (a). But the suspension isomorphism may be iterated on a class a H n (X)
till its degree becomes of the form 2r 1 + j with r large. This proves the Adem
relations.

8.5 The Steenrod Algebra

The Steenrod algebra A is the graded Z2 -algebra generated by indeterminates Sqi

(in degree i) and subject to the Adem relations and to Sq0 = 1. The properties of
Steenrod squares imply that the cohomology H (X) of a space X is an A-module.
The algebraic study of A-modules is a rich subject (see, e.g. [212] for a survey).

Lemma 8.5.1 As an algebra, A is generated by {Sqn | n = 2r }.

r 
Proof Let m = 2r +s with s < 2r (r 1). As 2 s1 1 mod 2, the Adem relation

[s/2]

r 2r k1
Sqs Sq2 = Sqm + s2k Sqmk Sqk (0 < i < 2j)
k=1

expresses Sqm as a sum of Sqi Sqj . If s > 0 then i, j < m, which permits us to prove
the lemma by induction on m. 
8.5 The Steenrod Algebra 347

Here are a few examples of decompositions of Sqi according to Lemma 8.5.1 and
its proof:

Sq3 = Sq1 Sq2

Sq5 = Sq1 Sq4
. (8.5.1)
Sq6 = Sq2 Sq4 + Sq5 Sq1 = Sq2 Sq4 + Sq1 Sq4 Sq1
Sq7 = Sq Sq = Sq Sq Sq
1 6 1 2 4

For 0 = a H 1 (RP ), the formula Sq(a2 ) = a2 + a2 shows that Sq2 is

n n n+1 n

not a sum of Sqi Sqj with i, j < 2n . On the other hand, the Adem relations imply that

Sq2 Sq2 = Sq3 Sq1 = Sq1 Sq2 Sq1 .

r
Therefore, the Sq2 do not generate A freely. In order to achieve that, we shall take
another system of generators.
For a sequence I = (i1 , . . . , ik ) of positive integers, we set SqI = Sqi1 Sqik
A. The degree of I is i1 + + ik . The sequence I is called admissible if ij 2ij+1 .
Let Admn be the (finite) set of admissible sequences of degree n. A monomial SqI
is called admissible if I is admissible.

Proposition 8.5.2 A is the polynomial algebra over the admissible monomials.

The family of admissible monomials is sometimes called the Cartan-Serre basis

of A.
Before proving Proposition 8.5.2, we develop some preliminaries. Fix an integer n
and consider

wn = x1 xn H ((RP )n ) Z2 [x1 , . . . , xn ] . (8.5.2)

As Sq(xi ) = xi + xi2 = xi (1 + xi ), one has


n 
n
Sq(wn ) = Sq(xi ) = wn (1 + xi ) .
i=1 i=1

Hence,

Sqk (wn ) = wn k (8.5.3)

where k is the k-th elementary symmetric polynomial:

1 = x1 + + xn
2 = x
1 x2 + xn1 xn (8.5.4)
k = i1 <<ik xi1 xik .
348 8 Steenrod Squares

For an integer p, we use the joker notation Lp for any polynomial in 1 , . . . , p . For
instance, the equations Lp = Lp+q and Sqi (Lp ) = Lp+i hold true but, as Sqp (p ) = p2
and Sqi (p ) = 0 for i > p, one has Sqi (Lp ) = L2p1 for all i.

Lemma 8.5.3 Let I = (i1 , . . . , ir ) be an admissible sequence. Then, for w = wn

with n i1 , one has

SqI (w) = w(i1 Li1 1 + Li1 1 ) = wLi1 .

Proof Only the first equation has to be proven. We proceed by induction on r. For
r=1, the lemma follows from (8.5.3). Suppose that k 2.
Let I = (i2 , . . . , ir ). By induction hypothesis, Eq. (8.5.3) and the Cartan formula,
one has

SqI (w) = Sqi1 SqI (w)
= Sqi1 (wLi2 )

= Sqi1 (w)Li2 + j1 Sqi1 j (w)Sqj (Li2 )

= wi1 Li2 + w j1 i1 j L2i2 1
= w(i1 Li1 1 + Li1 1 )

since i1 2i2 . 

Proof of Proposition 8.5.2 The Adem relations imply that any monomial SqI is a
sum of admissible ones. It remains to see that the admissible monomials are linearly
independent. 
Suppose that II I SqI = 0 for I Admn and I Z2 . We must prove that
I = 0 for all I. We proceed by induction on the cardinality of I. Equation (8.5.3)
for n i1 proves the assertion for I = {(i1 )}.
Let maxj I = max{0, ij | (i1 , . . . , ik ) I} and let w = wn as in (8.5.2) with
n max I. Set I = I0 I1 where I0 = {(i1 , . . . , ik ) I | i1 = max1 I}.
By Lemma 8.5.3, one has
  
0= I SqI (w) = I SqI (w) + I SqI (w) = w( max I Lmax I 1 + Lmax I 1 ) .
II II0 II1

Therefore,
 
I SqI (w) = 0 and I SqI (w) = 0 . (8.5.5)
II0 II1

(This uses a very easy case of the uniqueness of the expression of a symmetric
polynomial as a polynomial in the i s).
If the decomposition I = I0 I1 is non-trivial (i.e. I = I0 ), (8.5.5) permits
us to apply the induction hypothesis. Otherwise, we decompose I0 with respect to
8.5 The Steenrod Algebra 349

max2 (I0 ): I0 = I00 I 01 , where I0 = {(i1 , . . . , ik ) I | i2 = max2 I0 } and, as

in (8.5.5), obtain that
 
I SqI (w) = 0 and I SqI (w) = 0 . (8.5.6)
II00 II01

If  I > 1, iterating this process will once produce a non-trivial decomposition,

enabling us to use the induction hypothesis.
The proof of Proposition 8.5.2 shows that the map A  A(w) sends {SqI | I
Admn } into a free family of H ((RP )n ). This proves the following result.

Proposition 8.5.4 Let 0 = a H 1 (RP ). The evaluation map A H

((RP )n ) given by A  A(a a) (n times) is injective in degree n.

We no turn our interest to the cohomology ring H (Km ) of the Eilenberg-MacLane

complex Km . It contains the classes SqI () (0 = H m (Km ) = Z2 ) and the
admissible monomials play an important role. Define the excess e(I) of an admissible
sequence I = {i1 , . . . , ik } by

e(I) = (i1 2i2 ) + (i2 2i3 ) + + (ik1 2ik ) + ik = i1 i2 ik .

The excess of an admissible monomial SqI is the excess of I. Here is a famous

theorem of Serre [175, Sect. 2].

Theorem 8.5.5 H (Km ) is the polynomial algebra generated by SqI () for I admis-
sible of excess < m.

The proof of this theorem uses spectral sequences and will not be given here. The
condition e(I) < m is natural: SqI () = 0 is e(I) > n since i1 = e(I)+i2 + +ik >
n + i2 + + ik . If e(I) = n, then
r1
SqI () = (Sqi2 Sqik )2 = = (Sqir Sqik )2

where e(ir , . . . , ik ) < n.

Example 8.5.6 (1) Only the empty sequence has excess 0. Then H (K1 ) is the poly-
nomial algebra generated by H 1 (K1 ). This is not a surprise since K1 RP by
Proposition 3.8.3.
In order to formulate the other examples, observe that if I = (i1 , . . . , ik ) is admissible,
so is I + = (2i1 , i1 , . . . , ik ) and e(I + ) = e(I). We denote by F(I) the family of
admissible sequences obtained from I by iterating this construction.
(2) The family of admissible monomials with excess 1 is F(1). Thus H (K2 ) is a
polynomial algebra with one generator degree 2i + 1, i N. Its Poincar series is
 1
Pt (K2 ) = .
1 t 2 +1
i
iN
350 8 Steenrod Squares

For the Poincar series of Km , see Lemma 8.5.13.

(3) The set of admissible monomials with excess 2 is the union of the families
F(2r + 1, 2r , . . . , 2, 1) for r 0.
(4) The Poincar series of Km is computed in [175, Sect. 17].

Remark 8.5.7 The coefficient exact sequence 0 Z2 Z4 Z2 0 gives rise

to a Bockstein homomorphism : H (X) H +1 (X) (see [82, Sect. 3.E]). As is
functorial and not trivial, one has () = 0 in H n+1 (Kn ). But, by Theorem 8.5.5,
the only non-trivial element in H n+1 (Kn ) is Sq1 (). By naturality of and Sq1 , this
proves that = Sq1 . This argument illustrates the following corollary of Theorem
8.5.5, saying that the actions of the Steenrod algebra on H n () for all n N generate
all the mod 2 cohomology operations.

Corollary 8.5.8 Let Q be a cohomology operation, and let Q[n] its restriction to
H n (). Then, there exists An A such that Q[n] (x) = An x for all x H n (X) and
all spaces X.

Proof By functoriality, it suffices to prove the statement for X = K = K(Z2 , n)

and x = , the generator of H n (K). But Q[n] () H n (K) by Lemma 8.1.5 and
H i (K) = A by Theorem 8.5.5. 

We now list other corollaries of Theorem 8.5.5. The following one comes from
Proposition 8.5.4.

Corollary 8.5.9 Let 0 = a H 1 (RP ) and let y = a a H n ((RP )n ).

Let fy : (RP )n Kn such that Hfw () = y. Then, H f : H i (Kn ) H i
((RP )n ) is injective for i 2n.

Remark 8.5.10 The proofs of both Theorem 8.5.5 and Proposition 8.5.4, and then
that of Corollary 8.5.9, do not use the Adem relations. Thus, one can use Corollary
8.5.9 to give an alternative proof of the Adem relations, as in [175, Sect. 33], [154,
pp. 2931] or [27].

Corollary 8.5.11 The Adem relations are the only relations amongst the SqI s which
hold true for all spaces.

Proof A relation amongst the SqI s would be of the form P(SqI1 , . . . , SqIr ) = 0,
where P is a Z2 -polynomial in r variables. The Adem relations imply that any mono-
mial SqI is a sum of admissible ones. Therefore, there is a relation of the form

P(Sq J1 , . . . , SqJs ) = 0, where P is a Z -polynomial in s variables and J , . . . , J

2 1 s
are admissible sequences. Let m be the maximal excess of J1 , . . . , Js . For the
generator of H m (K(Z2 , m)), the equation P(Sq J1 (), . . . , SqJs ()) = 0 implies, by

Theorem 8.5.5, that P = 0. Hence, the original relation P(SqI1 , . . . , SqIr ) = 0 was
a consequence of the Adem relations. 

Another consequence of Theorem 8.5.5 is that Steenrod squares are characterized

by some of their properties listed in Theorem 8.2.1.
8.5 The Steenrod Algebra 351

Proposition 8.5.12 Suppose that for each CW -complex X, there exists a map
P : H (X) H (X) satisfying the following properties.
(a) If g : Y X is a continuous map, then H g P = P H g.
(b) P(H n (X)) H 2n (X).
(c) If a H 1 (RP ) then P(a) = a + a  a.
(d) P(x  y) = P(x)  P(y) for all x, y H(X).
Then P = Sq.
Proof Using (a) and (d) together with the definition of the cross product, we get
(d) P(x y) = P(x) P(y) for all x H(X) and y H(Y ).
Let w = x1 . . . xn H ((RP )n ) Z2 [x1 , . . . , xn ]. Using (c) and (d) we prove,
as for (8.5.3), that P(w) = w k . But, by Formula (8.5.3) again, this shows that
P(w) = Sq(w). Using (a), (b) and Corollary 8.5.9, we deduce that P = Sq on
H n (Kn ). By (a), this proves that P = Sq in general. 
As a last application of Theorem 8.5.5, we compute the Poincar series of Km ,
following [175, Sect. 17]. By Theorem 8.5.5, one has

 1
Pt (Km ) = ,
1 t m+a(r)
r=0

where

a(r) =  {I | Iadmissible, e(I) < m and deg(I) = r} .

To compute a(r), we note that an admissible sequence I = (i1 , . . . , ik ) is determined

by its excess components 1 = i1 2i2 , . . . , k1 = ik1 2ik , k = ik . Therefore,


k 
k
a(r) =  {(1 , . . . , k ) | i < m and i (2i 1) = r} . (8.5.7)
i=1 i=1

k
Set 0 = m 1 i=1 i . Then


k
m+r =1+ i 2i = 1 + 20 + 
+ 20 + + 2k + 
+ 2k . (8.5.8)
i=0 0 k


Using that ki=0 i = m 1 and writing the power of 2 in (8.5.8) in decreasing
order h1 hm1 , we get
m + a(r) =  {(h1 , . . . , hm1 ) Nr1 ) | h1 hm1 and2h1 + + 2hr + 1 = m + r} .
352 8 Steenrod Squares

This proves the following result of [175, Sect. 17].

Lemma 8.5.13 The Poincar series of Km is
 1
Pt (Km ) = .
2h1 ++2hm1 +1
h1 hm1 0 1 t

Let r < m. If I is an admissible sequence with deg(I) = r, the condition e(I) < m
is automatic since e(I) deg(I). Using (8.5.7), we see that a(r) is equal to the number
of partitions of r into integers of the form 2i 1. Also, H m+r (Km ) only contains
classes of the form SqI () (products like SqI ()SqJ () have higher degree). This
proves the following result of [191, p. 37].

Lemma 8.5.14 If r < m, then dim H m+r (Km ) is equal to the number of partitions
of r into integers of the form 2i 1.

8.6 Applications

Suspension of the Hopf maps. Recall that the non triviality of the cup-square map
(a) = a  a is H (KP2 ) for K = R, C, H or O implies that the Hopf maps

h1,1 : S 1 S 1 , h3,2 : S 3 S 2 , h7,4 : S 7 S 4 and h15,8 : S 15 S 8

are not homotopic to a constant maps (see Corollary 6.1.9). This argument cannot
be applied to the suspensions of the Hopf maps  k hp,q : S p+k S q+k since the cup
product in H >0 ( k KP2 ) vanish by dimensional reasons (also by Corollary 4.4.4).
But, for instance in RP2 , (a) = Sq1 (a). As  Sq = Sq  , one deduces that
Sq1 is not trivial on  k RP2 and therefore  k h1,1 is not homotopic to a constant map
for all k N [though H k+1  k h1,1 vanishes on H k+1 ( k RP2 )]. The same argument
applies for the other Hopf maps, so we get the following proposition.

Proposition 8.6.1 For all k 0, the k-th suspension of the Hopf maps

 k h1,1 : S k+1 S k+1 ,  k h3,2 : S k+3 S k+2 ,

 k h7,4 : S k+7 S k+4 and  k h15,8 : S k+15 S k+8

are not homotopic to a constant maps.

Actually, for k 1,  k h3,2 represents the generator of k+3 (S k+2 ) Z2 (see

[197, Proposition 5.1]).
Restrictions on cup-squares. The action of the Steenrod algebra on the coho-
mology imposes strong restrictions for the existence of classes with non-vanishing
cup-square. Let A<n be the subalgebra of A generated by {Sqi | i < n}.
8.6 Applications 353

Proposition 8.6.2 Let X be a topological space and let a H m (X). If m is not a

power of 2, then a  a A<m (a).
Proof By Lemma 8.5.1, A is generated by {Sqk | k = 2i }, so Sqm A<m if m is
not a power of 2. As a  a = Sqm (a), this proves the proposition. 
Corollary 8.6.3 Let X be a topological space. Let a H m (X) such that a  a = 0.
Then, there exists k m with k = 2i such that Sqk (a) = 0.
As a consequence of Proposition 8.6.2, if a H m (X) satisfies a  a = 0 with
m not a power of 2, there must be a non-zero group H m+k (X) with 0 < k < m. This
is not the case for the space X = Cf = D2m f S m used in Sect. 6.3 to define the
Hopf invariant of f : S 2m1 S m . Therefore, Proposition 8.6.2 has the following
corollary, a result due to Adem [4].
Corollary 8.6.4 Suppose that f : S 2m1 S m satisfies Hopf (f ) = 1. Then m = 2r .
When m is a power of 2, Proposition 8.6.2 is wrong, as seen with the projective
spaces KP2 . Using secondary cohomology operations, Adams proved deeper results
[2, Theorem 4.6.1 and Sect. 1.2] implying the following theorem.
Theorem 8.6.5 Let X be a space such that H m+k (X) = 0 for 1 k 2r . If r 3,
r+1
then Sq2 (H m (X)) = 0.
Combining this theorem with Corollary 8.6.4, Adams got his famous result [2,
Theorem 1.1.1]:
Theorem 8.6.6 Continuous maps f : S 2m1 S m with Hopf invariant 1 exist only
for m = 1, 2, 4, 8.
Remark 8.6.7 Recall that a non-singular map : Rm Rm Rm (see Sect. 6.2.2)
determines, using (6.2.2), a continuous multiplication : S m1 S m1 S m1 ,
giving rise to a map f : S 2m1 S m with Hopf (f ) = 1 (Proposition 6.3.3). Thus,
by Theorem 8.6.6, non-singular maps Rm Rm Rm exit only for m = 1, 2, 4, 8.
Note that Corollary 8.6.4 gives another proof of Proposition 6.2.5.
Somehow related to Theorem 8.6.6 are the results of E. Thomas on the cohomol-
ogy of H-spaces (see e.g. [194]). They also heavily use Steenrod square techniques.
A variant of the Sullivan conjecture. If f : R S is a continuous map, then
H f is a morphism of algebras over the Steenrod algebra A. The following result,
conjectured by H. Miller, was proven by Lannes [124, Theorem 0.4].
Theorem 8.6.8 Let Y be a simply connected space of finite cohomology type. Let
B = (RP )n . Then the map

[B, Y ] homA (H (Y ), H (B))

is a bijection.
In particular, if Y is a finite dimensional CW -complex, then [B, Y ] is a singleton.
This is a weak version of the Sullivan conjecture (see Remark 6.1.4).
354 8 Steenrod Squares

8.7 Exercises for Chapter 8

8.1. Let T = S 1 S 1 and let K be the Klein bottle. Is T homotopy equivalent to

K?
8.2. In H (RP RP ) Z2 [a, b], compute Sq6 (a5 b7 ).
8.3. Let a H m (X) such that Sq(a) = a. Prove that the smallest integer i such
that Sqi (a) = 0 is a power of 2.
8.4. Verify the relations in the Steenrod algebra given in (8.5.1).
8.5. Check that Sq2 Sq2 Sq2 Sq2 = 0 and that Sq3 Sq3 Sq3 = 0
8.6. Find a minimal set of generators of H (RP ) as a module over the Steenrod
algebra.
8.7. Let X be a space such that there is an GrA-epimorphism Z2 [a] H (X),
with a of degree r. Prove that r = 2 with k 3.
k
p
8.8. Let S r1 S n1 M be a locally trivial fiber bundle, where M is a closed
manifold of positive dimension. Prove that n = kr and r = 1, 2, 4, 8.
8.9. Let M be a closed connected manifold with total Betti number 3. What can be
said about its dimension?
8.10. Let M be a smooth closed G-manifold (G of order 2). Let f : M R
be a G-invariant Morse-Bott function satisfying the hypotheses of Theorem
7.6.6. Suppose that f has only two critical values 1, with M1 = pt and
dim M1 > 0. Suppose that n = dim M is odd. Prove that the pair (M, M1 ) is
homeomorphic to (RPn , RPn1 ). [Hint: this uses Exercise 8.8.]
8.11. Let Sq be the cohomology operation defined as follows: for a H n (X),
we set Sqi (a) = Sqni (a) (these Sqi s were the first operations defined by
Steenrod: see [40, Chap. 6]).
(a) Prove that Image(Sq0 ) ker Sq1 .
(b) Prove that Image(Sq0 ) = ker Sq1 for X = RP .
(c) Is (b) true for X = CP ?
Chapter 9
Stiefel-Whitney Classes

In Sect. 4.7.6, we encountered the Euler class e( ) H n (B) of a vector bundle

of rank n over B, which sometimes distinguishes between non-isomorphic vector
bundles. This is an example of a characteristic class, meaning that it is natural
with respect to induced bundles: e( f ) = H f (e( )) for any continuous map
f : B  B. Characteristic classes play a major role in the study of vector bundles.
In Sect. 9.4, we shall see that the Euler class belongs to a family wi ( ) H i (B) of
characteristic classes called the Stiefel-Whitney classes. Historically, these were the
first characteristic classes discovered, in the simultaneous work of Eduard Stiefel and
Hassler Whitney starting around 1935 (see [40, pp. 421426]). Using Morse theory,
we shall see in Sect. 9.5 that the cohomology of the Grassmannians is GrA-generated
by the Stiefel-Whitney classes of their tautological bundles. This implies that any
characteristic class in mod 2 cohomology is a polynomial in the Stiefel-Whitney
classes (see Corollary 9.5.10). The Chern classes of a complex vector bundle, when
reduced mod 2, are also Stiefel-Whitney classes (see Remark 9.7.6).
Geometric interpretations of w1 and w2 are given in Sects. 9.2 and 9.3, related to
orientations and spin structures. Stiefel-Whitney classes have important applications
in the topology of manifolds, for instance via the Wu formula (see Sect. 9.8) or in
the work by R. Thom on cobordism, surveyed in Sect. 9.9.2.

9.1 Trivializations and Structures on Vector Bundles

We recall below some classical facts about vector bundles (defined on p. 186).
Two vector bundles = ( p : E X ) and  = ( p  : E  X ) over the
same space X are isomorphic if there exists a homeomorphism h : E E  such
that p  h = p and such that the restriction of h to each fiber is linear. Two such
isomorphisms h 0 , h 1 : E E  are isotopic if there exists a family of isomorphisms
h t : E E  depending continuously on t [0, 1] and joining h 0 to h 1 . Unless
otherwise mentioned, the total space of a vector bundle is denoted by E( ) and the
bundle projection by p.
Springer International Publishing Switzerland 2014 355
J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_9
356 9 Stiefel-Whitney Classes

Let be a vector bundle of rank r over Y and let f : X Y be a continuous map.

The induced vector bundle f is the vector bundle of rank r over X defined by

E( f ) = {(x, z) X E( ) | f (x) = p(z)},

with the projection onto X as the bundle projection. The projection to E( ) gives a
map f : E( f ) E( ) which is a linear isomorphism on each fiber and such that
the diagram

f
E( f ) / E( )
pr X p (9.1.1)
 f

X / Y

is commutative. Note that if f and g coincide over A X , the induced bundles

f and g restrict to the same bundle over A. Two such maps are said homotopic
relative to A if there is a homotopy F : X I X between f and g such
that F(a, t) = f (a) = g(a) for all a A. The next proposition is proven in
[181, Theorem 11.3].

Proposition 9.1.1 Let be a vector bundle over a space B. Let f, g : X B be

two maps, where X is a paracompact space. Suppose that f and g coincide over
A X . If f and g are homotopic relative to A, then there exists an isomorphism
between f and g which is the identity over A.

A trivialization of a vector bundle over X is an isomorphism of with the product

bundle r = ( pr X : X Rr X ). A vector bundle admitting a trivialization is
called trivial. Denote by T ( ) the space of trivializations of (endowed with the
compact-open topology). Then, T ( ) = 0 (T ( )) is the set of isotopy classes of
trivializations of is 0 (T ( )).
By post-composition, the topological group Aut (r ) of the automorphisms of the
product bundle r acts continuously on the left on T ( ). As usual we denote by
G L(r, R) the general linear group, i.e. the topological group of automorphisms of
Rr . Alternatively, G L(r, R) is the group of invertible (r r )-matrices with real coef-
2
ficients, topologized as an open set of Rr . Note that Aut (r ) Map(X, G L(r, R)),
so the group 0 (Map(X, G L(r, R))) = [X, G L(r, R)] acts on T ( ).

Lemma 9.1.2 Let be a trivial vector bundle over a topological space X . Then the
action

[X, G L(r, R)] T ( ) T ( ) (9.1.2)

is simply transitive. In particular, given a trivialization T0 of , the correspondence

T0 induces a bijection [X, G L(r, R)] T ( ).
9.1 Trivializations and Structures on Vector Bundles 357

Proof Let h, h  T ( ). Then h  = (h  h 1 ) h and h  h 1 Aut (r ). This shows

that Aut (r ) acts transitively on T ( ), whence the action (9.1.2) is transitive.
On the other hand, if h 1 = h 0 T ( ) is isotopic to h 0 via a homotopy
h t T ( ), the formula t = h t h 1 0 defines a continuous path t Aut (r )
joining 0 = id to 1 = . This shows that [X, G L(r, R)] acts simply on T ( ). 

Remark 9.1.3 The bijection of Lemma 9.1.2 depends on the choice of [h 0 ] T ( ).

In general, there is no natural choice, except for the product bundle (like in the next
example).

Example 9.1.4 We see the projection S 1 C S 1 as the product vector bundle

of rank 2 over the circle, identifying C with R2 . By Lemma 9.1.2

T ( ) [S 1 , G L(2, R)] [S 1 , O(2)] {1} Z

(by Gram-Schmidt orthonormalization, G L(2, R) has the homotopy type of O(2)

which is homeomorphic to {1} S 1 ). The fiber linear map corresponding to
(1, k) is given by (x, z) x k z and that corresponding to (1, k) is given by
(x, z) x k z .

Example 9.1.5 Let be a vector bundle over Y and let f : X Y be a continuous

map. If is trivial, so is f . More precisely, let h : E( ) Y Rr be a trivialization
of . Write h under the form h(z) = ( p(z), h(z)) where h : E( ) Rd is a
map whose restriction to each fiber is a linear isomorphism. Then the map f h :
E( f ) X Rr given by f h(x, z) = (x, h(z)) is a trivialization of f . This
process descends to a map

f : T ( ) T ( f )

which is equivariant for the actions defined in (9.1.2), via the homomorphism f :
[Y, G L(r, R)] [X, G L(r, R)] induced by f . Indeed, for : Y
G L(r, R) and T T ( ), the following formula holds true in T ( f ).

f ( T ) = ( f ) f T. (9.1.3)

Proposition 9.1.6 Let be a vector bundle over a space X . If X is paracompact

and contractible, then is trivial. Moreover, any trivialization of the restriction of
over a point x0 X extends to a trivialization of which is unique up to isotopy.

Proof Let = ( p : E X ) and let 0 = ( p : E 0 {x0 }) be the restriction of

over the point x0 . The vector bundle is induced by the identity of X : = idX . As X
is contractible, id X is homotopic to a constant map c onto x0 and, by Proposition 9.1.1,
c 0 . The contractability of X also implies that c : [{x0 }, G L(r, R)]
[X, G L(r, R)] is an isomorphism. Proposition 9.1.6 then follows from Lemma 9.1.2
and the considerations of Example 9.1.5. 
358 9 Stiefel-Whitney Classes

We now consider a vector bundle over a space X = Z Y obtained by

attaching a pair (Z , A) to a space Y using an attaching map : A Y (see p. 97).
Let jY : Y X and j Z : Z X be the natural maps. Let Y = jY , Z = j Z
and denote by A the restriction of Z over A.
Lemma 9.1.7 Let be a vector bundle over X = Z Y as above. Suppose that
there exists trivializations h Y T (Y ) and h Z T ( Z ) such that the restriction
of h Z to A is isotopic to h Y . Suppose that Z is paracompact and that the pair
(Z , A) is cofibrant. Then, the bundle is trivial.
Proof Let h A T ( A ) be the restriction of h Z to A . An isotopy from h A to h Y
produces a trivialization h T ( A I ) (where A I is the vector bundle p id :
E( A ) I A I ) restricting to h A on A {0} and to h Y on A {1}. The pair
(Z , A) being cofibrant, there exists a retraction r : Z I Z {0} A I of Z I
onto Z {0} A I (see Lemma 3.1.37). Let 1 be the restriction of r Z above
Z {1} Z . As Z is paracompact and r is actually a strong deformation retraction
(see Remark 3.1.42), 1 is isomorphic to Z relative to A by Proposition 9.1.1. Then,
r h produces at level 1 a trivialization h Z of Z which is equal to h Y on E( A ).
This condition implies that the formula

h Z (z, u) if p(u) = i Z (z)
H (u) =
h Y (y, u) if p(u) = i Y (y)

defines a trivialization H : E( ) X Rr of . 
The above lemmas have analogues for principal bundles. Let A be a topological
group. Recall that a A-principal bundle over X consists of a continuous map p : P
X , a continuous right action of A on P such that p(z) = p(z) for all z P and all
A. In addition, the following local triviality should hold: for each x X there
is a neighbourhood U of x and a homeomorphism h : U A p 1 (U ) such that
p h(x, a) = x and h(x, a) = h(x, a). In consequence, A acts simply transitively
on each fiber and p is a surjective open map, thus descending to a homeomorphism

P/A X (use [44, Sect. I of Chap. VI]). An isomorphism of A-principal bundles
from p : P X to p  : P  X is a A-equivariant homeomorphism h : P P 
such that p  h = p. Two such isomorphisms h 0 , h 2 : P P  are isotopic if there
exists a family of isomorphisms h t : P P  depending continuously on t [0, 1]
and joining h 0 to h 1 . The notion of induced A-principal bundle works as for vector
bundles and Proposition 9.1.1 also follows from [181, Theorem 11.3].
A trivialization of an A-principal bundle is an isomorphism with the product
bundle X A X . An A-principal bundle is trivial if and only if it admits a
section (see, e.g. [181, I.8]). This gives a bijection between sections and trivializations
and between homotopy classes of sections and isotopy classes of trivialization. If
1 , 2 : X P are two sections of p, then 2 (x) = 1 (x)(x) for a unique
: Map(X, A) whence the analogue of Lemma 9.1.2: the action of Map(X, A) on
the trivializations of P is simply transitive. Also, Proposition 9.1.6 holds true for
principal bundles.
9.1 Trivializations and Structures on Vector Bundles 359

Let p : P X be an A-principal bundle, p  : P  X be an A -principal bundle

and let : A A be a continuous homomorphism. A continuous map g : P P 
is called a -map if p  g = p and g(z) = g(z) () for all z P and A.
Let : A G L(r, R) be a representation (i.e. a continuous homomorphism)
of the topological group A. Let P X be an A-principal bundle. Then the Rr -
associated bundle with total space

P (A,) Rr = P Rr {(z, u) (z, ()u)}

is a vector bundle of rank r over X . Most of the time, the representation is implicit
and we just write P A Rr for P (A,) Rr . A (A, )-structure (or just A-structure)
for a vector bundle of rank r over X is an A-principal bundle P X together
with a vector bundle isomorphism f : P A Rr E( ). Two A-structures (P, f )
and (P  , f  ) are
strongly equivalent if there exists an isomorphism of A-principal bundles h : P
P  such that f  (h idRr ) = f .
weakly equivalent if there exists an isomorphism of A-principal bundles h : P
P  such that f  (h idRr ) is isotopic to f .
Here are a few examples.

9.1.8 A = {1}, the trivial group. An {1}-structure on a vector bundle is just a

trivialization of . Strong equivalence coincides here with equality. Weak equivalence
classes of {1}-structures correspond to isotopy classes of trivializations.

9.1.9 Each vector bundle of rank r admits a G L(r, R)-structure which is unique
up to strong equivalence. Indeed, consider the space of frames of :

Fra( ) = { : Rr E( ) | is a linear isomorphism onto some fiber of }.

with the map pFra : Fra( ) X given by pFra () = p (0) (The image by
Fra( ) of the standard basis of Rr is a frame (basis) of (Rr ), whence the name space
of frames). By precomposition, the topological group G L(r, R) acts continuously and
freely on the right upon Fra( ). The map pFra descends to a continuous bijection
Fra( )/G L(r, R) X . Using local trivializations, one checks that this map is a
homeomorphism. Also, a local trivialization of gives rise to a local section of pFra .
Hence, pFra is a G L(r, R)-principal bundle, called the framed bundle of .
Consider the evaluation map f Fra : Fra( ) G L(r,R) Rr E( ) sending [, t] to
(t). This map is a continuous bijection which is linear on each fiber and, using
local trivializations of again, one checks that it is a homeomorphism. Hence,
f Fra is a G L(r, R)-structure on . We claim that f Fra is a universal structure in
the following sense. Each A-structure (for a representation : A G L(r, R))
360 9 Stiefel-Whitney Classes

f : P Rr E( ) determines a -map f : P Fra( ). The map f sends z P

to the map f (z, ) Fra( ). We check that two A-structure (P, f ) and (P  , f  ) are
(a) strongly equivalent if there exists an isomorphism of A-principal bundles h :
P P  such that f h = f.
(b) weakly equivalent if there exists an isomorphism of A-principal bundles h :
P P  such that f h is homotopic to f.
The case A = G L(r, R) give the uniqueness claimed above: any G L(r, R)-structure
on (for = id) is strongly equivalent to f Fra .

9.1.10 G L + (r, R)-structures and orientations (see also Sect. 9.2). Recall that an
orientation of a finite dimensional real vector space is an equivalence class of
(ordered) basis, where two basis are equivalent if their change-of-basis matrix is
in G L + (r, R), i.e. has positive determinant. An orientation of a vector bundle is an
orientation of each fiber which varies continuously, i.e. there are local trivializations

p 1 (U ) U Rr whose restriction to each fiber is orientation-preserving (for
the standard orientation of Rr ). A vector bundle admitting an orientation is called
orientable and the choice of an orientation makes it oriented.
If P is a G L + (r, R)-principal bundle, then P G L + (r,R) Rr is oriented, using the
standard orientation of Rr . Hence, a G L + (r, R)-structure (P, f ) on a vector bundle
makes it oriented. Conversely, an oriented vector bundle admits a G L + (r, R)-
structure: one just restricts the canonical (Fra( ), f Fra ) of 9.1.9 to Fra+ ( ), where

Fra+ ( ) = { Fra( ) | preserves the orientation}. (9.1.4)

(for the standard orientation of Rr ). As in 9.1.9, we check that (Fra+ ( ), f Fra ) is

universal for the A-structures on with a representation : A G L + (r, R).

9.1.11 O(r )-structures. Consider the orthogonal group O(r ) with its standard rep-

resentation O(r )  G L(r, R). Let f : P O(r ) Rr E( ) be a O(r )-structure on
. Then, the standard inner product on R gives, via f , an Euclidean structure on
r

(see p. 187) for which f is an isometry on each fiber. On the other hand, an Euclidean
bundle of rank r admits an O(r )-structure: one restricts f Fra to the subbundle of
Fra( ) formed by orthonormal frames:

Fra ( ) = { : Rr E( ) | is a linear isometry onto some fiber of }.

As in (9.1.9), one shows that such an O(r )-structure compatible with a given
Euclidean structure on is unique up to strong equivalence. This process provides
a bijection between Euclidean structures on and strong equivalences of O(r )-
structures.
9.1 Trivializations and Structures on Vector Bundles 361

On the other hand, a vector bundle over a paracompact space admits Euclidean
structures which form a convex space. Let (, et ) be the vector bundle endowed
with an Euclidean structure et depending continuously on t I . Then

Frae () = {t :Rr E( ) | t I, t is a linear isometry onto some fiber of (, et )}

is the total space of an O(r )-principal bundle over X I . Note that Frae ( ) is
the union indexed by I of Fra (, et ), the bundle of orthonormal frames for the
Euclidean structure et . If X is paracompact, Frae ( ) X I is isomorphic to the
principal bundle Fra (, e0 ) I X I [105, Chap. 4, Theorem 9.8]. Hence, there
is an isomorphism h : Fra (, e0 ) Fra (, e1 ) such that i 1 h is homotopic to i 0
(where i t : Fra (, et ) Fra( ) denotes the inclusion). This proves the following
statement: a vector bundle over a paracompact space admits an O(r )-structure which
is unique up to weak equivalence.

9.1.12 The case of orthogonal representations. Let : A O(r ) be an orthogonal

representation of a topological group A. Let f : P A Rr E( ) be an A-structure
on the vector bundle (for the representation ). As is orthogonal, P A Rr
inherits a natural Euclidean structure which is transported on via f (see 9.1.11).
As in 9.1.9, there is an O(r )-structure (PO , f O ) on such that f = f O f for a
-map f : P PO . If h : P P  induces a strong equivalence between the
A-structures (P, f ) and (P  , f  ), then PO = PO and f h = f. If h only induces
a weak equivalence, it descends to an isomorphism h O : PO PO making the
following diagram commutative

P
h / P
f f .
 
hO
PO / P
O

We can pull-back f over PO via h O , getting another representative of the weak

equivalence class of (P  , f  ). This permits us to assume that PO = PO and h O = id;
this also means that the Euclidean structure induced by f and f  coincide. Now, using
the Gram-Schmidt orthonormalization process in each fiber of , the isotopy between
f  (h idRr ) and f may be deformed into an isotopy of isometries. This implies
that f and f h are homotopic.
These considerations drive us to the following point of view for an A-structure
on , in the case of an orthogonal representation. One first fix an Euclidean structure
on and consider only A-structures (P, f ) on for which f is an isometry. In other
words, an A-structure may be seen as a -map f : P Fra ( ) and strong or weak
equivalences are described as in (a) and (b) of 9.1.9.
Note that, if is an oriented Euclidean bundle, one can consider the oriented
orthonormal frames
362 9 Stiefel-Whitney Classes

Fra+ ( ) = Fra ( ) Fra+ ( ) (9.1.5)

which is the total space of an S O(r )-principal bundle over X . Then (Fra+ ( ), f Fra )
is an S O(r )-structure which is universal for the A-structures associated to a repre-
sentation A S O(r ). The special case of the representation Spin(r ) S O(r )
(spin structures) is treated in Sect. 9.3 (compare [130, Sect. II.1]).
The best known example of r ep+ ( ) is for = T S n , the tangent bundle to the
standard unit sphere S n . One sees T S n as the space of couples (v, w) Rn+1 Rn+1
such that |v| = 1 and v, w = 0 (up to translation, w is tangent to v S n ). Then
the map q : S O(n + 1) S n defined by q(A) = Ae1 (first column vector) is the

oriented frame bundle for T S n . The bundle isomorphism S O(n + 1) Fra+ (T S n )
sends A to the map A : Rn T S n defined by

 
n

A (t1 , . . . , tn ) = Ae1 , ti Aei+1 .
i=1

9.1.13 Complex vector bundles. By replacing R by C in the definition of a vector

bundle, we get the notion of a complex vector bundle. The notion of a A-structure is
defined accordingly, using a complex representation : A G L(r, C). As in 9.1.9,
a complex vector bundle of rank r admits a G L(r, C)-structure which is unique
up to strong equivalence. Also, as in 9.1.11, using a Hermitian structure on (those
form a contractible space), admits an U (r ) structure which is unique up to weak
equivalence.

9.1.14 Classifying spaces and structures. Let E G L(r, R) BG L(r, R) be the

universal bundle for the principal G L(r, R)-bundles and let its associated vector
bundle. Recall that, for X paracompact, the correspondence [c] c provides a
bijection from [X, BG L(r, R)] and the set of isomorphism classes of vector bundles
over X of rank r (using (9.1.9)). We shall use the following consequence of this
result.

Lemma 9.1.15 Let (X, A) be a cofibrant pair with X paracompact. Let be a vector
bundle over X whose restriction over A is trivial. Then, there exists a vector bundle
over X/A such that , where : X X/A is the quotient map.

Proof Let c : X BG L(r, R) be a classifying map for (r = rank of ), i.e.

c ( ). In the definition of a cofibrant pair, A is supposed to be closed, so A is
paracompact. The restriction of c to A is then null-homotopic. As (X, A) is cofibrant,
there is a homotopy from c to c such that c|A is a constant map. Therefore, c descends
to c : X/A BG L(r, R). Hence, c (c ). 

Structures on vector bundles and the classifying spaces are related as follows. Let
be a vector bundle of rank r over a paracompact space X . Fix a characteristic map
c : X BG L(r, R) for Fra( ). A representation : A G L(r, R) of A induces
a continuous map B : B A BG L(r, R) which may be made a Serre fibration.
9.1 Trivializations and Structures on Vector Bundles 363

A lifting of c is a continuous map c : X B A such that B c = c and two such

liftings c0 and c1 are homotopic if there exists a homotopy h : X I B A between
c0 and c1 such that B h(x, t) = c(x) for all (x, t) X I . Then, the set of weak
equivalence of (A, )-structures on is in bijection with the homotopy classes of
liftings of c. For details (For details, see [23, Sect. 4]).

9.2 The Class w1 Orientability

Let O(V ) be the set of the two orientations of a finite dimensional vector space V .
Let be a vector bundle of rank r over a topological space X . Using the canonical
G L(r, R)-structure (Fra( ), f Fra ) of 9.1.9, we define

O( ) = Fra( ) G L(r,R) O(Rr ).

The projection O( ) X is a locally trivial bundle whose fiber over x X is, via
f Fra , in bijection with O( p 1 (x)). In consequence, O( ) X is a two fold covering.
An orientation of (see 9.1.10) is clearly a continuous section of O( ) X .
The characteristic class w(O( ) X ) H 1 (X ) (see Sect. 4.3.2), is called the
first Stiefel-Whitney class of and is denoted by w1 ( ).

Proposition 9.2.1 A vector bundle over a CW-complex X is orientable if and only

if w1 ( ) = 0 in H 1 (X ). The set of orientations of an orientable bundle is in bijection
with H 0 (X ).

Proof The 2-covering O( ) is trivial if and only if it admits a continuous section,

that is to say if and only if is orientable. On the other hand, if X is a CW-complex,
the 2-covering O( ) X is trivial if and only if its characteristic class vanish
(see Lemma 4.3.6). There are two orientation for the restriction of over each con-
nected component, whence the last assertion. 

Remark 9.2.2 If is a trivial vector bundle, then w1 ( ) = w1 ( ). Indeed,

there exists a natural map O( ) O( ) which is an isomorphism of 2-fold
coverings.

Proposition 9.2.3 Vector bundles over an 1-dimensional CW-complex are classified

by their rank and their first Stiefel-Whitney class.

Proof Let and  be two vector bundles over a CW-complex X . If and  are
isomorphic, they have the same rank and the 2-coverings O( ) and O(  ) are iso-
morphic, which implies that w1 ( ) = w1 (  ).
To prove the converse, note that, when X is 1-dimensional, any vector bundle
over X is the Whitney sum of a line bundle with the trivial vector bundle
[105, Chap. 8, Theorem 1.2]. By Remark 9.2.2, w1 ( ) = w1 (). We are thus reduced
to and  being both of rank 1, in which case we use Proposition 9.2.4. 
364 9 Stiefel-Whitney Classes

Let L(X ) be the set of isomorphism classes of real lines bundles over a space
X . The tensor product (see (7.5.5)) of two line bundles is again a line bundle. This
provides an operation on L(X ).

Proposition 9.2.4 Let X be a CW-complex. Then the first Stiefel-Whitney class pro-
vides an isomorphism

w1 : (L(X ), )
(H 1 (X ), +).

In particular, (L(X ), ) is an elementary abelian 2-group.

Proof Let = ( p : E B) be a line bundle over X . Endow with a Euclidean

structure. The unit sphere bundle S(E) X is then a 2-fold covering which is
clearly isomorphic to O( ). By (4.3.5), w(O( )) determines O( ) and then S(E).

But S(E) determines by the isomorphism S(E) O(1) R E (an O(1)-structure
on ). Hence, w1 is injective. For the surjectivity, let a H 1 (X ). By (4.3.5), there
is a 2-covering X X with characteristic class a. We see it as an O(1)-principal
bundle. Then, X O(1) R X is a Euclidean line bundle whose sphere bundle is
X . By the above, w1 ( ) = a.
It remains to show that w1 (  ) = w1 ( ) + w1 (  ) for two line bundles and 
over X . We start with some preliminaries. The linear group G L(R) is isomorphic
the multiplicative group R = R {0}. Set K = R R . Using the R-vector

space isomorphism R R R such that x y x y, the diagram

G L(R) G L(R) / G L(R R)



K / R

is commutative, where is the natural homomorphism (see p. 302) and is just

the standard product, which is a continuous homomorphism.
Let F = Fra( ), F  = Fra(  ) and F = Fra(  ) seen, using the above, as
principal R -bundles. Let : X X X be the diagonal inclusion. The definition
of F in (7.5.5) becomes

F = (F F  ) ( (F F  )).

Let ,  : X BR be characteristic maps for F and F  . One has morphisms

of R -principal bundles, i.e. commutative diagrams
9.2 The Class w1 Orientability 365

E(F ) / E(F F  ) K R / (ER ER ) K R

 (9.2.1)
  
/ XX / BR BR
X
(,  ) 3

where the maps on the top line are R -equivariant.

Recall from (7.4.3) that the two projections of K onto R induce a homotopy

equivalence P : B K BR BR . Let P  be a homotopy inverse of P. One has
more morphisms of R -principal bundles:

/ E K K BR / ER
(ER ER ) K R
(9.2.2)
  
P / BK B
/ BR
BR BR 

where is defined by ([(ti (gi , gi ), ]) = (ti , gi gi ). Diagrams (9.2.1) and (9.2.2)
imply that B P  (,  ) is a characteristic map for the R -principal bundle F .
As the inclusion {1}  R is a homomorphism and a homotopy equivalence,
one has BR  B{1}  RP  K (Z2 , 1). The map B P  thus defines contin-
uous multiplication on K (Z2 , 1). One checks that it is homotopy commutative and
admits a homotopy unit. Let u be the generator of H 1 (BR ) = Z2 . By Proposition
4.7.54 and Lemma 4.7.56, one has

H (B P  )(u) = u 1 + 1 u. (9.2.3)

Hence,

w1 (  ) = H (,  )(u 1 + 1 u) = H (u) + H  (u)

= w1 ( ) + w1 (  ). 

The above shows that the correspondence [ X X ] [ X O(1) R X ] induces

a bijection between the set Cov2 (X ) of equivalence classes of 2-fold coverings of X
and L(X ), with a commutative diagram

/ L(X )
Cov2 (X )
KKK u
KKK uu
w KK% uu .
zuu w1
H 1 (X )

By Proposition 4.7.36, we get the following corollary.

366 9 Stiefel-Whitney Classes

Corollary 9.2.5 Let be a line bundle over a CW-complex X . Then w1 () H 1 (X )

coincides with the Euler class e() of .

Example 9.2.6 The tautological line bundle over RP n . The 2-covering = (S n

RP n ) (1 n ) is an O(1)-principal bundle. Its associated line bundle =
(S n O(1) R RP n ) satisfies w1 () = w( ) = 0 in H 1 (RP n ) = Z2 (see the
proof of Proposition 4.3.10). Seeing RP n as the space of vector lines in Rn+1 ,
identifies itself with the tautological line bundle over RP n , i.e.

E() = {(a, v) RP n Rn+1 | v a} , p(a, v) = a.

The identification S n O(1) R
E() is given by [z, t] (R z, t z).

Proposition 9.2.7 Let be a vector bundle over a CW-complex X . Then, the fol-
lowing conditions are equivalent.
(1) The restriction of over the 1-skeleton of X is trivial.
(2) w1 ( ) = 0.

Proof Let i : X 1 X denote the inclusion and let 1 i be the restriction of

over X 1 . By Lemma 4.3.6, w1 (1 ) = i (w1 ( )) and w1 (1 ) = 0 if 1 is trivial.
As i : H 1 (X ) H 1 (X 1 ) is injective, this proves the implication (1) (2).
Conversely, if w1 ( ) = 0, then w1 (1 ) = 0 and 1 is trivial by Proposition 9.2.3. 

We finish this section by describing a singular cocycle representing w1 ( ) in terms

of transporting orientations. Let c : I X be a path and let O(E c(0) ) (where
E x is the fiber of over x X ). We see as an element of the fiber of Fra( ) over
c(0). The unique lifting c : I Fra( ) of c with c(0)
= provides an orientation
c = c(1)
O(E c(1) ). We say that c is obtained from by transport along c.
Choose a set-theoretic section : X Fra( ), i.e. an assignation of an orienta-
tion of E x for each x X (which has not to vary continuously). We see a singular
simplex S1 (X ) as a path : I X via the identification I 1 sending t to
(t, 1 t). The cochain w 1 (, ) defined by

1 if ( (0)) = ( (1))
w 1 (, ),  =
0 otherwise.

The cochains w 1 (, ) and w(Fra(

), ) of Sect. 4.3.2 clearly coincide. Hence, by
Proposition 4.3.7, w 1 (, ) is a cocycle representing w1 ( ).
Let x0 X and 0 O(E x0 ). We say that a loop c : I X at x0 preserves the
orientation if c 0 = 0 (this condition does not depend on the choice of 0 ). Note
that, if d is another loop at x0 , then (dc) 0 = d (c 0 ). Also, c 0 depends only on
the homotopy class of the loop c, by the homotopy lifting property of the covering
Fra( ) X . Hence the correspondence
9.2 The Class w1 Orientability 367


0 if c preserves the orientation
c
1 otherwise.

provides a homomorphism w : 1 (X, x0 ) Z2 which corresponds to w1 ( )

H 1 (X ) via the isomorphism H 1 (X )
hom(1 (X, x0 ), Z2 ) of Lemma 4.3.1.

9.3 The Class w 2 Spin Structures

In this section, we define a cellular second Stiefel-Whitney class w 2 ( ) H 2 (X ),

when is an orientable vector bundle over a regular CW-complex X (recall that H
denotes the cellular cohomology introduced in Sect. 3.5). A more general w2 ( )
H 2 (X ) (X any space) is considered in Sect. 9.4, but the results below are used to
establish the relationship between w2 ( ) and the existence of spin structures on .
Let be an orientable vector bundle of rank r 2 over a CW-complex X . Denote
by i the restriction of over the i-skeleton X i of X . By Propositions 9.2.1 and
9.2.7, 1 is trivial. Fix an orientation of . Choose a trivialization T1 of 1 which is
compatible with this orientation. The restriction T0 of T1 over X 0 is thus uniquely
determined up to isotopy.
Let e be a 2-cell of X with characteristic map : (D 2 , S 1 ) (X, X 1 ). As D 2 is
contractible, there is a unique (up to isotopy) trivialization Te of over D 2 which is
compatible with the orientation (see Proposition 9.1.6). Thus, T1 and Te provides
two trivialization of restricted to the boundary S 1 of D 2 . By Lemma 9.1.2, these
two trivializations differ by the action of a map from S 1 to G L(r, R), whose range is
G L + (r, R) since T1 and Te are both compatible with the chosen orientation. Let

w 2 (e) [S 1 , G L + (r, R)] 1 (G L + (r, R)) 1 (S O(r )) (9.3.1)

be the homotopy class of this map. Here, the isomorphism [S 1 , G L + (r, R)]
1 (G L + (r, R)) holds true since G L + (r, R) is a topological group and G L + (r, R)
has the homotopy type of S O(r ) by the Gram-Schmidt orthonormalization process.
If r 3, then 1 (S O(r )) = Z2 . If r = 2 then 1 (S O(r )) Z and, by conven-
tion, we take w 2 (e) mod 2. The correspondence e w 2 (e) thus defines a cellular
2-cochain w 2 = w 2 (, T1 ) C 2 (X ).

Lemma 9.3.1 Suppose that X has no 3-cells or that X 3 is regular complex. Then,
the cochain w 2 (, T1 ) is a cellular cocycle. Its cohomology class w 2 ( ) H 2 (X )
depends only of the isomorphism class of .

The class w 2 ( ) H 2 (X ) is called the second Stiefel-Whitney class of the vector

bundle . If the rank of is 1, we set by convention that w 2 ( ) = 0.

Proof We may assume that X is connected. Observe first that the cochain w 2 (, T1 )
does not depend on the orientation of , since the other choice would just change all
368 9 Stiefel-Whitney Classes

the orientations under consideration. Let us see how w 2 (, T1 ) depends on the isotopy
class of the trivialization [T1 ] T (1 ). Let T1 T (1 ) be another trivialization
compatible with the orientation. As seen above, this compatibility implies that the
restriction T0 of T1 to 0 coincides with T0 up to isotopy. Such an isotopy may be
realized in a neighbourhood of X 0 in X 1 , so one may assume that T0 = T0 . By
Lemma 9.1.2, there is a unique map a : X 1 G L + (r, R) such that T1 = a T1 .
As T0 = T0 , the restriction of a to X 0 is constant to the identity of Rr . Hence, each
1-cell with characteristic map : D 1 X 1 gives rise to a homotopy class

a() = [a ] [(D 1 , S 0 ), (G L + (r, R), id)] 1 (G L + (r, R), id) 1 (S O(r ), id) = Z2

which does not depend on the choice of (since 1 (S O(r ), id) = Z2 ; again,
if r = 2, one takes by convention the value mod 2 of a 1 (S O(2)) Z).
The correspondence a() determines a cellular 1-cochain a C 1 (X ). Let
e be a 2-cell of X with characteristic map : (D 2 , S 1 ) (X, X 1 ). Using the
cellular boundary formula (3.5.4) and writing the abelian group 1 (G L + (r, R)) =
Z2 additively, we get that [a ] = (a). In the same way, using (9.1.3), one gets the
following formula.

w 2 (, T1 ) = w 2 (, a T1 ) = w 2 (, T1 ) + (a). (9.3.2)

We now prove that w 2 is a cellular cocycle. If X has no 3-cells, there is nothing

to prove. Let  be a 3-cell of X which, as X 3 is now supposed to be regular, can
be identified with a subcomplex (also called ) of X . As  is contractible, there is
a unique (up to isotopy) trivialization T  of | , compatible with the orientation. As
above, one may assume that T  coincides with T0 over  0 . A trivialization T1 T (1 )
may be thus defined by

T  (z) if p(z) 
T1 (z) =
T1 (z) otherwise.

By (9.3.2) and the construction of T1 , one has

(w 2 (, T1 ))() = (w 2 (, T1 ))() = 0. (9.3.3)

Equation (9.3.3) shows that (w 2 (, T1 ))() = 0 for all 3-cell , proving that
(w 2 (, T1 )) = 0. Also, formula (9.3.2) show that [w 2 (, T1 )] H 2 (X ) does not
depend on the choice of T1 and thus depends only on . Finally, if  isomorphic to

via a homeomorphism h : E(  ) E( ) over the identity of X , the trivialization
T1 = T1 h satisfies w 2 (  , T1 ) = w 2 (, T1 ). Therefore, w 2 (  ) = w 2 ( ). 
Lemma 9.3.2 Let be an orientable vector bundle of rank r 2 over a CW-complex
X satisfying the hypotheses of Lemma 9.3.1. If is a trivial vector bundle over X ,
then w 2 ( ) = w 2 ( ).
9.3 The Class w 2 Spin Structures 369

Proof Note that is orientable since is so, thus w 2 ( ) is defined. Let

us represent w 2 ( ) by a cocycle w 2 (, T1 ) as above. If T : X Rs E() is
a fixed trivialization of , one checks that w(, T1 T ) = 1 j w(,
T1 ), where
j : S O(r ) S O(r + s) is the inclusion. But 1 j is an isomorphism if r 3 or the
reduction mod 2 if r = 2, which proves the lemma in these cases.
If r = 1, then w 2 ( ) = 0 by convention. Also, as is orientable, it is trivial by
Proposition 9.2.4. Thus, is trivial and w 2 ( ) = 0 by the proof that (2)
implies (3) in the next proposition (this part of the proof is valid for any r ). 
Example 9.3.3 Let X = S 2 , with its cellular decomposition with one 0-cell and
one 2-cell: X = D 2 S 1 pt. Let : S 1 G L + (r, R), representing []
1 (G L + (r, R)). Then

E([] ) = D 2 Rr pt Rr

is the total space of a vector bundle [] of rank r over X . This process gives a
bijection between 1 (G L + (r, R)) and the isomorphism classes of vector bundles of
rank r over S 2 (compare [181, Sect. 18]). The two cochain associating to the 2-cell of
X the element [] 1 (G L + (r, R)) = Z2 if r 3 (or its reduction mod 2 if r = 2)
represents w 2 ([] ) H 2 (X ) = Z2 . Summing up, there are two vector bundles
(up to isomorphism) of rank 3 over S 2 : the trivial bundle , satisfying w 2 () = 0,
and the non-trivial bundle , characterized by the property that w2 ( ) = 0. This is
an example of Proposition 9.3.4.
Proposition 9.3.4 Let be a vector bundle of rank r 3 over a CW-complex X .
Suppose that X has no 3-cell or that X 3 is a regular complex. Let i be the restriction
of over the X i . Then, the following conditions are equivalent.
(1) 3 is trivial.
(2) 2 is trivial.
(3) w1 ( ) = 0 and w 2 ( ) = 0.
In the next section, Proposition 9.3.4 will be generalized, using the singular second
Stiefel-Whitney class (see Proposition 9.4.6).
Proof By Proposition 9.2.7, w1 ( ) = 0 if 2 is trivial. Also, is orientable by
Proposition 9.2.1, so w 2 ( ) is defined. We can represent w 2 ( ) by a cocycle w 2 (, T1 )
where T1 is the restriction of  T (2 ). For each 2-cell e, we thus have T e = |e ,
thus w 2 (, T1 ) = 0, proving that w 2 ( ) = 0. Thus, (2) implies (3).
Conversely, suppose that w1 ( ) = 0 and w 2 ( ) = 0. Choose a trivialization
T1 of 1 and let : 2 (D 2 , S 1 ) (X, X 1 ) be a global characteristic map
for the 2-cells of X . Then, w 2 (, T1 ) = (a) for a C 1 (X ). As in the proof
of Lemma 9.3.1, the cochain a may be used to modify (relative to 0 ) T1 into a
trivialization T1 such that w 2 (, T1 ) = 0 (this uses that r 3). This means that, over
2 S 1 , the trivialization T1 coincides up to isotopy with the unique (up to isotopy)
trivialization over 2 D 2 compatible with the orientation. By Lemma 9.1.7, this
implies that 2 is trivial.
370 9 Stiefel-Whitney Classes

We have so far proven that (2) (3). We now prove the equivalence (1) (2),
which is true for any CW-complex. The implication (1) (2) is trivial. Conversely,
let  T (2 ). Let : 3 (D 3 , S 2 ) (X, X 2 ) be a global characteristic map
for the 3-cells of X . As above, one compares the trivialization  over 3 S 2
with the unique (up to isotopy) trivialization over 3 D 3 which is compatible
with the orientation. Their isotopy classes differ by the action of an element of
[3 S 2 , G L + (r, R)]. But 2 (S O(r )) = 0 (see, e.g. [181, 22.10]), thus

[S 2 , G L + (r, R)] [S 2 , S O(r )] 2 (S O(r )) = 0.

As above, using Lemma 9.1.7, this implies that 3 is trivial. 

We now see the second Stiefel-Whitney class w 2 ( ) as the obstruction to the

existence of the existence of a spin structure on . This refers to the standard orthog-
onal representation 0 : Spin(r ) S O(r ) which is a 2-covering. As in 9.1.12,
it is natural to fix an Euclidean structure and an orientation on , in which case, a
spin structure is seen a 0 -map f : P Fra+ ( ), where P is a Spin(r )-principal
bundle. Note that f is a 2-covering whose restriction to each fiber is modeled by 0 ,
and any such covering would be a spin structure. Alternative equivalent definitions
of spin structures are to be found in [150] or [130, Sect. 2.1].
Strong and weak equivalences for spin structures may be expressed as in (a) and (b)
in 9.1.9 but then, by the homotopy lifting property for coverings, these two notions of
equivalence coincide. Since 1 (S O(r )) is cyclic, 0 is, up to equivalence, the only 2-
covering of S O(r ) which is non-trivial (see Sect. 4.3.1). These considerations prove
the following lemma.

Lemma 9.3.5 Let be an oriented Euclidean bundle of rank r over a CW-complex

X . Then, there is a bijection between
the strong (or weak) equivalence classes of spin structures on .
the isomorphism classes of 2-coverings of Fra+ ( ) whose restriction to the fibers
of pFra is non-trivial.

In general, the above sets are empty. For the existence of a spin structure, one has
the following proposition (see Proposition 9.4.6 for a more general framework).

Proposition 9.3.6 Let be a vector bundle of rank r 2 over a regular CW-complex

X . Then, the following conditions are equivalent.
(a) admits a spin structure.
(b) w1 ( ) = 0 and w 2 ( ) = 0.
Moreover, if (a) or (b) holds true, then the set of strong (or weak) equivalence classes
of spin structures on is in bijection with H 1 (X ).

Proof Let (P, f ) be a spin structure on and let P2 X 2 be the restriction of P

onto the 2-skeleton of X . Suppose first that r 3. As 1 (Spin(r )) = 0, the principal
bundle P2 X 2 admits a section by [181, Corollary 34.4] and is therefore trivial.
9.3 The Class w 2 Spin Structures 371

This implies that 2 is trivial and (b) holds by Proposition 9.3.4. When r = 2, we
apply the above argument to where is a trivial bundle and use Remark 9.2.2
and Lemma 9.3.2.
Conversely, suppose that w1 ( ) and w 2 ( ) vanish. Fix an orientation and an
Euclidean structure on (thus inducing an orientation and an Euclidean structure on
k ). Suppose first that r 3. By Proposition 9.3.4, 2 is trivial and thus admits a spin
structure f2 : P2 Fra+ (2 ). Consider the commutative diagram

2 (X 2 ) / 1 (S O(r )) i
/ 1 (Fra + (2 )) / 1 (X 2 ) / 0 (S O(r ))

    
2 (X ) / 1 (S O(r )) i
/ 1 (Fra + ( )) / 1 (X ) / 0 (S O(r ))

(9.3.4)
where the rows are the homotopy exact sequences of the bundles Fra+ ( ) for = 2
or . By Lemma 9.3.5, the set of strong equivalence classes of spin structures on
is in bijection with the set

E( ) = { hom(1 (Fra+ ( )), Z2 ) | i = 0}.

Diagram (9.3.4) implies that 1 (Fra+ (2 )) 1 (Fra+ ( )) is an isomorphism (five-

lemma) and that E( ) = E(2 ). Hence, the spin structure on 2 extends to .
In the case r = 2, we apply the above argument to where is a trivial
bundle of rank 1. Note that w1 ( ) and w 2 ( ) vanish by Remark 9.2.2 and
Lemma 9.3.2. Hence, admits a spin structure. Taking the sum with a fixed
frame of gives a (S O(2) S O(3))-map Fra+ ( ) Fra+ ( ) and thus a
commutative diagram

2 (X ) / 1 (S O(2)) i
/ 1 (Fra + ( )) / 1 (X ) / 0 (S O(2))

    
2 (X ) / 1 (S O(3)) i
/ 1 (Fra + ( )) / 1 (X ) / 0 (S O(3))

(9.3.5)
As E( ) = , Diagram (9.3.5) shows that E( ) = and hence admits a
Spin(2)-structure.
Finally, note that, in (9.3.4), the last horizontal map vanish (0 (S O(r ))
0 (Fra+ ( )) is injective). Also, hom(1 (Fra+ ( )), Z2 ) is an Abelian group. Hence,
if E( ) = , it is in bijection with

{ hom(1 (Fra+ ( )), Z2 ) | i = 0} hom(1 (X ), Z2 ) H 1 (X ),

the last isomorphism being established in Lemma 4.3.1. 

372 9 Stiefel-Whitney Classes

9.4 Definition and Properties of Stiefel-Whitney Classes

Let = ( p : E X ) be a vector bundle of rank r over a paracompact space X . In

Theorem 4.7.37 was established the Thom isomorphism

 : H k (X )
H k+r (E, E 0 ),

where E 0 E is the complement of the zero section. Using the Steenrod squaring
Sq : H (E, E 0 ) H (E, E 0 ), the (total) Stiefel-Whitney class w( ) H (X ) of
is defined by

w( ) = 1 Sq (1), (9.4.1)

where 1 H 0 (X ) is the unit class. The component of w( ) in H i (X ) is denoted by

wi ( ) and is called the i-th Stiefel-Whitney class of .
Equation (9.4.1) is one of the multiple definitions of the Stiefel-Whitney class and
it is due to Thom [190, Sects. II and III]. For a history of Stiefel-Whitney classes,
see [153, p. 38] and [40, Chap. IV, Sect. 1]. The main properties of Stiefel-Whitney
class are given in the following proposition.
Theorem 9.4.1 Let = ( p : E X ) be a vector bundle of rank r over a para-
compact space X .
(1) w( ) = 1 + w1 ( ) + wr ( ). In particular, wi ( ) = 0 if i > r .
(2) If f : Y X be a continuous map, then w( f ) = H f (w( )). In particular,
if is isomorphic to  , then w( ) = w(  ).
(3) If is trivial, then w( ) = 1.
(4) Let  be a vector bundle over a paracompact space X  . Then,

w(  ) = w( ) w(  ) H (X X  ). (9.4.2)

If X  = X , then

w(  ) = w( )  w(  ) H (X ). (9.4.3)

(5) If is a trivial vector bundle over X , then w( ) = w( ).

(6) wr ( ) is the Euler class of .
Property (2) says that the total Stiefel-Whitney and its homogeneous compo-
nents are characteristic classes. Historically, they were the first characteristic classes
defined, in the simultaneous work of Eduard Stiefel and Hassler Whitney starting
around 1935 (see [40, pp. 421426]).
Proof Recall that (1) is the Thom class U ( ) H r (E, E 0 ); thus, (9.4.1) is
equivalent to
9.4 Definition and Properties of Stiefel-Whitney Classes 373

w( ) = 1 Sq(U ( )). (9.4.4)

Now, Theorem 9.4.1 comes from the properties of Sq established in Theorem 8.2.1.
Since Sq0 = id, one has w0 ( ) = 1 Sq0 (U ( )) = 1. As Sqi (U ( )) = 0 for i > r ,
this proves (1). The naturality (2) comes from the naturality of all the ingredients
of (9.4.4): the Thom class is natural (Lemma 4.7.30), and so is , and Sq is also
natural, being a cohomology operation. Now, (3) is a consequence of (2) since a
trivial bundle is induced by a map to a point.
To prove (4), one has

w(  ) = 1 Sq(U 
 ( ))  
1
=  Sq U ( ) U ( ) (9.4.5)
 using (4.7.24)
= 1 Sq(U ( )) Sq(U (  )) by (3) of Theorem 8.2.1

On the other hand, if a H (X ) and a  H (X  ), one has

(a a  ) = H ( p p  )(a a  )  U (  )
= [H p(a) H p  (a  )]  [U ( ) U (  )]
(9.4.6)
= [H p(a)  U ( )] [H p(a  )  U (  )]
= (a) (a  ).

Thus, (9.4.5) together with (9.4.6) proves (9.4.2). If X  = X , then  = (  )

where : X X X is the diagonal inclusion. Therefore, (9.4.3) comes from (2)
already proven, (9.4.2) and Remark 4.6.1:

w(  ) = H (w(  )) = H (w( ) (  )) = w( )  w(  ).

Property (5) is a consequence of (3) and (4). Finally, (6) follows from

wr ( ) = 1 Sqr (U ( )) = 1 (U ( )  U ( )) = e( ),

the last equality coming from (4.7.22). 

Remark 9.4.2 Versions of Properties (1), (2), (9.4.3) and (6) uniquely characterize
the total Stiefel-Whitney class. See Proposition 9.6.4, [153, Theorem 7.3] or [105,
Chap. 16, Sect. 5]. This is the philosophy of the axiomatic presentation of Stiefel-
Whitney class (see [153]), inspired by that of the Chern classes introduced by Hirze-
bruch [96, p. 58].

Remark 9.4.3 As the Steenrod squares are used for Definition (9.4.1), the Adem rela-
tions provide constraints amongst Stiefel-Whitney classes. For instance, the relation
Sq2i+1 = Sq1 Sq2i (see Example 8.4.4) implies that w2i+1 ( ) = 0 if w2i ( ) = 0.
Also, if w2k ( ) = 0 for k = 1 . . . , r , then, by Lemma 8.5.1, w j ( ) = 0 for
0 < j < 2r +1 .
374 9 Stiefel-Whitney Classes

We now discuss the relationship with the classes w1 and w2 defined in Sects. 9.2
and 9.3.

Proposition 9.4.4 Let be a vector bundle over a CW-complex X . Then, the first
Stiefel-Whitney class w1 ( ) H 1 (X ) defined above coincides with that defined in
Sect. 9.2. In particular, w1 ( ) = 0 if and only if is orientable.

Proof Both definitions enjoy naturality for induced bundles. We can then restrict
ourselves to X being 1-dimensional, since H 1 (X ) H 1 (X 1 ) is injective. In this
case, where is a line bundle and a trivial vector bundle (see e.g. [105,
Chap. 8, Theorem 1.2]). By Remark 9.2.2 and (5) of Theorem 9.4.1, we are reduced
to the case of a line bundle. Then, both definitions coincide with the Euler class by
Corollary 9.2.5 and Point (6) of Theorem 9.4.1. 

A similar result holds for the second Stiefel-Whitney class.

Proposition 9.4.5 Let be an orientable vector bundle over a CW-complex X .

Suppose that X has no 3-cells or that X 3 is a regular complex. Then, the second
Stiefel-Whitney class w2 ( ) H 2 (X ) defined above coincides with the cellular one
w 2 ( ) H 2 (X ) defined in Sect. 9.3.

Proof Recall that the condition on X (and the orientability of ) was necessary
for us to define w 2 ( ). The coincidence between w2 ( ) H 2 (X ) and w 2 ( )
H 2 (X ) holds under the identification of H 2 (X ) and H 2 (X ) as the same subgroup of
H 2 (X 2 ) (see (3.5.5)). The class w2 is natural by Point (2) of Theorem 9.4.1 and, by
construction, w 2 is natural for the restriction to a subcomplex. We can thus suppose
that X = X 2 and that X is connected.
As is orientable, its restriction over X 1 is trivial. By Lemma 9.1.15, p ,
where p : X X = X/ X 1 . Again, w2 ( ) = H p(w2 ( )) and, by construction of
w 2 , w 2 ( ) = H p(w 2 ( )). We can thus suppose that X is a bouquet of 2-sphere, or
even that X = S 2 with its minimal cell decomposition.
If is a trivial bundle, both equations w2 ( ) = w2 ( ) and w 2 ( ) = w 2 ( )
hold true, by Point (5) of Theorem 9.4.1 and Lemma 9.3.2. We can thus suppose that
has rank 3. As seen in Example 9.3.3, there is only one non-trivial such bundle over
S 2 , characterized by w 2 ( ) = 0. Let C be the tautological bundle over CP 1 S 2 .
By Proposition 6.1.10, one has

0 = e(C ) = w2 (C ) = w2 (C ).

which finishes the proof of our proposition. Incidentally, we have proven that C is
stably non-trivial. 

Proposition 9.4.5 permits us to generalize the framework of Propositions 9.3.4 and

9.3.6.

Proposition 9.4.6 Let be a vector bundle of rank r 3 over a CW-complex X .

Then, the following conditions are equivalent.
9.4 Definition and Properties of Stiefel-Whitney Classes 375

(1) w1 ( ) = 0 and w2 ( ) = 0.
(2) the restriction 3 of over X 3 is trivial.

Proof By Theorem 9.4.1, (1) implies w1 (2 ) = 0 and w2 (2 ) = 0. As 2 has no 3-

cells, w 2 (2 ) is defined and, by Proposition 9.4.5, w 2 (2 ) = 0. By Proposition 9.3.4,
2 is trivial which, as seen in the proof of Proposition 9.3.4, implies that 3 is trivial.
Thus, (1) implies (2). To prove that (2) implies (1), let j : X 3 X denote the
inclusion. Then j (wi ( )) = wi (3 )) = 0 for and j : H k (X ) H k (X 3 ) is
injective for k 3 (We have also proven that (1) implies w3 ( ) = 0, but this is
already known by Remark 9.4.3). 

Proposition 9.4.7 Let be a vector bundle of rank r 2 over a CW-complex X .

Then, the following conditions are equivalent.
(1) w1 ( ) = 0 and w2 ( ) = 0.
(2) admits a spin structure
Moreover, if (2) holds true, then the set of strong (or weak) equivalence classes of
spin structures on is in bijection with H 1 (X ).

Proof Suppose first that r 3. If admits a spin structure, then 2 is trivial (see the
proof of Proposition 9.3.6), which implies (1) by Proposition 9.4.6. Conversely, if (1)
holds true, then 2 is trivial by Proposition 9.4.6 and thus 2 admits a spin-structure.
That this structure extends to is established as in the proof of Proposition 9.3.6. For
the case r = 2 as well as for the last assertion of the proposition, the proofs are the
same as those for Proposition 9.3.6. 

9.5 Real Flag Manifolds

Most of the results of this section come from [15], but we do not use spectral
sequences. The Leray-Hirsch Theorem 4.7.17 for locally trivial bundles, together
with some perfect Morse theory, is sufficient for our needs. We shall deal with homo-
geneous spaces of the form / 0 , where  is a Lie group and 0 a compact subgroup
(therefore, a Lie subgroup). Then / 0 inherits a smooth manifold structure [37,
Chap. 1, Proposition 5.3]. More generally, [20, Chap. II, Theorem 5.8] implies the
following lemma.

Lemma 9.5.1 Let  be a Lie group and H G be compact subgroups of . Then,

the quotient map /H /G is a smooth locally trivial fiber bundle with fiber
G/H . If H = {1}, then the quotient map  /G is a smooth G-principal
bundle.
376 9 Stiefel-Whitney Classes

9.5.1 Definitions and Morse Theory

Let n 1 , . . . nr be positive integers and let n = n 1 + n 2 + nr . By the flag

manifold Fl(n 1 , . . . , nr ), we mean any smooth manifold diffeomorphic to the homo-
geneous space

Fl(n 1 , . . . , nr ) O(n) O(n 1 ) O(n 2 ) O(nr ). (9.5.1)

Here are some examples.

(1) Nested subspaces. Fl(n 1 , . . . , n
r ) is the set of nested vector subspaces V1
Vr Rn with dim Vi = ij=1 n j .
(2) Mutually orthogonal subspaces. Fl(n 1 , . . . , nr ) is the set of r -tuples (W1 , . . . ,
Wr ) of vector subspaces Rn which are mutually orthogonal and satisfy
dim Wi = n i . The correspondence from this definition to Definition (1) asso-
ciates to (W1 , . . . , Wr ) the nested family {Vi } where Vi is the vector space
generated by W1 Wi .
(3) Isospectral symmetric matrices. Let 1 > > r be real numbers. Consider
the manifold S M(n) of all symmetric real (n n)-matrices, on which O(n) acts
by conjugation. Then Fl(n 1 , . . . , nr ) occurs as the orbit of the diagonal matrix
having entries i with multiplicity n i .
  
Fl(n 1 , . . . , nr ) = R dia 1 , . . . , 1 , , r , . . . , r R 1 | R O(n) .

 
 
n1 nr
(9.5.2)
In other words, Fl(n 1 , . . . , nr ) is here the space of
symmetric real (n n)-
r
i=1 (x i ) . Indeed,
matrices with characteristic polynomial equal to ni

elementary linear algebra teaches us that two matrices in S M(n) are in the
same O(n)-orbit if and only if they have the same characteristic polynomial.
The correspondence from this definition to Definition (2) associates, to a matrix
M, its eigenspaces for the various eigenvalues.
Concrete definition (3) will be our working definition for Fl(n 1 , . . . , nr ) throughout
this section. Special classes of flag manifolds are given by the Grassmannians

Gr(k; Rn ) = Fl(k, n k) O(n) O(k) O(n k)

of k-planes in Rn . This is a closed manifold of dimension

dim Gr(k; Rn ) = dim O(n) dim O(k) dim O(n k) = k(n k).

For example, Gr(1; Rn ) RP n1 , of dimension n 1. Using Definition (3) above,

our concrete Grassmannian will be
9.5 Real Flag Manifolds 377
  
Gr(k; Rn ) = R dia 1, . . . , 1, 0, . . . , 0 R 1 | R O(n) . (9.5.3)

 
 
k nk

In other words, Gr(k; Rn ) is the space of orthogonal projectors on Rn of rank k.

Another interesting flag manifold is the complete flag manifold

Fl(1, . . . , 1) O(n) O(1) O(1)

with dim Fl(1, . . . , 1) = dim O(n) = n(n1)

2 .
We now define real functions on the flag manifolds by restriction of the weighted
trace on f : S M(n) R defined by


n
f (M) = j Mjj
j=1

where Mi j denotes the (i, j)-entry of M.

Proposition 9.5.2 Let Fl(n 1 , . . . , nr ) S M(n) be the flag manifold as presented
in (9.5.2). Then, the restriction f : Fl(n 1 , . . . , nr ) R of the weighted trace is a per-
fect Morse function whose critical points are the diagonal matrices in Fl(n 1 , . . . , nr ).
The index of the critical point dia(x1 , . . . , xn ) is the number of pairs (i, j) with i < j
and xi < x j .
For a general discussion about such Morse functions on flag manifolds, see
[13, Chap. 8].
Example 9.5.3 For Gr(2; R5 ) = Fl(2, 3), we get the following (52 ) = 10 critical
points, with their index and value by f .

Critical point Index Value

dia(1, 1, 0, 0, 0) 0 3
dia(1, 0, 1, 0, 0) 1 4
dia(1, 0, 0, 1, 0), dia(0, 1, 1, 0, 0) 2 5
dia(1, 0, 0, 0, 1), dia(0, 1, 0, 1, 0) 3 6
dia(0, 0, 1, 1, 0), dia(0, 1, 0, 0, 1) 4 7
dia(0, 0, 1, 0, 1) 5 8
dia(0, 0, 0, 1, 1) 6 9

Remark 9.5.4 The function f : Gr(k; Rn ) R given by

k(k + 1)
f(M) = + f (M)
2

is a Morse function which is self-indexed, i.e. f(M) = j if M is a critical point of

index j.
378 9 Stiefel-Whitney Classes

Proof of Proposition 9.5.2 We introduce precise notations which will be used later.
For 1 i < j n, let r i j : M2 (C) Mn (C) defined by requiring that the entries
of r i j (N ) are those of the identity matrix In , except for

r i j (N )ii = N11 , r i j (N )i j = N12 , r i j (N ) ji = N21 , r i j (N ) j j = N22 .

The restriction of r i j to S O(2) gives an injective homomorphism r i j : S O(2)

S O(n) whose image is formed by the matrices
 
ij cos t sin t
Rt =r ij
(t R).
sin t cos t

ij
The action of Rt on Fl(n 1 , . . . , nr ) S M(n) by conjugation produces a flow and
ij
thus a vector field V i j on Fl(n 1 , . . . , nr ), whose value VM at M Fl(n 1 , . . . , nr ) is
ij i j i j
VM = dt d
(Rt M Rt )|t=0 (we identify TM Fl(n 1 , . . . , nr ) as a subspace of S M(n)).
A direct computation gives that

ij ij
(Rt M Rt )ii = Mii cos2 t Mi j sin 2t + M j j sin2 t
ij ij
(Rt M Rt ) j j = Mii sin2 t + Mi j sin 2t + M j j cos2 t (9.5.4)
ij ij
(Rt M Rt )i j = Mi j cos 2t + (Mii M j j ) sin t cos t.

Moreover,
ij ij
(Rt M Rt )ik = Mik cos t M jk sin t if i = k = j
ij ij
(Rt M Rt )k j = Mki sin t + Mk j cos t if i = k = j (9.5.5)
ij ij
(Rt M Rt )kl = Mkl if i = k and j = l.

ij ij
Let g i j (t) = f (Rt M Rt ). The first derivative g i j (t) satisfies

g i j (t) = ( j i)(Mii M j j ) sin 2t + 2( j i)Mi j cos 2t.

Hence,
ij
VM f = g i j (0) = 2( j i)Mi j , (9.5.6)

which proves that only the diagonal matrices in Fl(n 1 , . . . , nr ) may be critical points
of the weighted trace.
Suppose that Fl(n 1 , . . . , nr ) is a diagonal matrix. Let

J = {(i, j) | 1 i < j n and ii = j j }.

9.5 Real Flag Manifolds 379

ij
and let V = {V | (i, j) J } T Fl(n 1 , . . . , nr ). By (9.5.4) and (9.5.5),
ij ij
dt (Rt Rt )(0) has only non-zero term away from the diagonal, namely dt
d d
ij ij
(Rt Rt )i j (0) = ii j j . Hence, vectors of V are linearly independent. But

n(n 1)  n k (n k 1) 
r r
J = = dim O(n) dim O(n k ) = dim Fl(n 1 , . . . , nr ).
2 2
k=1 k=1

Therefore, V is a basis of T Fl(n 1 , . . . , nr ). Using (9.5.6), this proves that the

diagonal matrices in Fl(n 1 , . . . , nr ) are exactly the critical points of the weighted
trace. The matrix of the Hessian form H f on T Fl(n 1 , . . . , nr ) is

ij
H f (V kl , V ) = V kl (V i j f )
 
= V kl M 2( j i)Mi j by (9.5.6)
 d kl 
= 2( j i) dt (Rt Rt )|t=0 i j .
kl

Using (9.5.4) and (9.5.5), we see that the matrix of H f in the basis V is diagonal,
with diagonal term

ij ij
H f (V , V ) = 2( j i)( ii j j ).

As (i, j) J , this proves that f is a Morse function as well as the assertion on the
Morse index of .
It remains to prove that f is perfect. Let  be the subgroup of O(n) formed by the
diagonal matrices (with coefficients 1). The O(n)-action on S M(n) by conjugation
may be restricted to  and f is -invariant. Moreover, the diagonal matrices in
Fl(n 1 , . . . , nr ) are exactly the fixed points of the -action. The perfectness of f then
follows from Theorem 7.6.6.
Here is a first consequence of Proposition 9.5.2.
Corollary 9.5.5
n!
dim H (Fl(n 1 , . . . , nr )) = .
n 1 ! nr !

In particular,

dim H (Fl(k, n k)) = Gr(k; Rn ) = (nk ) and dim H (Fl(1, , 1)) = n!.

Proof By Proposition 9.5.2, the weighted trace f : Fl(n 1 , . . . , nr ) R is a perfect

Morse function. Hence, by Proposition 7.6.4, dim H (Fl(n 1 , . . . , nr )) = Crit f .
But Crit f consists of the diagonal matrices in Fl(n 1 , . . . , nr ), which are all conjugate
to
 
dia 1 , . . . , 1 , , r , . . . , r

 
 
n1 nr
380 9 Stiefel-Whitney Classes

by a permutation matrix. Hence, Crit f is an orbit of the symmetric group Symn ,

with isotropy group Symn 1 Symnr , whence the formulae. 
Remark 9.5.6 The critical points of f in Proposition 9.5.2 are related to the Schubert
cells (see Sect. 9.5.3).
Consider the inclusion S M(n) S M(n + 1) with image the matrices with van-
ishing last row and column. Seeing Gr(k; Rn ) S M(n) as in (9.5.3), this gives an
inclusion Gr(k; Rn ) Gr(k; Rn+1 ).
Lemma 9.5.7 The homomorphism H j (Gr(k; Rn+1 )) H j (Gr(k; Rn )) induced
by the inclusion is surjective for all j and is an isomorphism for j n k.
Proof Let us use the Morse function f : Gr(k; Rn+1 ) R of Remark 9.5.4 and
let f be its restriction to Gr(k; Rn ). Then, f and f are self-indexed and Crit f
Crit f N. For m N, let Wm = f1 ((, m + 1/2]) and Wm = ( f )1 ((, m +
1/2]). For the first assertion, we prove, by induction on m that H (Wm ) H (Wm )
is surjective for all m N. The induction starts with m = 0, since W0  W0  pt.
The induction step involves the cohomology sequences

0 / H (Wm , Wm1 ) / H (Wm ) / H (Wm1 ) / 0

i m,m1
i m1
im (9.5.7)
  
0 / H (Wm , W  ) / H (W  ) / H (W  ) / 0
m1 m m1

obtained by Lemma 7.6.3, since f and f are perfect by Proposition 9.5.2. From
Proposition 9.5.2 again and its proof, the critical points of f have the negative normal

directions in Wm1 or in Wm1 . Hence, using excision, the Morse lemma and Thom
isomorphisms, we get the commutative diagram


H (Wm , Wm1 ) / CCrit f f1 (m) H m (C)

i m,m1 pr oj
 


H (Wm , Wm1 ) / CCrit f f1 (m) H m (C)

which proves that i m,m1
is onto. If i m1 is surjective by induction hypothesis, we

get that i m is surjective by diagram-chasing.
Note that the point D Crit f Crit f of lowest index is

D = dia(1, . . . , 1, 0, . . . , 0, 1) S M(n + 1)

satisfies f(D) = index(D) = n k + 1 (the number of zeros in D). Hence,

Crit f Wnk = Crit f Wnk
 . The same induction argument as above shows that

H (Gr(k; R )) H (Gr(k; Rn )) is an isomorphism for j n k.

j n+1 j 
9.5 Real Flag Manifolds 381

9.5.2 Cohomology Rings

The cohomology ring of a flag manifold V will be generated by Stiefel-Whitney

classes of some tautological bundles over V . Consider a flag manifold Fl(n 1 , . . . , nr ),
with n = n 1 + + nr . Consider the following closed subgroups of O(n).

Bi = O(n 1 ) {1} O(nr ) O(n 1 ) O(nr ) O(n),

where {1} sits at the i-th place. Then


Pi = O(n)/Bi
O(n) O(n 1 ) O(nr ) = Fl(n 1 , . . . , nr )

is an O(n i )-principal bundle over Fl(n 1 , . . . , nr ). Indeed, if K is a compact sub-

group of a Lie group G, then G G/K is a principal K -bundle (see, e.g. [12,
Theorem 2.1.1, Chap. I]). Let i be the vector bundle of rank n i associated to Pi ,
i.e. E(i ) = Pi O(n i ) Rn i . The vector bundle i is called the i-th-tautological vec-
tor bundle over Fl(n 1 , . . . , nr ). Being associated to an O(n i )-principal bundle, i is
endowed with an Euclidean structure and its space of orthogonal frames Fra (i ) is
equal to Pi .
In the mutually orthogonal subspaces description (presentation (2), p. 376) of
Fl(n 1 , . . . , nr ), we see that

E(i ) = {(W1 . . . , Wr , v) Fl(n 1 , . . . , nr ) Rn | v Wi }.

Note that 1 r is trivial. Indeed,

E(1 r ) = {((W1 , . . . , Wr ), (v1 , . . . , vr )) Fl(n 1 , . . . , nr ) (Rn )r | vi Wi }

and the correspondence

((W1 , . . . , Wr ), (v1 , . . . , vr )) v1 + + vr (9.5.8)

restricts to a linear isomorphism on each fiber. Such a map thus provides a trivial-
ization of 1 r .
If one sees Fl(n 1 , .
. . , nr ) as the space of matrices M S M(n) with characteristic
polynomial equal to ri=1 (x i )n i (presentation (3), p. 376), then

E(i ) = {(M, v) Fl(n 1 , . . . , nr ) Rn | Mv = i v}. (9.5.9)

The vector bundle 1 over Fl(k, n k) = Gr(k; Rn ) is called the tautological

vector bundle over the Grassmannian Gr(k; Rn ); it is of rank k and is denoted by ,
k or k,n . The space of Fra (k ) is the Stiefel manifold Stief(k, Rn ) of orthonormal
k-frames in Rn .
382 9 Stiefel-Whitney Classes

The inclusion Rn Rn {0}  Rn+1 induces an inclusion Gr(k; Rn ) 

Gr(k; Rn+1 ) and we may consider the inductive limit

Gr(k; R ) = lim Gr(k; Rn )

n

which is a CW-space. The tautological vector bundle k is also defined over

Gr(k; R ) and induces that over Gr(k; Rn ) by the inclusion Gr(k; Rn ) 
Gr(k; R ). It is classical that i (Stief(k, Rn )) = 0 for i < n k (see [181,
Theorem 25.6]), thus Stief(k, R ) = Fra(k ) is contractible. Hence, the O(k)-
principal bundle Stief(k, R ) Gr(k; R ) is a universal O(k)-principal bundle
(see [181, Sect. 19.4]) and thus homotopy equivalent to the Milnor universal bundle
E O(k) B O(k). In particular, Gr(k; R ) has the homotopy type of B O(k). As
a consequence, any vector bundle of rank k over a paracompact space X is induced
from k by a map X Gr(k; R ) (for a direct proof of that (see [153, Theorem 5.6]).

Theorem 9.5.8 The cohomology ring of B O(k) is GrA-isomorphic to the polyno-

mial ring

H (B O(k)) H (Gr(k; R )) Z2 [w1 , . . . , wk ]

generated by the Stiefel-Whitney classes wi = wi (k ) of the tautological bundle k .

Proof Slightly more formally, we consider the polynomial ring Z2 [w1 , . . . , wk ] with
formal variables wi of degree i. The correspondence wi wi (k ) provides a GrA-
morphism : Z2 [w1 , . . . , wk ] H (B O(k)) which we shall show that it is
bijective.
For the injectivity, we consider the tautological line bundle over RP and its
n-times product n over (RP )n . As seen above, k is universal so n is induced by
a map f : (RP )n B O(n). Recall from Proposition 4.3.10 that H (RP ) =
Z2 [a] with a of degree 1 and, by Theorem 9.4.1, w( ) = 1 + a. By the Knneth

theorem, there is a GrA-isomorphism Z2 [a1 , . . . , an ] H ((RP )n ) and, by
Theorem 9.4.1, w( n ) = (1 + a1 )   (1 + an ). As H f (w(k )) = w( n ),
there is a commutative diagram

Z2 [w1 , . . . , wk ] / H (B O(k))

H f
 
/ H ((RP )n )
Z2 [a1 , . . . , an ]

with

(w(k )) = (1 + a1 ) (1 + an ) = 1 + 1 + n ,
9.5 Real Flag Manifolds 383

where i = i (a1 , . . . , an ) is the i-th elementary symmetric polynomial in the vari-

ables a j (see (8.5.4)). Thus, (wi ) = i . Now, if 0 = A Z2 [w1 , . . . , wk ] satisfies
(A) = 0, then (A) = 0 would be a non-trivial polynomial relation between the
i s. But the elementary symmetric polynomials are algebraically independent (see
e.g. [122]). Thus, is injective.
For d N, let


k
Bd = {(d1 , . . . , dk ) Nk | j d j = d}.
j=1

The correspondence (d1 , . . . , dk ) w1d1 wkdk is a bijection from Bd onto a basis

of the vector subspace Z2 [w1 , . . . , wk ][d] formed by the elements in Z2 [w1 , . . . , wk ]
which are of degree d. On the other hand, consider Gr(k; Rn ) S M(n) as in (9.5.3),
with n large. Let Critd f Gr(k; Rn ) be the set of critical points of index d for the
weighted trace. Then the correspondence

(d1 , . . . , dk ) dia(0, . . . , 0, 1, 0, . . . , 0, 1, . . . , 0, . . . , 0, 1, 0, . . . , 0)

 
 
 
dk dk1 d1

provides a bijection Bd
Critd f . As f is a perfect Morse function by Proposition
9.5.2, one has

 Bd =  Critd f = dim H d (Gr(k; Rn )) = dim H d (B O(k)),

the last equality coming from Lemma 9.5.7 when n is large enough. Therefore,

dim Z2 [w1 , . . . , wk ][d] = dim H d (B O(k)).

As is injective, it is then bijective. 

Define
1
Q r (t) = = 1 + t r + t 2r + Z[[t]], (9.5.10)
1 tr

which is the Poincar series of Z2 [x] if x is of degree r . Here is a direct consequence

of Theorem 9.5.8.

Corollary 9.5.9 The Poincar series of B O(k) is

Pt (B O(k)) = Q 1 (t) Q k (t).

As any vector bundle of rank k over a paracompact space is induced from the
universal bundle k [153, Theorem 5.6], Theorem 9.5.8 has the following corollary.
384 9 Stiefel-Whitney Classes

Corollary 9.5.10 Any characteristic class in mod 2 cohomology for vector bun-
dles of finite rank over paracompact spaces is a polynomial in the Stiefel-Whitney
classes wi .

Also, Theorem 9.5.8 together with Lemma 9.5.7 gives the following corollary.

Corollary 9.5.11 The cohomology ring H (Gr(k; Rn )) is generated, as a ring, by

the Stiefel-Whitney classes w1 (k ), . . . , wk (k ) of the tautological bundle k .

Theorem 9.5.8 permits us to compute the cohomology of B S O(k). The latter also
has a tautological bundle k = (E S O(k) S O(k) Rk B S O(k)) which is orientable.

Corollary 9.5.12 The cohomology ring of B S O(k) is GrA-isomorphic to the poly-

nomial ring

H (B S O(k)) Z2 [w2 , . . . , wk ]

generated by the Stiefel-Whitney classes wi = wi (k ) of the tautological bundle k .

Proof Let i : S O(k) O(k) denote the inclusion. By Example 7.2.4, the map
Bi : B S O(k) B O(k) is homotopy equivalent to a two fold covering, which
is non-trivial since B S O(k) is connected. By Lemma 4.3.6, its characteristic class
w(Bi) H 1 (B O(k)) is not trivial. By Theorem 9.5.8, the only non-zero element in
H 1 (B O(k)) is w1 (k ), so w(Bi) = w1 (k ).
By Theorem 9.5.8 and the transfer exact sequence (Proposition 4.3.9), the ring
homomorphism H Bi : H (B O(k)) H (B S O(k)) is surjective with kernel the
ideal generated by w1 (k ). As Bi is covered by a bundle map from k to k , one has
H Bi(wi (k ) = wi (k ). The corollary follows. 

Remark 9.5.13 In contrast with the simplicity of H (B S O(k)), the cohomology

ring H (B Spin(k)) is complicated and its computation requires spectral sequences
(see [167]). The stable case B Spin = limk B Spin(k) is slightly simpler (see [193]).

We are now in position to give a GrA-presentation of H (Fl(n 1 , . . . , nr )). Let

w( j ) = 1 + w1 ( j ) + + wn j ( j ) H (Fl(n 1 , . . . , nr )) (9.5.11)

be the Stiefel-Whitney class of the tautological vector bundle j . As seen in (9.5.8),

1 r is trivial. By Theorem 9.4.1, the equation

w(1 )   w(r ) = 1 (9.5.12)

holds true. Hence, the homogeneous components of w(1 )   w(r ) in positive

degrees vanish, giving rise to n equations.

Theorem 9.5.14 The cohomology algebra H (Fl(n 1 , . . . , nr )) is GrA-isomorphic

to the quotient of the polynomial ring
9.5 Real Flag Manifolds 385

Z2 [wi ( j )] , 1 i r j , j = 1, . . . , r

by the ideal generated by the homogeneous components of w(1 ) w(r ) in

positive degrees.

Proof We first prove that H (Fl(n 1 , . . . , nr )) is, as a ring, generated by the Stiefel-
Whitney classes wi ( j ) (1 i r j , j = 1, . . . , r ). This is done by induction on
r (note that r 2 in order for the definition of Fl(n 1 , . . . , nr ) to make sense). For
r = 2, as Fl(n 1 , n 2 ) = Gr(n 1 ; Rn 1 +n 2 ), the result comes from Corollary 9.5.11.
For the induction step, let us define a map : Fl(n 1 , . . . , nr ) Fl(n nr , nr ) by
(W1 . . . , Wr ) = (W1 Wr 1 , Wr ) (using the mutually orthogonal defini-
tion (2) of the flag manifolds). By Lemma 9.5.1, this gives a locally trivial bundle

Fl(n 1 , . . . , nr 1 )
Fl(n 1 , . . . , nr )
Fl(n nr , nr ).

By induction hypothesis, H (Fl(n 1 , . . . , nr 1 )) is generated, as a ring, by the Stiefel-

Whitney classes of its tautological bundles, say wi ( j ). Note that these bundles are
induced by the tautological bundles (called j ) over Fl(n 1 , . . . , nr ): j = j . Then,
H is surjective and wi ( j ) wi ( j ) is a cohomology extension of the fiber (see
p. 172). On the other hand,

Fl(n nr , nr ) Gr(n nr ; Rn ) Gr(nr ; Rn ),

the last isomorphism sending an (n nr )-dimensional subspace of Rn to its

orthogonal complement. By Corollary 9.5.11, H (Gr(nr ; Rn )) is GrA-generated by
w1 (nr ), . . . , wnr (nr ) and H (wi (nr )) = wi (r ). By the Leray-Hirsch Theorem
4.7.17, H (Fl(n 1 , . . . , nr )) is then GrA-generated by wi ( j ) (1 i n j and
j = 1 . . . r ).
Let  = O(n 1 ) O(nr ) O(n) and consider the commutative diagram

O(n) / E O(n) / B O(n)

= (9.5.13)
  
O(n)/  / E O(n)/  / B O(n)

where the top line is the O(n)-universal bundle. Hence, the bottom line is a locally
trivial bundle with fiber equal to O(n)/  = Fl(n 1 , . . . , nr ); as E O(n) is con-
tractible, there are homotopy equivalences

E O(n)/   B  B O(n 1 ) B O(nr ),

the last homotopy equivalence coming from (7.4.3). Hence, Diagram (9.5.13) may
be rewritten in the following way.
386 9 Stiefel-Whitney Classes

= / 


 
O(n) / E O(n) / B O(n) (9.5.14)
=
  

Fl(n 1 , . . . , nr ) / B B / B O(n)

where denotes the inclusion of  in O(n). The left column is a -principal bundle
which is a -structure on = 1 r . The central column is the -principal
bundle associated to the vector bundle = n 1 nr over B  B O(n 1 )
B O(nr ). Thus, the map is a classifying map for the -structure on : it
lifts the map B , which is classifying for as a vector bundle (that B is
null-homotopic is coherent with the triviality of , seen in (9.5.8)). Hence

n j j ,

and thus

H (wi (n j )) = wi ( j ) , H (wi ( )) = wi ( ).

As H (Fl(n 1 , . . . , nr )) is GrA-generated by the classes wi ( j ), H is surjec-

tive and one may apply the Leray-Hirsch Theorem 4.7.17 and its corollaries. By
Theorem 9.5.8 and the Knneth theorem, there is a GrA-isomorphism

H (B) , (1 i r j , j = 1, . . . , r ).
Z2 [wi (n j )]

On the other hand, H (B O(n)) Z2 [wi (n )] and H B(wi (n )) = wi ( ). By

Corollary 4.7.19, H (Fl(n 1 , . . . , nr )) is then GrA-isomorphic to the quotient of
H (B) by the ideal generated by wi ( ) (i > 0). Hence, one has the commuta-
tive diagram

 
Z2 [wi ( j )] wi ( ), i > 0
/ Z2 [wi ( j )]wi ( ), i > 0
TTTT jjj
TTTT jjjj
TTTT jjj
T) ju jj
H (Fl(n 1 , . . . , nr ))

which proves Theorem 9.5.14. 

In the following corollary, we use the notations Q i of (9.5.10).
Corollary 9.5.15 The Poincar polynomial of Fl(n 1 , . . . , nr ) is given by the formula
r
j=1 [Q 1 (t) Q n j (t)]
Pt (Fl(n 1 , . . . , nr )) = .
Q 1 (t) Q n (t)
9.5 Real Flag Manifolds 387

In particular,

Q 1 (t) Q k (t)
Pt (Gr(k; Rn )) = Pt (Fl(k, n k)) =
Q nk+1 (t) Q n (t)

and

Q 1 (t)n (1 t)(1 t 2 ) (1 t n )
Pt (Fl(1, . . . , 1)) = = .
Q 1 (t) Q n (t) (1 t)n

Remark 9.5.16 The above formulae, evaluated at t = 1 using LHospitals rule, give
dim H (Fl(n 1 , . . . , nr )), etc., giving again the formulae of Corollary 9.5.5.

Proof of Corollary 9.5.15 We have seen in the proof of Theorem 9.5.14 that

B
Fl(n 1 , . . . , nr )
B B O(n)

is a locally trivial bundle satisfying the hypotheses of the Leray-Hirsch theorem.

We know by Corollary 9.5.9 that Pt (B O(k)) = Q 1 (t) Q k (t). By the Knneth
formula, we get that


r
Pt (B) = Pt (B O(n 1 ) B O(nr )) = [Q 1 (t) Q n j (t)].
j=1

The first formula then comes from Corollary 4.7.20. The other formulae are conse-
quences of the first one.
We now give some illustrations of Theorem 9.5.14.

Example 9.5.17 Consider the case of the complete flag manifold Fl(1, . . . , 1).
Theorem 9.5.14 says that H (Fl(1, . . . , 1)) is generated by xi = w1 (i ) for i =
1 . . . , n. In this generating system, wi (1 n ) = i , the ith elementary
symmetric polynomial in the variables xi . Hence, by Theorem 9.5.14,
 
H (Fl(1, . . . , 1)) Z2 [x1 , . . . , xn ] 1 , . . . , n .

Example 9.5.18 Consider the case of the Grassmannian

Gr(k; Rn ) = Fl(k, n k).

Set = 1 , with Stiefel-Whitney class w( ) = w = 1 + w1 + w2 + , and

= 2 , with w( ) = w = 1 + w 1 + w 2 + . Note that the fiber of the vector
bundle over P Gr(k; Rn ) is the set of vectors in Rn which are orthogonal to
P. Equation (9.5.12) becomes
w  w = 1, (9.5.15)
388 9 Stiefel-Whitney Classes

which is equivalent to the following system of equations:


k
w i = wr w ir (i = 1, . . . , n k) and w i = 0 if i > n k. (9.5.16)
r =1

This system has the following unique solution.

Lemma 9.5.19 With the convention w i = wi = 0 for i < 0, the equation

 
holds true in H r (Gr(k; Rn )). The symmetric formula wr = det w i+1 j 1i, jr
holds true as well. These equalities are both equivalent to Eq. (9.5.15).
Proof The first equation is proved by induction on r , starting, for r = 1, with
w 1 = w1 (this also gives the uniqueness of the solution). The induction step is
achieved by expanding the determinant with respect to the first column: the (s, 1)-th
minor is equal to w r s by induction hypothesis and the result follows from (9.5.16).
The symmetric equation follows from the symmetry in wi and w i of (9.5.16) (coming
from the symmetry of (9.5.15)). 
Below are two special case of Example 9.5.18.
Example 9.5.20 Consider the case of Gr(1; Rn ) = Fl(1, n 1) RP n1 . The
relation w  w = 1 gives rise to the system of equations

w1 + w 1 = 0
w i + w i1 w1 = 0 (i = 2, . . . , n 1)
w n1 w1 = 0

from which we deduce the usual presentation H (RP n1 ) Z2 [w1 ]/(w1n ).

Example 9.5.21 In the case Gr(2; R4 ) = Fl(2, 2), the relation w  w = 1 gives
rise to four equations

w 1 = w1
w 2 = w12 + w2
(9.5.17)
w13 = 0
w2 w12 + w22 = 0.

and Theorem 9.5.14 says that H (Gr(2; R4 )) is generated by w1 , w2 , w 1 and w 2 ,

subject to Relations (9.5.17). The first two equations imply that H (Gr(2; R4 )) is
9.5 Real Flag Manifolds 389

generated by w1 and w2 , as known since Corollary 9.5.11. We check that an addi-

tive basis of H (Gr(2; R4 )) is given by 1, w1 , w2 , w12 , w2 w1 and w2 w12 = w22 . The
Poincar polynomial of Gr(2; R4 ) is given by Corollary 9.5.15:

Q 1 (t)Q 2 (t) (1 t 3 )(1 t 4 )

Pt (Gr(2; R4 )) = = = 1 + t + 2t 2 + t 3 + t 4 .
Q 3 (t)Q 4 (t) (1 t)(1 t 2 )

For any bundle of rank k over a space X the dual (or normal) Stiefel-Whitney
class w r ( ) are defined by the equation of Lemma 9.5.19. Set w(
) = 1+ w 1 ( )+
for the total dual Stiefel-Whitney class. Equations (9.5.15) and (9.5.16) are satisfied.
If there exists a vector bundle over X such that is trivial, then w() = w(
).
Thus, if is of rank r , then w i ( ) = 0 for i > r . The same condition is necessary
for being induced from the tautological bundle by a map f : X Gr(k; Rk+r ).
For example, let M be a smooth manifold of dimension k which admits an immer-
sion : M Rk+r . Let x M. By identifying T(x) Rk+r with Rk+r , the k-
vector space Tx (T M) becomes an element of Gr(k; Rk+r ). This produces a map
: M Gr(k; Rk+r ) and T M = . We thus get the following result.
Proposition 9.5.22 If a smooth manifold M of dimension k admits an immersion
into Rk+r , then w i (T M) = 0 for i > r .
For improvements of Proposition 9.5.22 concerning also smooth embeddings, see
Proposition 9.8.23 and Corollary 9.8.24. Usually, Proposition 9.5.22 does not give
the smallest integer r for which M immerses into Rk+r . This is however the case in
the following example, taken from [153, Theorem 4.8].
Proposition 9.5.23 For k = 2 j ( j 1), the projective space RP k immerses into
R N if and only if N 2k 1.
Proof That a manifold of dimension k 2 immerses into R2k1 is a classical
theorem of Whitney [210]. Conversely, we shall see in Proposition 9.8.10 that
j j +1 j
w(T RP 2 ) = (1 + a)2 = 1 + a + a2 ,
j
where 0 = a H 1 (RP 2 ) = Z2 . Hence,
j j 1
w(T
RP 2 ) = 1 + a + a 2 + + a 2 ,
j
which, using Proposition 9.5.22 implies that RP 2 does not immerses into
R2 2 .
j+1


9.5.3 Schubert Cells and Stiefel-Whitney Classes

Let f : M R be a Morse function on a manifold M. It is classical that M has

the homotopy type of a CW-complex whose r -cells are in bijection with the critical
390 9 Stiefel-Whitney Classes

points of index r of f (see, e.g. [13, Theorem 3.28]). For the weighted trace f (or
f) defined on M = Gr(k; Rn ) in Proposition 9.5.2 (or Remark 9.5.4), a very explicit
such CW-structure is given, using the Schubert cells (there are generalizations for
flag manifolds). Inspired by works of H. Schubert on enumerative geometry in the
XIXth century (see e.g. [169]), Schubert cells were introduced in 1934 (for complex
Grassmannians) by Ch. Ehresmann [47] ([48] for the real Grassmannians). See [40,
22425] for a history. We restrict ourselves here to a very elementary point of view,
Schubert calculus being a huge subject in algebraic geometry.
Recall that Crit f are diagonal matrices in S M(n). We write dia(1 , . . . , n ) =
dia(), where  = 1 n is a binary word of length n. Let [nk ] be the set or
such words with i = k (they are (nk ) in number). The correspondence 0
identifies [nk ] with a subset of [n+1
k ], permitting us to define [ k ] as the direct limit
of [nk ].
Let F = (F1 Fn ) be a complete flag in Rn (adding the convention that
F0 = {0}). For [nk ], the Schubert cell CF with respect to F is defined by


i
CF = {P Gr(k; Rn ) | dim(P Fi ) = j } Gr(k; Rn ).
j=1

(This convention is close to that of [119], except for the binary words being written in
the reverse order, so it works for n = , in the spirit of [153, Sect. 6]). The following
facts may be proven.
(1) The Schubert cells {CF | [nk ]} are the open cells of a CW-structure X F on
Gr(k; Rn ) (see [153, Sect. 6]). The dimension of CF is

k(k + 1) 
d() = index (dia()) = f(dia()) = + i .
2
i1

By Proposition 9.5.2, the cellular chains have then the same Poincar polynomial
as the homology. Therefore, X F is a perfect CW-structure.
(2) The closure C F , called the Schubert variety, satisfies


i
C F = {P Gr(k; Rn ) | dim(P Fi ) j } Gr(k; Rn )
j=1

and is a subcomplex of X F (see e.g. [47, Sect. 10]). As X F is perfect, so is C F

and thus C F defines a homology class

[] = [C F ] Hd() (Gr(k; Rn )) (n )

which does not depend on F since, by Proposition 9.5.2, the complete flag man-
ifold is path-connected. It corresponds, under the isomorphism (3.5.6) between
9.5 Real Flag Manifolds 391

cellular and singular homology, to the cellular homology class for X F indexed
by . It follows that the S = {[] H (Gr(k; Rn )) | [nk ]} is a basis of
H (Gr(k; Rn )) (n ).
(3) Let P Gr(k; Rn ). Using a basis of Rn compatible with the flag F, let M P be
the matrix of a linear epimorphism Rn Rnk with kernel P. The condition
P C F is equivalent to the vanishing of various minors of M P . Therefore,
P C F is a compact real algebraic variety. This is another proof of the existence
of the class [], since such a variety carries a fundamental class (see [192, p. 67]
or [16, Theorem 3.7 and Sect. 3.8]).
(4) Suppose that F is the standard flag (Fi = Ri 0). Then f(C F ) = [0, d()]
and C F f1 (d()) = dia(). Recall from Proposition 9.5.2 that d() is equal
to the number of pairs (i, j) with 1 i < j n such that i < j . For such
ij
a pair (i, j), let Rt be the one-parameter subgroup of S O(n) considered in the
proof of Proposition 9.5.2. Then, the curve Rt dia()Rt is contained in C F for
ij ij

t R and stays in C when |t| < /2. By the proof of Proposition 9.5.2, these
F

curves generate the negative normal bundle for f at dia().

Example 9.5.24 Consider the case of Gr(1; Rn+1 ) RP n . For F the standard flag
in Rn+1 , the Schubert cells give the standard CW-structure on RP n , the cell CF for
= 0r 10nr being of dimension r . The Schubert variety C F is equal to RP r (a rare
case where it is a smooth manifold).
Note that H ([]) = [0] where : Gr(k; Rn )  Gr(k; Rn+1 ) is induced by
the inclusion Rn Rn 0  Rn R. We often identify [] with [0]. In this way,
for instance, [100101] determines a class in H5 (Gr(k; Rn )) for n 6.
Let S  = {[] | [nk ]} (n ) be the additive basis of H (Gr(k; Rn ))
which is dual for the Kronecker pairing to the basis S (see (2) above): the class []
is defined by

[] , [] = ,

where is the Kronecker symbol. The basis S  was studied in [31, 32]. Because
of intersection theory, a more widely used additive basis for H (Gr(k; Rn )) (defined
only or n < ) is S P D , is formed by the Poincar duals [] P D for all [nk ].
Though some intersection theory in used in the proof of Proposition 9.5.29, we shall
not use S P D . We just note the following result.
Lemma 9.5.25 For any k n < , the two sets S  and S P D in H (Gr(k; Rn ))
are equal.
Proof Let F be the standard flag (Fi = Ri 0) and let F be the anti-standard
one (Fi = 0 Ri ). For [nk ], define [nk ] by i = n+1i . The cycles

C F and C F are of complementary dimensions and, by (4) above, they intersect
transversally in a single point (the k-plane generated by i ei for i = 1, 2, . . . , n). In

the same way, if [nk ] satisfies d() = d() but = , then C F C F = .
Analogously to Proposition 5.4.12, one has
392 9 Stiefel-Whitney Classes

[ ] P D  [] P D = [C F C F ] (9.5.18)

(see Remark 9.5.26). This implies that

[ ] P D  [] P D , [Gr(k; Rn )] =

But, using (4.5.13),

[ ] P D  [] P D , [Gr(k; Rn )] = [ ] P D , [] P D  [Gr(k; Rn )]

= [ ] P D , [].

This proves that [] = [ ] P D (or [] P D = [ ] ). 

Remark 9.5.26 In the above proof, (9.5.18) is not a consequence of Proposi-

tion 5.4.12, which would require that C F and C F are submanifolds of Gr(k; Rn ). In
this simple situation, one could use the Morse function f to isolate the intersection
point around the critical level d() and deal with an intersection of submanifolds
(with boundaries). For more general situation (see the proof of Proposition 9.5.29),
one must rely on the intersection theory for real algebraic varieties (see, e.g. [16,
(1.12) and Sect. 5]).

In addition to the above ambient inclusion : Gr(k; Rn )  Gr(k; Rn+1 ) we

also consider the fattening inclusion : Gr(k; Rn ) Gr(k + 1; Rn+1 ) sending
P to R P R Rn . Then H j ([]) = [1] for all [kn ]. This drives us to
decompose a word [nk ] into its prefix, stem and suffix, delimited by the first 0 and
the last 1 of :

= 111111111

  00101101
  .

  0000
prefix stem suffix

Given n and k, a word [nk ] (and then a class [] H (Gr(k; Rn )) or []

H (Gr(k; Rn ))) is determined by its stem. For example, 0101 is, for k = 4 and
n = 8, the stem of the unique class [11010100] H 3 (Gr(4; R8 )). The stem of 1
H 0 (Gr(k; Rn )) is just 0. Here is a first use of the prefix-stem-suffix decomposition.

Proposition 9.5.27 Let n, k and i be integers with 0 i k. Then, for k + 1

n , the Stiefel-Whitney class wi = wi (k ) is the class in H i (Gr(k; Rn )) with
stem 01i . For example, w3 (4 ) = [1011100] H 3 (Gr(4; R7 )).

Proof The proposition is true if i = 0, since w0 (k ) = 1. Let us assume that i 1.

We first prove that wi (k ) = [01i ] in H i (Gr(i; R )). Recall H (Gr(i; R ))
Z2 [w1 , . . . , wi ] (where w j = w j (i )) so
 
K = ker H : H (Gr(i; R )) H (Gr(i 1; R ))
9.5 Real Flag Manifolds 393

is the ideal generated by wi . Hence, K i is one-dimensional generated by wi . As

H ([]) = [1], one has

 [] if = 1
H ([] ) = (9.5.19)
0 otherwise.

Hence 0 = [01i ] K i . Therefore, wi (k ) = [01i ] . Proposition 9.5.27 follows

from the above particular case since H (wi (k )) = wi (k ) and H (wi (k )) =
wi (k1 ). 

Let , [nk ]. As S is a basis for H (Gr(k; Rn )) and S  is the Kronecker dual

basis for H (Gr(k; Rn )), we can write

[]  [] =
 []
[nk ]

where

 = []  [] , [] Z2 .

is a version of the Schubert calculus

Computing the structure constants 
for
(mod 2). The usual Schubert calculus deals with the structure constants C
the basis S P D , defined by


[]PD  []PD = C []PD .
[nk ]

= C
By Lemma 9.5.25 and its proof, one has 
. Again, Schubert calculus
was initiated by Ehresmann in [47, 48] and further developed in e.g. [31, 32, 66, 75].
For a more recent as well as an equivariant version, see [119]. Note that  = 0

unless d() + d() = d().

A binary word [nk ] is determined by its Schubert symbol, i.e. the k-tuple of
integers indicating the positions of the 1s in . For instance, the Schubert symbol
of 0100101 is (2, 5, 7). We use the Schubert symbol of for all the cohomology
classes [0 j ] ( and 0 having the same symbol). For the reverse correspondence,
we decorate the Schubert symbol by a flat sign . Example:

[01001010r ] = (2, 5, 7) in H 7 (Gr(3; R7+r ))


(2, 5, 7) = [01001010 ] in H7 (Gr(3; R7+r ))
r

Our notation for Schubert symbols are that of [153], close to the original one of
[47]. Other conventions are used in e.g. [32, 75].
394 9 Stiefel-Whitney Classes

Remark 9.5.28 Fix the integers k n and let a = (a1 , . . . , ak ) be a k-tuple of

integers. In order for a to be a Schubert symbol determining a class in H (Gr(k; Rn )),
it should satisfy

1 a1 < a2 < < an n. (9.5.20)

When this is not the case, we decide by convention, that a represents the class 0.

A Schubert cell CF will be also labeled by the Schubert symbol of : CF = CaF

if a = [] . For the Poincar duality (see Lemma 9.5.25), we set

a = [ ] = [] P D .

If a = (a1 , . . . , ak ) then a = (n + 1 ak , . . . , n + 1 a1 ). The definition of 
is also transposed for Schubert symbols:

ab= ab
c
c
c

where the sum runs over all Schubert symbols c and

ab
c
= a  b, c  Z2 .

The following proposition and its proof is a variant, in our language, of Reduction
Formula I of [75, p. 202].

Proposition 9.5.29 (Reduction formula) Let k 2 be an integer. Let r , s and t be

positive integers k satisfying t = r + s 1. Let a, b and c be Schubert k-symbols.
Then

0 if ct < ar + bs 1
ab =
c
ac b if ct = ar + bs 1,

b and c are the Schubert (k 1)-symbols

where a,

a = (a1 , . . . , ar 1 , ar +1 1, . . . , ak 1)
b = (b1 , . . . , bs1 , bs+1 1, . . . , bk 1)
c = (c1 , . . . , ct1 , ct+1 1, . . . , ck 1)

Example 9.5.30 Let us use the formula for s = 1 and suppose that bs = 1. Thus
b = [] = [1] with [n1
k1 ]. The condition t = r + s 1 reduces to t = r and
ct = ar + bs 1 becomes cr = ar . Writing it in terms of a  = [] and c = [] this
means that if r = r = 1 for some r , one can remove r from and r from and
replace by .
For instance,
9.5 Real Flag Manifolds 395

10101,11010
01101 = 1010,1010
0110 s = 1 and r = t = 5
= 100,010
010 s = 1 and r = t = 3
=1 since [100] = 1 in H 0 (Gr(1; R3 )).

Proof of Proposition 9.5.29 Let F, F  and F  be three complete flags in Rn . If chosen

 
generically, then C xF , C yF and C zF are pairwise transverse for any Schubert symbols
x, y and z. Therefore, if d(a) + d(b) = d(c), C a C b C c is 0-dimensional and
c = a  b, c 
ab
= a  b, (c ) P D  [Gr(k; Rn )]
(9.5.21)
= a  b  (c ) P D , [Gr(k; Rn )]
= (C a C b C c ) mod 2,

the last equality coming from the intersection theory analogous to Proposition 5.4.12
but for algebraic cycles (see Remark 9.5.26).
Let P Gr(k; Rn ). If P C a C b C c then

dim(P Fn+1ar ) k + 1 r

dim(P Fn+1b ) k+1s
s
dim(P Fct ) t.

Therefore, the condition t = r + s 1 implies that


dim(P Fn+1ar Fn+1b s
Fct ) 1.

On the other hand, as F, F  and F  are transverse flags,


dim(Fn+1ar Fn+1b s
Fct ) = t r s + 2.

Thus, abc = 0 if c < a + b 1. If c = a + b 1, then F  

t r s t r s n+1ar Fn+1bs Fct
is a line L, which must be contained in any P C a C b C c . Let L be the

orthogonal complement of L and let : Rn L be the orthogonal projection.

For 1 i n 1, define

(Fi ) if i n ar
Fi =
(Fi+1 ) if i n + 1 ar .

As L Fn+1ar but L  Fnar , the sequence of vector spaces Fi constitutes a

complete flag F in for L . Define F  accordingly and F  by

 (Fi ) if i ct
Fi =  ) if i c + 1.
(Fi+1 t
396 9 Stiefel-Whitney Classes

F  and F  are transverse flags and, by linear algebra, one checks that
Then, F,

P = (P) L C a C b C c (P) C (a) C c .

C (b)

Hence,
   
(C aF C bF C cF ) = (C (Fa) F C cF )
C (b)

c =  c .
which, using (9.5.21), proves that ab a b

Corollary 9.5.31 Let a, b and c be Schubert k-symbols. Then ab c = 0 unless

ci max{ai + b1 1, bi + a1 1} for all 1 i k. In particular, ab

c = 0 unless

ci max{ai , bi } for all 1 i k.

Proof If cr < ar + b1 1 for some integer r , then ab

c = 0 by the reduction formula

for s = 1. As ab = ba , this proves the corollary.

c c 

We now compute, for a Schubert symbol a, the expression of wi  a in the basis

S  . For J {1, 2, . . . , k}, we define a map a a J from Nk to itself by

ai + 1 if i J
aiJ =
ai if i
/ J.

Proposition 9.5.32 Let a be a Schubert k-symbol. The equation


wi  a = aJ (9.5.22)
J {1,2,...,k}
J =i

holds in H (Gr(k; Rn )) (with the convention of Remark 9.5.28).

Since in the right side of (9.5.22), we use the convention of Remark 9.5.28,
Proposition 9.5.32 holds true for any n and any i (wi = 0 if i k).
Example 9.5.33

(2, 3, 5, 6) + (2, 3, 4, 7) + (1, 3, 5, 7) in H 6 (Gr(4; Rn )) for n 7.
w2  (1, 3, 4, 6) =
(2, 3, 5, 6) in H 6 (Gr(4; R6 )).

Proof It suffices to prove the proposition for n = . We identify wi with it Schubert

symbol which, by Proposition 9.5.27, is

wi = (1, 2, . . . , k i, k i + 1, . . . , k + 1).

Then, the notation a,w

c
i
is meaningful. Let c be a Schubert k-symbol such that
a,wi = 0. Then d(c) = d(a) + i and, as c j a j by Corollary 9.5.31, there is
c
9.5 Real Flag Manifolds 397

K {1, 2, . . . , k} with K = k i such that c j = a j for j K . By iterating the

reduction formula for s = 1 with the indices in K , we get that a,wc c
= a,w
i i , where
a and c are Schubert i-symbols and wi = (2, 3, . . . , i + 1). By Corollary 9.5.31,
we have c j a j + 1 for all 1 j i. This implies that c = a J for J =
{1, 2, . . . , k} K .
Conversely, let J {1, 2, . . . , k} with |J | = i. We have to prove that (J ) =
a,wi = 1 if a J is a Schubert symbol for Gr(k; R ). By repeating the reduction
a J

formula for s = 1 with all the indices not in J , we get that

a J
(J ) = a,w
i

where a is a Schubert i-symbol, J = {1, 2, . . . , i} and wi = (2, 3, . . . , i + 1). By

iterating again the reduction formula for s = 1 with the indices i, i 1, etc., till 2,
we get that

(J ) = ((aa11),(2)
+1)
.

This coefficient is equal to 1, as (u)  w1 = (u + 1) in H (Gr(1; R ))

H (RP ). 

For [nk ], let [nnk ] be obtained from by exchanging 0s and 1s and

reverse the order: 100101 = 010110; in formula:

j = 0 n+1 j = 1. (9.5.23)

Note that d( ) = d(). This formal operation is related to the homeomorphism

h : Gr(k; Rn ) Gr(n k; Rn ) sending k-plane P to its orthogonal complement
P .

Lemma 9.5.34 H h([] ) = [ ] .

Proof Let F = (F1 Fn ) be a complete flag in Rn and let F be the dual

flag, defined by Fi = Fni
(we add the convention that F = {0} = F ). To
0 0
(F) (F )
establish Lemma 9.5.34, we shall prove that h(C ) = C .
Let P Gr(k; Rn ). Write P = (P Fi ) Q i . Then

P Fni = {v Fni | v, Q i  = 0}.

Hence

codim F (P Fni ) = dim Q i = codim P (P Fi ) . (9.5.24)
ni

(F) (F )
Suppose that P C for [nk ]. Then, P C [nnk ]. We must prove

that = , that is to say (i = 0 n+1i = 1). But, using (9.5.24)
398 9 Stiefel-Whitney Classes

i = 0 dim(P Fi ) = dim(P Fi1 )

codim F (P Fni ) = codim F (P Fn+1i )
ni n+1i

dim(P Fn+1i ) = dim(P Fni )+1
n+1i = 1. 

Let w = w(k ) = 1 + w 1 + w nk be the total Stiefel-Whitney class of the

tautological (n k)-vector bundle over Gr(k; Rn ) (see Example 9.5.18).

Proposition 9.5.35 Suppose that n i + k. Then w i H i (Gr(k; Rn ) is the class

of stem 0i 1. Its Schubert symbol is (1, 2, . . . , k 1, k + i).

Example: w 5 = [111000001] = (1, 2, 3, 9) in H 5 (Gr(4; Rn ) for n 9.

Proof The homeomorphism h : Gr(k; Rn ) Gr(n k; Rn ) is covered by the

tautological bundle map k nk . Hence, h nk = k and thus w i = H
h(wi (nk )). Therefore, w i = (wi ) by Lemma 9.5.34. As stem ( ) = stem () ,
Proposition 9.5.35 follows from Proposition 9.5.27. 

We now give the expression of w i  a for a Schubert k-symbol a. As w 1 = w1 ,

we can use Formula (9.5.22) for i = 1:

w1  a = w 1  a = b (9.5.25)
b

where the sum runs over all the Schubert k-symbols b such that


k
a j b j a j + 1 and (b j a j ) = 1.
j=1

Example:

w12 = w1  (1, 2, . . . , k 1, k + 1)
= (1, 2, . . . , k 2, k, k + 1) + (1, 2, . . . , k 1, k + 2)
= w2 + w 2 .

Formula (9.5.25) admits the following generalization, called the Pieri formula, which
is a sort of a dual of Proposition 9.5.32.

Proposition 9.5.36 (Pieris formula) Let a be a Schubert k-symbol. The equation


w i  a = b (9.5.26)
b
9.5 Real Flag Manifolds 399

holds in H (Gr(k; Rn )), where the sum runs over all the Schubert k-symbols b such
that


k
a j b j < a j+1 and (b j a j ) = i (9.5.27)
j=1

(with the convention of Remark 9.5.28).

Proof For Schubert k-symbols a, b, the proposition says that

w i  a, b  = 1 (a, b) satisfies (9.5.27). (9.5.28)

(Note that the implication follows from Corollary 9.5.31 and from w i being of
degree i). Rewriting (9.5.28) with , [nk ] gives that w i  [] , [] = 1 if and
only if and satisfy the following pair of conditions

(i) = A1 10r1 A2 10r2 As 10rs As+1 , with sj=1 r j = i, and
(ii) = A1 0r1 1A2 0r2 1 Ars 0rs 1As+1 .
(Intuitively: a certain quantity of 1s are shifted by one position to the right of total
amount shifting being i). The pair of conditions (i) and (ii) is equivalent to the
following ones
s
(i) = A rs
s+1 1 0 As 1
rs1 0 A 1r1 0 A with
2 1 j=1 r j = i, and

(ii) = As+1 01 As 01r s r s1
A2 01 .
r 1

Recall from Lemma 9.5.34 and the proof of Proposition 9.5.35 that the homeo-
morphism h : Gr(k; Rn ) Gr(n k; Rn ) satisfies H ([]) = [ ] and w i =
H (wi (nk )). Therefore

w i  [] , [] = 1 wi (nk )  [ ] , [ ] = 1. (9.5.29)

By Proposition 9.5.32, the right hand equality in (9.5.29) is equivalent to the pair of
conditions (i) and (ii) , which proves Proposition 9.5.36. 
We finish this subsection by mentioning the Giambellis formula, which express
a cohomology class given by a Schubert symbol as a polynomial in the w i s. The
Giambelli and the generalized Pieri formulae together provides a procedure for com-
puting the structure constants ab
c .

Proposition 9.5.37 (Giambellis formula)

 
(a1 , . . . , ak ) = det w ai j 1i, jk .

with the convention that w u = 0 if u < 0.

For wr = (1, 2, . . . , k r, k r + 1, . . . , k + 1), Proposition 9.5.37 reproves the
second formula of Lemma 9.5.19.
400 9 Stiefel-Whitney Classes

Proof By induction on k, starting trivially if k = 1. The lengthy induction step, using

the Pieri formula, may be translated in our language from [75, pp. 204205] (see also
[32, p. 366]). 

9.6 Splitting Principles

Let :  O(n) denotes the inclusion of the diagonal subgroup of O(n)

 O(1) O(1) O(n).

This induces an inclusion B : B B O(n) between the classifying spaces. The

symmetric group Symn acts on O(1) O(1) by permuting the factors, and
then on B. As in Sect. 9.5, n denotes the tautological vector bundle on B O(n) 
Gr(n; R ). It is the vector bundle associated to the universal O(n)-bundle E O(n)
B O(n).

Theorem 9.6.1 The GrA-morphism H B : H (B O(n)) H (B) is injective

and its image is H (B)Symn . The induced vector bundle B n splits into a Whitney
sum of line bundles.

Proof We have seen in (9.5.13) that the homotopy equivalence


B  E O(n) O(1) O(1)

makes B homotopy equivalent to the locally trivial bundle

B
F1(1, . . . , 1)
B B O(n). (9.6.1)

We have also established in the proof of Theorem 9.5.14 that H is surjective.

Hence, by Corollary 4.7.19, H B is injective. Also, using (7.4.3) and that O(1)
{1}. one has a homotopy equivalence

B / B O(1) B O(1)  RP RP


and thus a GrA-isomorphism

: Z2 [x1 , , xn ]
H (B)

where xi has degree 1. By Theorem 9.5.8,

H (B O(n)) Z2 [w1 (n ), . . . , wn (n )].

9.6 Splitting Principles 401

Note that B is covered by a morphism of principal bundles

E O(1) E O(1)
E / E O(n)

 
B O(1) B O(1)
B / B O(n).

One has a similar diagram for the associated vector bundles = (E O(1) O(1) R

B O(1)) (corresponding to the tautological line bundle over

n RP ) and n . This
implies that B n . As w( ) = i=1 (1 + xi ), one has

H B(wi (n )) = wi ( ) = i ,

where i is i-th elementary symmetric polynomial in the variables x j . The second

assertion of Theorem 9.6.1 follows, since the elementary symmetric polynomials
GrA-generate Z2 [x1 , , xn ]Symn H (B)Symn .
Finally, the homotopy equivalence is of the form = (1 , . . . , n ), with
i : B B O(1). In other words, coincides with the composition

1 n
B
B B B O(1) B O(1),

where is the diagonal map. Hence,

B n ( ) = (1 n ) = 1 n ,

which shows that B n is isomorphic to a Whitney sum of line bundles. 

Theorem 9.6.1 may be generalized as follows. Consider the inclusion homomor-
phism

n 1 ,...,nr : O(n 1 ) O(nr ) O(n)

sending (A1 , . . . , Ar ) to the diagonal-block matrix with blocks A1 , . . . , Ar . Using

the homotopy equivalence B O(n 1 ) B O(nr )  B(O(n 1 ) O(nr ))
(see (7.4.3)), the homomorphism n 1 ,...,nr induces a continuous map

Bn 1 ,...,nr : B O(n 1 ) B O(nr ) B O(n).

Theorem 9.6.2 The map Bn 1 ,...,nr satisfies the following properties.

(1) The GrA-morphism

H Bn 1 ,...,nr : H (B O(n)) H (B O(n 1 ) B O(nr ))

is injective.
402 9 Stiefel-Whitney Classes

(2) H Bn 1 ,...,nr (wi ) = wi (n 1 nr ) for each i 0. In particular, the image

of H Bn 1 ,...,nr is generated by wi (n 1 nr ) (i 0).
(3) The induced vector bundle Bn1 ,...,nr n splits into a Whitney sum of vector
bundles of ranks n 1 , . . . , nr .
Proof Using the inclusion factorization

/ O(1)n 1 O(1)nr / O(n 1 ) O(nr ) / O(n),

O(1)n

where = n 1 ,...,nr , the injectivity of H B comes from that of H B1,...,1 , estab-

lished in Theorem 9.6.1. As B is covered by a morphism of principal bundles

/ E O(n)
E O(n 1 ) E O(nr )
,
 
B / B O(n)
B O(n 1 ) B O(nr )

one deduces (2) and (3) as in the proof of Theorem 9.6.1. 

Proposition 9.6.3 Let be a vector bundle over a paracompact space X . Then,
there is a map f : X X such that
(1) H f is injective.
(2) f splits into a Whitney sum of line bundles.
Proposition 9.6.3 is called the splitting principle. For = n over B O(n),
Theorem 9.6.1 says that on can take B O(n)n and f = B.
Proof As X is paracompact, admits an Euclidean structure and there is a classifying
map : X B O(n) for , i.e. n . Consider the pull-back diagram

f
X / X

,
 
B
B / B O(n)

where B is defined as in (9.6.1). As B is a locally trivial bundle with fiber

Fl(n 1 , . . . , nr ), so is f (this is the Fl(n 1 , . . . , nr )-bundle associated to Fra ).
We saw in the proof of Theorem 9.6.1 that H (B) H (Fl(n 1 , . . . , nr )) is sur-
jective. Then, so is H (X ) H (Fl(n 1 , . . . , nr )). Hence, by Corollary 4.7.19,
H f is injective. Now,

f = f n = B n .

As, by Theorem 9.6.1, B n is a Whitney sum of line bundles, so does f . 

9.6 Splitting Principles 403

One consequence of the splitting principle is the uniqueness of Stiefel-Whitney

classes (compare [153, Theorem 7.3] or [105, Chap. 16, Sect. 5]).

Proposition 9.6.4 Suppose that w is a correspondence associating, to each vector

) H (X ), such that
bundle over a paracompact space X , a class w(
(1) if f : Y X be a continuous map, then w( f ) = H f (w(
)).
) = w(
(2) w(  )  w( ).

(3) if is the tautological line bundle over RP , then w(

) = 1 + a, where
0 = a H 1 (RP ).
Then w = w, the total Stiefel-Whitney class.

Proof Condition (2) implies that

([Link]) w(
1 n ) = w(
1 )   w(
n ).
As in the proof of Theorem 9.6.1, Conditions (1), ([Link]) and (3) imply that the map
B : B B O(n) satisfies

H B(w(
n )) = (1 + xi )n H (B) Z2 [x1 , , xn ].

Thus, still by the proof of Theorem 9.6.1, H B(w(

n )) = H B(w(n )). As H B
n ) = w(n ). The bundle n being universal (see
is injective, this implies that w(
9.1.14), Condition (1) implies that w( ) = w( ) for any vector bundle over a
paracompact space X . 

Another consequence of the splitting principle is the action of the Steenrod

algebra on Stiefel-Whitney classes. The following proposition was proved by
Wu wen-Tsn [214].

Proposition 9.6.5 Let be a vector bundle over a paracompact space X . Then

  ji+k1
Sqi w j ( ) = k wik ( ) w j+k ( ). (9.6.2)
0ki

Example 9.6.6 Setting wi = wi ( ), we get

Sq1 w j = w1 w j + ( j 1)w j+1

 j1
Sq2 w j = w2 w j + ( j 2)w1 w j+1 + 2 w j+2
 j2  j3
Sq w j = w3 w j + ( j 3)w2 w j+1 +
3
2 w1 w j+2 + 3 w j+3 .

Proof of Proposition 9.6.5 By naturality of w and Sq, it suffices to prove (9.6.2) for
= n , the tautological vector bundle on B O(n). By Theorem 9.6.1 and its proof,

H B(w j (n )) = j Z2 [x1 , , xn ]
404 9 Stiefel-Whitney Classes

where x j corresponds to the non-trivial element in H 1 (RP ), and j is the

j-th elementary symmetric polynomial in the variables xr . As H B is injec-
tive by Theorem 9.6.1, Formula (9.6.2) reduces to the computation of Sqi j in
Z2 [x1 , , xn ]Symn , using that Sq(xr ) = xr + xr2 . This technical computation may
be found in full details in [15, Theorem 7.1]. The reader may, as an exercise, prove
the special cases of Example 9.6.6.
The splitting principle also gives the following result about the Stiefel-Whitney
classes of a tensor product (for a more general formula, see [153, p. 87]).

Lemma 9.6.7 Let and be vector bundles over a paracompact space X . Suppose
that is of rank r and that is a line bundle. Then


r
w( ) = (1 + w1 ())k wr k ( ). (9.6.3)
k=0

Proof Set u = w1 (). Suppose first that splits into a Whitney sum of r line bundles
j , of Stiefel-Whitney class 1 + v j . Then, letting k = (v1 , . . . , vr ) denote the k th
elementary symmetric polynomial, one has

w( ) = w(rj=1 ( j ))

= rj=1 w( j ) by (9.4.3)
r
= j=1 (1 + u + v j ) by Proposition 9.2.4
r
= k=0 (1 + u) r k (v1 , . . . , vr ).
k

Since r k (v1 , . . . , vr ) = wr k ( ), we have shown (9.6.3) when splits into a

Whitney sum of r line bundles. If this is not the case, Formula (9.6.3) still holds true
by the splitting principle of Proposition 9.6.3. 

9.7 Complex Flag Manifolds

The plan of this section follows that of Sect. 9.5. We shall indicate the slight changes
to get from the real flag manifolds to the complex ones, without repeating all the
proofs.
Let n 1 , . . . nr be positive integers and let n = n 1 + n 2 + nr . By the complex
flag manifold FlC (n 1 , . . . , nr ), we mean any smooth manifold diffeomorphic to the
homogeneous space

FlC (n 1 , . . . , nr ) U (n) U (n 1 ) U (n 2 ) U (nr ). (9.7.1)

The most usual concrete occurrence of complex flag manifolds are as below.
9.7 Complex Flag Manifolds 405

(1) Nested subspaces. FlC (n 1 , . . . , nr ) is the

set of nested complex vector subspaces
V1 Vr Cn with dimC Vi = ij=1 n j .
(2) Mutually orthogonal subspaces. FlC (n 1 , . . . , nr ) is the set of r -tuples
(W1 , . . . , Wr ) of complex vector subspaces Cn which are mutually orthogonal
(for the standard Hermitian product on Cn ) and satisfy dim Wi = n i . The cor-
respondence from this definition to Definition (1) associates to (W1 , . . . , Wr )
the nested family {Vi } where Vi is the complex vector space generated by
W1 Wi .
(3) Isospectral Hermitian matrices. Let 1 > > r be real numbers. Consider
the manifold H M(n) of all Hermitian (n n)-matrices, on which U (n) acts
by conjugation. Then FlC (n 1 , . . . , nr ) occurs as the orbit of the diagonal matrix
having entries i with multiplicity n i .
  
FlC (n 1 , . . . , nr ) = R dia 1 , . . . , 1 , , r , . . . , r R 1 | R U (n) .

 
 
n1 nr
(9.7.2)

In other words, FlC (n 1 , . . . , nr ) is here 

the space of Hermitian (n n)-matrices
with characteristic polynomial equal to ri=1 (x i )n i . Indeed, two matrices in
H M(n) are in the same U (n)-orbit if and only if they have the same characteristic
polynomial. The correspondence from this definition to Definition (2) associates,
to a matrix M, its eigenspaces for the various eigenvalues.
Concrete definition (3) is our working definition for FlC (n 1 , . . . , nr ) throughout this
section. Special classes of flag manifolds are given by the complex Grassmannians

Gr(k; Cn ) = FlC (k, n k) U (n) U (k) U (n k)

of complex k-planes in Cn . This is a closed manifold of dimension

dim Gr(k; Cn ) = dim U (n) dim U (k) dim U (n k) = 2k(n k).

For example, Gr(1; Cn ) CP n1 , of dimension 2(n 1).

Using Definition (3) above, our concrete Grassmannian will be
  
Gr(k; Cn ) = R dia 1, . . . , 1, 0, . . . , 0 R 1 | R U (n) . (9.7.3)

 
 
k nk

As, in the real case, we define the complete complex flag manifold

FlC (1, . . . , 1) U (n) U (1) U (1)

with dim FlC (1, . . . , 1) = dim U (n) n = n 2 n = n(n 1).

As in Sect. 9.5, we define real functions on the flag manifolds by restriction of the
weighted trace on f : H M(n) R defined by
406 9 Stiefel-Whitney Classes


n
f (M) = j Mjj
j=1

where Mi j denotes the (i, j)-entry of M.

Proposition 9.7.1 Let FlC (n 1 , . . . , nr ) H M(n) be the complex flag manifold as
presented in (9.5.2). Then, the restriction f : FlC (n 1 , . . . , nr ) R of the weighted
trace is a perfect Morse function whose critical points are the diagonal matrices in
FlC (n 1 , . . . , nr ). The index of the critical point dia(x1 , . . . , xn ) is twice the number
of pairs (i, j) with i < j and xi < x j . In consequence, dim FlC (n 1 , . . . , nr ) =
2 dim Fl(n 1 , . . . , nr ) and

Pt (FlC (n 1 , . . . , nr )) = Pt 2 (Fl(n 1 , . . . , nr )). (9.7.4)

Recall that dim Fl(n 1 , . . . , nr ) was computed in Corollary 9.5.5 and that the
Poincar polynomial Pt (Fl(n 1 , . . . , nr )) was described in Corollary 9.5.15. Equal-
ity (9.7.4) implies the following corollary.
Corollary 9.7.2 The cohomology groups of FlC (n 1 , . . . , nr ) vanish in odd
degrees.
Remark 9.7.3 The manifold FlC (n 1 , . . . , nr ) H M(n) admits an U (n)-invariant
symplectic form, induced from the non-degenerate symmetric form (X, Y )
trace (X Y ) on H M(n) (see [12, Chap. II, Example 1.4]). The weighted trace is
the moment map for the Hamiltonian circle action given by the conjugation by
dia(eit , e2it , . . . , enit ). The involution given on FlC (n 1 , . . . , nr ) by the complex
conjugation is anti-symplectic and anti-commutes with the circle action. Its fixed
point set is Fl(n 1 , . . . , nr ). Note that f is -invariant and the critical point of f or
f |Fl(n 1 , . . . , nr ) are the same. This, together with (9.7.4), is a particular case of a
theorem of Duistermaat [45] (see also Remark 9.7.9).
Proof of Proposition 9.7.1 We use the injective homomorphism r i j : SU (2) U (n),
introduced in the proof of Proposition 9.7.1, whose image contains the matrices
   
ij cos t sin t ij i j cos t 1 sin t
Rt =r ij
and Rt = r (t R).
sin t cos t 1 sin t cos t

Suppose that FlC (n 1 , . . . , nr ) is a diagonal matrix. Then, a basis of

T FlC (n 1 , . . . , nr ) is represented by the curves

ij ij i j (t) = R ti j R t
i j (t) = Rt Rt and
ij
.

As in the proof of Proposition 9.5.2, this shows that the critical points of f are exactly
the diagonal matrices and computes the indices.
As the critical points are all of even index, the function f is a perfect Morse
function by Lemma 7.6.2. One can also proceed as in the proof of Proposition 9.5.2,
9.7 Complex Flag Manifolds 407

using that f is invariant for the action of the diagonal subgroup T of U (n), which is
the torus U (1) U (1), and use Theorem 7.6.11. 
As in Sect. 9.5, consider the inclusion H M(n) H M(n+1) with image the matri-
ces with vanishing last row and column. Seeing Gr(k; Cn ) H M(n) as in (9.5.3),
this gives an inclusion Gr(k; Cn ) Gr(k; Cn+1 ). The proof of the following lemma
is the same as that of Lemma 9.5.7.

Lemma 9.7.4 The homomorphism H j (Gr(k; Cn+1 )) H j (Gr(k; Cn )) induced

by the inclusion is surjective for all j and is an isomorphism for j 2(n k).

Tautological bundles. Consider a complex flag manifold FlC (n 1 , . . . , nr ), with n =

n 1 + + nr and the following closed subgroups of U (n).

Bi = U (n 1 ) {1} U (nr ) U (n 1 ) U (nr ) U (n).

Then

Pi = U (n)/Bi
U (n) U (n 1 ) U (nr ) = FlC (n 1 , . . . , nr )

is an U (n i )-principal bundle (see p. 381) over FlC (n 1 , . . . , nr ). Its associated com-

plex vector bundle of rank n i , i.e. E(i ) = Pi U (n i ) Cn i , is called i-th-tautological
vector bundle over FlC (n 1 , . . . , nr ). Being associated to an U (n i )-principal bun-
dle, i is endowed with an Hermitian structure and its space of orthonormal frames
Fra (i ) is equal to Pi . In the mutually orthogonal subspaces presentation (2) of
FlC (n 1 , . . . , nr ), we see that

E(i ) = {(W1 . . . , Wr , v) FlC (n 1 , . . . , nr ) Cn | v Wi }.

Note that 1 r is trivial (see Sect. 9.5, p. 381).

The complex vector bundle 1 over FlC (k, n k) = Gr(k; Cn ) is called the tauto-
logical vector bundle over the complex Grassmannian Gr(k; Cn ); it is of (complex)
rank k and is denoted by or k . The space of Fra (k ) is the complex Stiefel manifold
Stief(k, Cn ) of orthonormal k-frames in Cn .
The inclusion Cn Cn {0}  Cn+1 induces an inclusion Gr(k; Cn ) 
Gr(k; Cn+1 ) and we may consider the inductive limit

Gr(k; C ) = lim Gr(k; Cn )

n

which is a CW-space. The tautological vector bundle k is also defined over

Gr(k; C ) and induces that over Gr(k; Cn ) by the inclusion Gr(k; Cn ) 
Gr(k; C ). It is classical that i (Stief(k, Cn )) = 0 for i < 2(n k) + 1
(see [181, 25.7]), thus Stief(k, C ) = Fra(k ) is contractible. Hence, the U (k)-
principal bundle Stief(k, C ) Gr(k; C ) is a universal U (k)-principal bundle
(see [181, Sect. 19.4]) and thus homotopy equivalent to the Milnor universal bundle
EU (k) BU (k). In particular, Gr(k; C ) has the homotopy type of BU (k). As a
408 9 Stiefel-Whitney Classes

consequence, any complex vector bundle of rank k over a paracompact space X is

induced from k by a map X Gr(k; C ).
To emphasize the analogy with Sect. 9.5, we introduce the total Chern classes
c( ) H 2 (X ) of a complex vector bundle of rank k over a space X by


k
c( ) = w2 j (R ),
j=1

where R is the vector bundle seen as a real vector bundle of (real) rank 2k. The
component of c( ) in H 2 j (X ) is

c j ( ) = w2 j (R ) H 2 j (X ) (9.7.5)

is called the i-th Chern class of .

Theorem 9.7.5 The cohomology ring of BU (k) is GrA-isomorphic to the polyno-

mial ring

H (BU (k)) = H (Gr(k; C )) Z2 [c1 , . . . , ck ]

generated by the Chern classes ci = ci (k ) of the tautological bundle k .

Remark 9.7.6 Our Chern classes c j ( ) are the reduction mod 2 of the integral Chern
classes (see [153, Sect. 14] or [105, Chap. 16]). That the restriction mod 2 of c j ( )
coincides with w2 j (R ) (whence our definition (9.7.5)) is proven in [181, Theo-
rem 41.8]. Note that, by Theorem 9.7.5, w2 j+1 (R ) = 0.

Proof of Theorem 9.7.5 It is the same as that of Theorem 9.5.8, using Proposition 9.7.1.
To see that the Chern classes are algebraically independent, we use the tautological
complex line bundle over CP and its n-times product n over (CP )n .
Theorem 9.7.5 together with Lemma 9.5.7 gives the following corollary.

Corollary 9.7.7 The cohomology ring H (Gr(k; Cn )) is generated, as a ring, by

the Chern classes c1 (k ), . . . , ck (k ) of the tautological bundle k .

Let c( j ) = 1 + c1 ( j ) + + cn j ( j ) H (FlC (n 1 , . . . , nr )) be the Chern class

of the tautological vector bundle j . The following theorem is proven in the same
way as Theorem 9.5.14. Actually, replacing Z2 [ci ( j )] by Z[ci ( j )], the statement is
true for the integral cohomology (as we wrote a minus sign in the last expression).

Theorem 9.7.8 The cohomology algebra H (FlC (n 1 , . . . , nr )) is GrA-isomorphic

to the quotient of the polynomial ring

Z2 [ci ( j )] , 1 i r j , j = 1, . . . , r

by the ideal generated by the homogeneous components of 1 c(1 ) c(r ).

9.7 Complex Flag Manifolds 409

Remark 9.7.9 By Theorems 9.7.8 and 9.5.14, the correspondence ci ( j ) wi

( j ) provides an abstract ring isomorphism

H (Fl(n 1 , . . . , nr )).
H 2 (FlC (n 1 , . . . , nr ))

Actually, FlC (n 1 , . . . , nr ) with the complex conjugation is a conjugation space

(see Sect. 10.2). Given Remark 9.7.3, this is established in [87, Theorem 8.3].

We have seen in Proposition 9.2.4 that the first Stiefel-Whitney class classifies
the real lines bundles. The full analogue for complex line bundles requires coho-
mology with Z-coefficients: the first integral Chern class provides an isomorphism

(LC (X ), ) H 2 (X ; Z), where LC (X ) be the set of isomorphism classes of com-
plex lines bundles over a CW-complex X (see [96, pp. 6263]). But, staying within
the mod 2 cohomology, one can prove the following result.

Proposition 9.7.10 Let and  be two complex line bundles over a CW-complex
X . Then c1 (  ) = c1 ( ) + c1 (  ).

Proof The argument follows the end of the proof of Proposition 9.2.4. One has to
replace R by C and K by K C = C C . The only thing to prove is that the
composed map

P / B KC B
/ BC
BC BC 

corresponding to that of Diagram (9.2.2) satisfies

H (B P  )(v) = v 1 + 1 v, (9.7.6)

where v is the generator of H 2 (BC ) = H 2 (BU (1)) = H 2 (CP ). The complex

conjugation of C induces an involution on BC corresponding to the conjugation
on CP , with fixed point RP = BR . The map B P  is -equivariant. Hence,
Eq. (9.7.6) follows from (9.2.3), using that the inclusion j : RP CP satisfies
H j (v) = u 2 (see Proposition 6.1.11). 

Finally, the splitting principle results of Sect. 9.6 have their correspondents for
complex bundles. One uses the inclusion of the diagonal subgroup of U (n)

 U (1) U (1) U (n).

The following result is proven in the same way as for Theorem 9.6.1.

Theorem 9.7.11 The GrA-morphism H B : H (BU (n)) H (B) is injec-

tive and its image is H (B)Symn . The complex vector bundle B n induced from
the universal bundle n splits into a Whitney sum of complex line bundles.
410 9 Stiefel-Whitney Classes

As for Theorems 9.6.2 and 9.7.11 generalizes in the following way for the
inclusion

n 1 ,...,nr : U (n 1 ) U (nr ) U (n).

Theorem 9.7.12 The map Bn 1 ,...,nr satisfies the following properties.

(1) The GrA-morphism

H Bn 1 ,...,nr : H (BU (n)) H (BU (n 1 ) BU (nr ))

is injective.
(2) H Bn 1 ,...,nr (ci ) = ci (n 1 nr ) for each i 0. In particular, the image
of H Bn 1 ,...,nr is generated by ci (n 1 nr ) (i 0).
(3) The induced complex vector bundle Bn1 ,...,nr n splits into a Whitney sum of
complex vector bundles of ranks n 1 , . . . , nr .

As in Sect. 9.6, we deduce from Theorem 9.7.11 the following proposition (split-
ting principle for complex bundles).

Proposition 9.7.13 Let be a complex vector bundle over a paracompact space X .

Then, there is a map f : X X such that
(1) H f is injective.
(2) f splits into a Whitney sum of complex line bundles.

As in Proposition 9.6.4, we get an axiomatic characterization of Chern classes.

Proposition 9.7.14 Suppose that c is a correspondence associating, to each complex

vector bundle over a paracompact space X , a class c(
) H 2 (X ), such that
(1) if f : Y X be a continuous map, then c( f ) = H f (c(
)).
(2) c( 
) = c( )  c(
).

(3) if is the tautological complex line bundle over CP , then c(

) = 1+a, where
0 = a H 2 (CP ).
Then c = c, the total Chern class.

Thanks to our definition of Chern classes via Stiefel-Whitney classes, the follow-
ing proposition is a direct consequence of Proposition 9.6.5.

Proposition 9.7.15 Let be a complex vector bundle over a paracompact space X .

Then
  ji+k1
Sq2i c j ( ) = k cik ( ) c j+k ( ). (9.7.7)
0ki
9.7 Complex Flag Manifolds 411

Remark 9.7.16 As in Sect. 9.5.3, the Schubert calculus may be developed for
complex Grassmannians. The degrees of (co)homology classes are doubled. The
Stiefel-Whitney classes wi are replaced by the Chern classes ci . The Stiefel-Whitney
classes w i corresponds, in the literature, to the Segre classes.

We finish this section with the complex analogue of Lemma 9.6.7.

Lemma 9.7.17 Let and be complex vector bundles over a paracompact space
X . Suppose that is of rank r and that is a line bundle. Then


r
c( ) = (1 + c1 ())k cr k ( ). (9.7.8)
k=0

Proof The proof is the same as that of Lemma 9.6.7. The use of Proposition 9.2.4
has to be replaced by that of Proposition 9.7.10. 

9.8 The Wu Formula

9.8.1 Wus Classes and Formula

Let Q be a closed manifold of dimension n. The map

Sqk ,[Q]
H nk (Q) H n (Q) Z2

is a linear form on H nk (Q). By Poincar duality (see Theorem 5.3.12), there is a

unique class vk (Q) H k (Q) such that Sqk (a), [Q] = vk (Q)  a, [Q] for all
a H nk (Q). In other words,

Sqk (a) = vk (Q)  a (9.8.1)

for all a H nk (Q). The left hand side of (9.8.1) vanishing if k > n k, one has
vk (Q) = 0 if k > n/2. The class vi (Q) is the i-th Wu class of Q (for Wu classes in
a more general setting, see [123, Sect. 3]). Note that v0 (Q) = 1. The total Wu class
v(Q) is defined by

v(Q) = 1 + v1 (Q) + + v[n/2] (Q) H (Q).

As an example the next lemma shows the role of vk (Q) when n = 2k. Let V be
a Z2 -vector space and let B : V V V be a bilinear form. A symplectic basis of
V for B is a basis {a1 , . . . , ak , b1 . . . , bk } of V such that B(ai , a j ) = B(bi , b j ) = 0
and B(ai , b j ) = B(b j , a j ) = i j . By convention, the empty basis for V = {0} is
also symplectic.
412 9 Stiefel-Whitney Classes

Lemma 9.8.1 Let Q be a closed smooth manifold of dimension 2k such that its Wu
class vk (Q) vanishes. Then the bilinear form B : H k (Q) H k (Q) Z2 given by
B(x, y) x  y, [Q] admits a symplectic basis.
Note that the lemma implies that H k (Q) has even dimension and thus, by
Poincar duality, (Q) is even. This can be also deduced from Corollary 5.4.16 and
Theorem 9.4.1 since, by the Wu formula (see below), w2k (T Q) = Sqk (vk (Q)) = 0.
Proof By definition of the Wu class, vk (Q) = 0 is equivalent to B being alternate,
i.e. B(x, x) = 0 for all x H k (Q). By Theorem 5.3.12, B is non-degenerate. We are
thus reduced to prove the following classical claim: on a Z2 -vector space V of finite
dimension, a non-degenerate bilinear form B which is alternate admits a symplectic
basis. As B is non-degenerate, there exists a1 , b1 V such that B(a1 , b1 ) = 1.
Hence B(b1 , a1 ) = 1 since alternate implies symmetric in characteristic 2. One has
an exact sequence

0 AV
Z2 Z2 0,

where is the linear map (v) = (B(a1 , v), B(v, b1 )) and A = ker . As B is
non-degenerate, so is its restriction to A A. This permits us to prove the claim by
induction on the dim V . 
The Wus formula below relates the Wu class of Q to the Stiefel-Whitney class
w(T Q) of the tangent bundle T Q of Q (often called the Stiefel-Whitney class of Q).
Theorem 9.8.2 (Wus formula) For any smooth closed manifold Q, one has

w(T Q) = Sq(v(Q)).

The Wu formula was proved by Wu wen-Tsn in 1950 [213] by direct compu-

tations in H (Q Q). We follow below the proof of Milnor-Stasheff [153, The-
orem 11.14] (for a proof using equivariant cohomology, see Remark 9.8.21). The
computations are lightened by the use of the slant product

/
H (X Y ) H (Y )
H (X ) (9.8.2)

/
(actually: H k (X Y ) Hm (Y ) H km (X )) which is defined as follows. Consider

the map H (X ) H (Y ) H (Y ) H (X ) defined by the correspondence a b
b, a, using the Kronecker pairing  , . As H (X Y ) H (X ) H (Y )
by the Knneth theorem (we assume that Y is of finite cohomology type), this gives
the linear map (9.8.2). The slant product is characterized by the equation

(a b)/ = b,  a

for all a H (X ), b H (Y ) and H (Y ). It is also a morphism of H (X )-

modules, i.e.
9.8 The Wu Formula 413

[(u 1)  c]/ = u  (c/) (9.8.3)

for all u H (X ), c H (X Y ) and H (Y ).

Proof of Wus formula Consider Q as the diagonal submanifold of M = Q Q,

with normal bundle = (Q, M). By Lemma 5.4.17, T Q is isomorphic to .
A Riemannian metric provides a smooth bundle pair (D(), S()) with fiber
(Dr , Sr 1 ) and there is a diffeomorphism from D() to a closed tubular neighbour-
hood W of Q in M. One has the diagram

/ M
Q
Y AA } >
AAi j }}
AA }}
A }}
W

where is the bundle projection and the other maps are inclusions. By excision,

j
H (M, M Q) o
H (W, Bd W ) H (D(), S()).

Hence, the Thom class of may be seen as an element U H q (W, Bd W ) satisfying

U = j (U  ) for a unique U  H q (M, M Q). Let U  H q (M) be the image of
U  under the restriction homomorphisms H q (M, M Q) H q (M).
By definition of the Stiefel-Whitney class w = w(T Q) = w(), one has

(w)  U = Sq U.

One has (1 w) = 1  w = w, whence j (1 w) = (w). Hence, the

equation (1 w)  U  = Sq U  holds true in H (M, M Q) which, in H (M),
implies

(1 w)  U  = Sq U  . (9.8.4)

Without loss of generality, we may assume that Q is connected. Let A = {a1 , a2 , . . . }

and B = {b1 , b2 , . . . } of H (Q) be an additive bases of H (Q) which are Poincar
dual, i.e. ai  b j , [Q] = i j . We suppose that a0 = 1. By Lemma 5.4.2 and
Eq. (5.4.1), one has

U  = ai bi
i

and therefore

U  /[Q] = (1 b0 )/[Q] = 1.
414 9 Stiefel-Whitney Classes

Applying this together with Eqs. (9.8.4) and (9.8.3) gives

Sq U  /[Q] = [(1 w)  U  ]/[Q] = w  (U  /[Q]) = w . (9.8.5)


We now express the Wu class v = v(M) in the A-basis: v = i i ai . Then,
v  b j , [Q] = j , which implies that
 
v= v  bi , [Q] ai = Sq bi , [Q]ai , . (9.8.6)
i i

Hence, using (9.8.5), we get


Sq v = Sq bi , [Q] Sq ai
i
 
= Sq ai Sq bi /[Q]
i
= Sq U  /[Q]
= w.

The remainder of this subsection is devoted to general corollaries of Wus for-

mula. The first one says that the Stiefel-Whitney class w(T Q) depends only on the
homology type of Q.

Corollary 9.8.3 Let f : Q  Q be continuous map between smooth closed man-

ifolds Q and Q  of the same dimension. Suppose that H f : H (Q) H (Q  ) is
surjective. Then H f (w(T Q)) = w(T Q  ).

Proof By Kronecker duality, H f is injective and thus 0 f is injective. The con-

nected components of Q out of the image of f play no role, so one may assume that
0 f is a bijection. This implies that H f ([Q  ]) = [Q].
Let v  = H f (v(Q)). For b H (Q), one has

v(Q)  b, [Q] = v(Q)  b, H f [Q  ]

= H f (v(Q)  b), [Q  ]
= v   H f (b), [Q  ].

On the other hand

v(Q)  b, [Q] = Sq b, H f [Q  ]

= Sq(H f (b)), [Q  ]
= v(Q  )  H f (b), [Q  ].
9.8 The Wu Formula 415

Therefore, the equality

v   H f (b), [Q  ] = v(Q  )  H f (b), [Q  ]

is valid for all b H (Q). As H f is surjective, Theorem 5.3.12, this implies that
v  = V (Q  ), so H f (v(Q)) = v(Q  ). By Wus formula,

w(T Q  ) = Sq v(Q  ) = Sq H f (v(Q)) = H f (Sq v(Q)) = H f (w(T Q)). 

Corollary 9.8.4 Let f : Q Q be continuous map of degree one. Then

H f (w(T Q)) = w(T Q).

Proof By Proposition 5.2.8, H f is surjective and then H f is injective by Kronecker

duality. As H (Q) is a finite dimensional vector space, this implies that H f (and
H f ) are bijective. The results then follows from Corrolary 9.8.3. 

9.8.2 Orientability and Spin Structures

A smooth manifold is orientable if its tangent bundle is orientable. The following

corollary generalizes Proposition 4.2.3.

Corollary 9.8.5 Let Q be a smooth closed n-dimensional manifold. Then Q is ori-

entable if and only if Sq1 : H n1 (Q) H n (Q) vanishes.

Proof By Proposition 9.4.4, Q is orientable if and only if w1 (T Q) = 0 which, by

Wus formula, is equivalent to v1 (Q) = 0. By the definition of v1 (Q), its vanishing
is equivalent to Sq1 : H n1 (Q) H n (Q) being zero. 

The same argument, using Proposition 9.4.7, implies the following result.

Corollary 9.8.6 Let Q be a smooth closed n-dimensional manifold which is ori-

entable. Then T Q admits a spin structure if and only if Sq2 : H n2 (Q) H n (Q)
vanishes.

Example 9.8.7 A closed manifold M such that H (M) is GrA-isomorphic to

H (RP n ) is orientable if and only if n is odd. Indeed, let 0 = a H 1 (M) = Z2 . By
the Cartan formula, Sq1 (a n1 ) = 0 if and only if n is odd. As H n1 (M) is generated
by a n1 , the assertion follows from Corollary 9.8.5. A similar argument, using Corol-
lary 9.8.6, proves that a closed manifold M such that H (M) is GrA-isomorphic to
H (CP n ) admits a spin structure if and only if n is odd.

In the particular case n = 4, Corollary 9.8.6 gives the following result.

Corollary 9.8.8 Let Q be a smooth connected closed 4-dimensional manifold which

is orientable. Then,
416 9 Stiefel-Whitney Classes

(1) a  a = w2 (T Q)  a for all a H 2 (Q).

(2) T Q admits a spin structure if and only if the cup-square map H 2 (Q) H 4 (Q)
vanishes.
(3) w4 (T Q) = w2 (T Q)  w2 (T Q).
Point (2) is the analogue of Proposition 4.2.3 for surfaces. In particular, T CP 2
does not admit a spin structure.
Proof If w1 (T Q) = 0, then w2 (T Q) = v2 (Q) by Wus formula. Hence, a 
a = Sq2 (a) = v2 (Q)  a = w2 (T Q)  a for all a H 2 (Q), which proves (a).
Point (b) thus follows from Corollary 9.8.6. For Point (c), we use Wus formula again:

w4 (T Q) = Sq2 (v2 (Q)) = Sq2 (w2 (T Q)) = w2 (T Q)  w2 (T Q). 

Corollary 9.8.9 Let Q be a smooth closed manifold of dimension n 7. If T Q

admits a spin structure, then w(T Q) = 1.
Proof The proposition is obvious for n 2. Otherwise, by Proposition 9.4.6, the
existence of a spin structure implies that the restriction of T Q over the 3-skeleton of
a triangulation of Q is trivial. Hence w3 (T M) also vanishes which, by Wus formula,
implies that vi (M) vanishes for i 3. As n 7, this implies that v(Q) = 1 and
thus w(T Q) = Sq v(Q) = 1. 
An interesting example is given by the projective spaces.
Proposition 9.8.10 Let 0 = a H 1 (RP n ). The Stiefel-Whitney class of the tangent
space of RP n is

w(T RP n ) = (1 + a)n+1

and the Wu class of RP n is

[n/2]
 ni  i
v(RP n ) = i a .
i=0

Here are a few examples.

n v(R P n ) w(T R P n )
2 1+a 1 + a + a2
3 1 1
4 1 + a + a2 1 + a + a5
5 1 + a2 1 + a2 + a4
6 1 + a + a3 1 + a + a2 + a3 + a4 + a5 + a6
7 1 1
9.8 The Wu Formula 417

Remark 9.8.11 The formulae of Proposition 9.8.10 imply the following.

(1) RP n is orientable if and only if n is odd (this is not a surprise!). More generally,
w2i+1 (T RP 2k+1 ) = 0.
(2) T RP n admits a spin structure if and only if n 3 mod 4. In this case, there
are two spin structures, since H 1 (RP n ) = Z2 . For a discussion about these two
structures for RP 3 S O(3), see [130, Example 2.5, p. 87].
(3) w(T RP n ) = 1 if and only if n = 2k 1. But T RP n is trivial if and only if
n = 1, 3, 7 by Adams Theorem [2, p. 21].

Proof of Proposition 9.8.10 The two formulae will be proved separately. Checking
Wus formula is left as an exercise. By (8.2.2),
ni  n ni  i
vi (RP n )  a ni = Sqi a ni = i a = i a a
ni
.
  i
which proves that vi (RP n ) = ni
i a . This proves the formula for the Wu class.
As for the Stiefel-Whitney class, the idea is the following. Recall that RP n =
Gr(1; Rn+1 ) = Fl(1, n). Write = 1 and = 2 for the tautological bundles.
Then,

T RP n hom( , ) (9.8.7)

(see [153, Lemma 4.4] for a proof). But is the trivial bundle n+1 of rank
n + 1. Adding to both side of (9.8.7) the bundle hom( , ) 1 , we get

T RP n 1 hom( , n+1 ).

The latter is the Whitney sum of (n + 1)-copies of = hom( , 1 ). But ,

using an Euclidean metric on . For details (see [153, Theorem 4.5]). Hence, the
formula for w(T RP n ) follows from (9.4.3).

Remark 9.8.12 The argument of the proof of Proposition 9.8.10 essentially works
for computing the Chern class of the tangent bundle to T CP n (which is a complex
vector bundle). The slight difference is that = hom( , 1 ) is not, as complex
vector bundle, isomorphic to but to the conjugate bundle (the complex structure
on each fiber is the conjugate of that of (see [153, pp. 169170]). But this does not
alter our Chern classes which are defined mod 2: ci (T CP n ) = w2i (T CP n ). Thus

[n/2]
 ni  i
c(T CP n ) = (1 + a)n+1 and v(CP n ) = i a
i=0

where 0 = a H 2 (CP n ). The first formula holds as well for the integral
Chern class with a suitable choice of a generator of H 2 (CP n ; Z) (see
[153, Theorem 14.10]).
418 9 Stiefel-Whitney Classes

9.8.3 Applications to 3-Manifolds

Wus formula has two important consequences for closed 3-dimensional manifolds.
The first one is the following.

Proposition 9.8.13 Let Q be a smooth closed 3-dimensional manifold which is

orientable. Then T Q is a trivial vector bundle (in other words: Q is parallelizable).

Proof For any smooth closed manifold, one has w1 (T Q) = v1 (Q) by Wus for-
mula. Thus, v1 (Q) = 0 if Q is orientable. In dimension 3, this implies that
v(Q) = 1 and, by Wus formula again, w(T Q) = 1. The result then follows from
Proposition 9.4.6. 

The second application is Postnikovs characterization of the cohomology ring

of a closed connected 3-dimensional manifold [164]. Let M be such a manifold.
Consider the symmetric trilinear form M : H 1 (M) H 1 (M) H 1 (M) Z2
defined by
M (a, b, c) = a  b  c, [M].

The first observation is that M determines the ring structure of H (M).

Lemma 9.8.14 Let M and M be two closed connected 3-dimensional manifolds.

such that
Suppose that there exists an isomorphism h 1 : H 1 (M) H 1 ( M)

M (h 1 (a), h 1 (b), h 1 (c)) = M (a, b, c).

Then, h 1 extends to a GrA-isomorphism h : H (M) H ( M).

Proof Let A = {a1 , . . . , am } be a Z2 -basis of H 1 (M). The set A = {a 1 , . . . , a m }

where a i = h 1 (ai ) is then a Z2 -basis of H 1 ( M). Let B = {b1 , . . . , bm } and B =
{b1 , . . . , bm } be the bases of H 2 (M) and H 2 ( M)
which are Poincar dual to A and
i.e. the equations
A,

ai  b j , [M] = i j and a i  b j , [ M]

= i j (9.8.8)

are satisfied for all i, j. Let h 2 : H 2 (M) H 2 ( M) be the isomorphism such

that h 2 (bi ) = bi and let h 3 be the unique isomorphism from H 3 (M) to H 3 ( M).
This produces a GrV-isomorphismh : H (M) H ( M). To prove that h is a
m
GrA-morphism, write ai  a j = l=1 li j bl and, using (9.8.8), note that


m
M (ai , a j , ak ) = ai  a j  ak , [M] = li j bl  ak , [M] = ikj .
l=1
9.8 The Wu Formula 419

Therefore,


m
=
h 2 (ai  a j )  a k , [ M]
M (ai , a j , al )h 2 (bl )  a k , [ M]
l=1

m
= M (ai , a j , al )bl  a k , [ M]

l=1
= M (ai , a j , ak )

and

= h 1 (ai )  h 1 (a j )  h 1 (ak ), [ M]

h 1 (ai )  h 1 (a j )  a k , [ M]
= M (h 1 (ai ), h 1 (a j ), h 1 (ak )).

Since M (h 1 (ai ), h 1 (a j ), h 1 (ak )) = M (ai , a j , ak ), this proves that h 2 (ai  a j ) =

h 1 (ai )  h 1 (a j ). On the other hand, h 3 formally satisfies h 3 (u), [ M] = u, [M].
Hence, the equality h (ai )  h (b j ) = h (ai  b j ) follows from (9.8.8). We have
1 2 3

thus established that h is a GrA-morphism. 

The trilinear form M is linked to the Wu class v(M) H 1 (M).
Lemma 9.8.15 Let M be a closed connected 3-dimensional manifold. Then, the Wu
class v = v1 (M) satisfies

M (v, b, c) = M (b, b, c) + M (b, c, c) (9.8.9)

for all b, c H 1 (M).

Proof This comes from that

v1 (M)  (b  c) = Sq1 (b  c) = Sq1 (b)  c + b  Sq1 (c)

= b  b  c + b  c  c. 
The following realizability result is due to Postnikov [164].
Proposition 9.8.16 Let (V, ) a symmetric trilinear form, with V a finite dimen-
sional Z2 -vector space. Let v V satisfying (9.8.9). Then, there exists a closed
connected 3-manifold M with an isometry (H 1 (M), M ) (V, ), sending v1 (M)
onto v.
Proof (indications) The full proof may be found in [164]. When M is orientable, the
form M is alternate
  vanishes (v = w1 (T M)
since the left hand side of (9.8.9) = 0).
Hence, M 3 H 1 (M). An alternate form 3 V may be lifted to 3 V ,
where V is a free abelian group with V Z2 V . In [187], D. Sullivan constructed
a closed connected orientable 3-manifold M with (H 1 (M; Z), M ) (V , ), which
thus proves Proposition 9.8.16 in the orientable case. 
420 9 Stiefel-Whitney Classes

9.8.4 The Universal Class for Double Points

The material of this section is essentially a rewriting in our language of results

of Haefliger [78]. Let M be a closed manifold of dimension m. Let G = {1, }
acting on M M by (x, y) = (y, x), with fixed point set (M M)G = M ,
the diagonal submanifold of M M. The diagonal inclusion : M M M

induces a diffeomorphism : M M . For N > 1, S N G (M M) is a closed
manifold containing RP N M as a closed submanifold of codimension m. Let
PDG,N (M) = PD(RP N M ) H m (S N G (M M)), the Poincar dual of
RP N M (see Sect. 5.4.1). If N is big enough,

HGm (M M) H m (S G (M M)) H m (S N G (M M))

is an isomorphism. Therefore, there is a unique class PDG (M) HGm (M M)

whose image in H m (S N G (M M)) is equal to PDG,N ( M ).
The class PDG (M) is called the universal class of double points for continuous
maps into M, a terminology justified by Lemma 9.8.17. For a space X , denote by
j : X 0 (X X ) the inclusion of X 0 = (X X ) X into (X X ). As G acts
freely on X 0 , the quotient space X = X 0 /G, called the reduced symmetric square
of X , has the homotopy type of X G 0 by Lemma 7.1.4.

The diffeomorphism : M M is covered by a bundle isomorphism :

TM (M M, M ) (see Lemma 5.4.17), which intertwines with the antipodal
the sphere bundle T 1 M becomes G-diffeomorphic with
involution on T M. Via ,
the boundary BdW of a G-invariant tubular neighbourhood of M in M M. Thus,

W (W M )G (BdW )G (T 1 M)G (T 1 M)/G. (9.8.10)

As j is G-equivariant, it induces HG j : HG (X X ) HG (X 0 ). Let f : Q M

be a continuous map. Consider the homomorphism

HG ( f f ) HG j
H m (Q ).
: HGm (M M) HGm (Q Q) HGm (Q 0 )

We denote by : HGm (M M) H m (T 1 Q/G) the post-composition of with

the homomorphism H m (Q ) H m (W ) H m (T 1 N /G). Define

(PDG (M)) HGm (Q ) and O G = (PDG (M)) HGm ((T 1 Q)/G).

f f
OG =

Lemma 9.8.17 Let f : Q M be a continuous map between closed manifolds.

Then
f
(1) if f is homotopic to an embedding, then OG = 0.
(2) if f is homotopic to an immersion, then O G = 0.
f
9.8 The Wu Formula 421

Proof The classes OG and O G depend only on the homotopy class of f . If f is

f f

injective, then ( f f )(Q 0 ) M 0 and thus ( f f )G (Q 0G ) MG0 = (M M)G

( M )G . By Lemma 5.4.2, PDG (M) has image zero in HGm (M 0 ), which proves (1)
(this does not use that Q is a manifold). Suppose that f is an immersion, so f is
locally injective. As Q is compact, there is a G-invariant tubular neighbourhood W
of Q in Q Q such that ( f f )(W 0 ) M 0 . We deduce (2) as above for (1). 

In order to get applications of Lemma 9.8.17, we now express PDG,N (M) within
the description of HGm (M M) given by Proposition 8.3.3, which can be rephrased
as follows. There is a GrA[u]-isomorphism from HG (M M) to (Z2 [u] D) N ,
where D is the Z2 -vector space generated by {x x | x H k (M), k 0} so that
: HG (M M) H (M M)G sends the elements of D N (elements of
u-degree 0) isomorphically to H (M M)G . The Z2 -vector space N is generated
by {x y + y x | x, y H (M)} and coincides with the ideal ann (u).

Proposition 9.8.18 Using the isomorphism HGm (M M) (Z2 [u] D) N , we

have
[m/2]
(1) PDG (M) k=0 u m2k (vk (M) vk (M)) mod N , where vk (M) is the k-th
Wu class of M.
(2) (PDG (M)) = PD( M ), the Poincar dual of M in M M.

Example 9.8.19 Let M = RP 2 . One has H (M) = Z2 [a]/(a 3 ) and v(M) = 1 + a.

Then, PDG (M) u 2 + a a mod N and, according to Eq. (5.4.1), (PDG (M)) =
PD( M ) = 1 a 2 + a a + a 2 1. Therefore,

PDG (RP 2 ) = u 2 + a a + N (1 a 2 ).
Proof It is enough to prove (1) for PDG,N (M) with N big enough. Let i : Q P
be the inclusion of a closed manifold Q into a compact manifold P. For x H j (P),
one has

x  PD(Q), [P] = x, PD(Q)  [P] = x, H i([Q]) = H i(x), [Q].

(9.8.11)
We shall apply (9.8.11) to Q = RP N M and P = S N G (M M) and x =
u N m+2i (a a) H N +m (P), where a H mi (M). One has

H i(x), [Q] = u N m+2i H i(a a), [Q]


= u N m+2i mi j=0 u
mi j Sq j (a), [Q] by definition of Sq(a)

= u Sq (a), [Q]
N i only non-zero term
= Sqi (a), [M].
(9.8.12)
For y, z H j (M), we have (y y) + (z z) (y + z) (y + z) mod N
and u N = 0. Therefore, for k > 0, u k PDG,N (Q) admits an expression of the form
[m/2]
u k PDG,N (Q) = u k j=0 u m2 j (y j y j ) with y j H j (M). Hence,
422 9 Stiefel-Whitney Classes

[m/2]
x  PDG,N (Q), [P] = u N m+2i (a a)  u m2 j (y j y j ), [P]
j=0
= u N (a a)  (yi yi ), [P]
= u N {(a  yi ) (a  yi )}, [P]
= a  yi , [M] (9.8.13)

Using (9.8.11), Formulae (9.8.12) and (9.8.13) imply that yi = vi (M) for all i =
0 . . . , [m/2]. This proves (1).
For Point (2) we must prove that N (PDG,N ) = PD(M) where N is induced by
the fiber inclusion M M S N (M M) P. But this map is transversal to
RP N M . Point (2) thus comes from Proposition 5.4.10. 

Proposition 9.8.18 enables us to compute the image of PDG (M) under the homo-
morphism

/ H ((M M)G ) / H (M)[u] ev1

/ H (M)
HG (M M) r
G

Corollary 9.8.20 ev1 r (PDG (M)) = w(T M), the total Stiefel-Whitney class of
the tangent bundle T M.

Proof
[m/2]  
ev1 r (PDG (M)) = k=0 ev1 r vk (M) vk (M) by Proposition 9.8.19
[m/2]
= k=0 Sq(vk (M)) by (8.3.5)
= Sq(v(M))
= w(T M) by the Wu formula. 

Remark 9.8.21 The formula of Corollary 9.8.20 may be proven directly in the
following way. By Lemma 5.4.4 and the considerations before (9.8.10), one has
r (PDG (M)) = eG (T M), where G acts on T M by the antipodal action on each
fiber. This equivariant Euler class satisfies ev1 (eG (T M)) = w(T M) (see (10.4.7) in
Sect. 10.4), which proves Corollary 9.8.20. Moreover, using the proof of
Corollary 9.8.20, with the last line removed, gives a new proof of the Wu formula.

Lemma 9.8.22 Let Q be a closed manifold of dimension p. There is a commutative

diagram

GysG
0 / H p ( Q ) / H (Q Q) / H (Q ) / 0
G G

HG r
  
/ H p (Q)  / H (Q) / H ((T 1 Q)/G) / 0
0 G G
9.8 The Wu Formula 423

where the rows are exact sequences. Here, (a) = a  eG (T Q), where G acts on
T Q via the antipodal involution.

Proof The diagram comes from Proposition 5.4.10 applied, for N big, to the pair
(S N G (Q Q), RP N Q ). The long diagram of Proposition 5.4.10 splits into
the above diagram because  is injective (see Proposition 7.5.14). The bottom line
is the Gysin sequence for the sphere bundle (T 1 Q)G Q G . 

The above results permit us to express an obstruction to embedding in terms

of the dual Stiefel-Whitney classes. Let f : Q M be a smooth map. Define
Q)  H f (w(T M)) H (Q) where w(T
w f = w(T Q) is the dual Stiefel-
Whitney classes of T Q (see p. 389).

Proposition 9.8.23 Let f : Q q M m be a smooth map between closed manifolds.

f
(1) If f is homotopic to a smooth immersion, then w k = 0 for k > m q.
(2) If f is homotopic to a smooth embedding, then

(w mq 1)  PD( Q ) = H ( f f )(PD( M ))
f

in H m (Q Q).

Proof We shall argue with the help of the diagram

HG ev1
HG
m p
( Q ) / H m p (Q) / H (Q)
G

GysG eG (T Q) w(T Q) (9.8.14)

  
ev1
HGm (Q Q) r / H m (Q) / H (Q)
G

The left square comes from Lemma 9.8.22 and is thus commutative. So is the right
square by Remark 9.8.21.
If f is homotopic to a smooth immersion, then O G = 0 by Lemma 9.8.17. By
f
mq
Lemma 9.8.22, this implies that there exists b HG (Q) such that the equations

ev1 r HG ( f f )(PDG (M)) = ev1 (b  eG (T Q)) = ev1 (b)  w(T Q)

hold true in H (Q). But

ev1 r HG ( f f )(PDG (M)) = H f ev1 r (PDG (M))

= H f (w(T M)) by Corollary 9.8.20.

Hence

ev1 (b)  w(T Q) = H f (w(T M)). (9.8.15)

424 9 Stiefel-Whitney Classes

Since w(T
Q)  w(T Q) = 1, multiplying both sides of (9.8.15) by w(T
Q) gives

ev1 (b) = H f (w(T M))  w(T

Q) = w f . (9.8.16)

f
As b is of degree m q, Eq. (9.8.16) implies that w k = 0 for k > m q. We have
thus proven (1). Also, (9.8.16) is equivalent to


mq
f
b= w i u mqi . (9.8.17)
i=0

To prove Point (2), we use the diagram

HG GysG
(Q) o / H m (Q Q)
m p m p
HG
HG ( Q ) G

(9.8.18)
  
H Gys
H m p (Q) o
H m p ( Q ) / H m (Q Q)

The left square is obviously commutative and the right square is so by Proposition
5.4.11, since the fiber inclusion Q Q S N G (Q Q) is transversal to RP N
f
Q . If f is homotopic to a smooth embedding, then OG = 0 by Lemma 9.8.17.
Also, (1) holds and, by the above and Diagram (9.8.14), one has

1 (b)
HG ( f f )(PDG (M)) = GysG (HG ) (9.8.19)

By Point (2) of Proposition 9.8.18, one has

HG ( f f )(PDG (M)) = H ( f f ) (PDG (M)) = H ( f f )(PD( M )).

(9.8.20)
Let : Q Q Q be the inclusion and pr2 : Q Q Q be the projection
onto the first factor. Then pr1 = id Q . Hence, any a H (Q) satisfies

a = H H H pr1 (a) = H H (a 1). (9.8.21)

Hence,

1 (b) =
GysG (HG ) 1 (b)
Gys (H ) by commutativity of (9.8.18)
=
Gys (H ) 1 (w mq
f
) by (9.8.17)
= f
Gys H (w mq 1) by (9.8.21)
f
= (w mq 1)  PD( Q ) by Lemma 5.4.8
(9.8.22)
Combining (9.8.19), (9.8.20) and (9.8.22) provides the proof of Point (2). 
9.8 The Wu Formula 425

Corollary 9.8.24 Let Q be a closed manifold of dimension q.

(1) If Q may be immersed in Rm , then w k (T Q) = 0 for k > m q.
(2) If Q may be embedded in Rm , then w mq (T Q) = 0.

Point (1) was already proven in Proposition 9.5.22.

Proof Let f 0 : Q Rm be a smooth map. Composing with the inclusion Rm 

S m = Rm {} gives a smooth map f : Q S m , homotopic to a constant
map. Then w f = w(T Q) and H ( f f ) = 0. Corollary 9.8.24 thus follows from
Proposition 9.8.23. 

Examples 9.8.25 1. Let Q be a closed non-orientable surface. Then, w 1 (T Q) =

w1 (T Q) = 0. Therefore, Q cannot be embedded in R3 . Note that M can be
embedded in R4 by Whitneys theorem and immersed in R3 using Boys surface.
2. Let Q be a closed 4-dimensional orientable manifold which is not spin (for
instance CP 2 ). Then, w 2 (T Q) = w2 (T Q) = 0. Therefore, Q cannot be embed-
ded in R6 . Note that CP 2 embeds in R7 . Indeed, CP 2 is diffeomorphic to the
space FlC (1, 2) of Hermitian (3 3)-matrices with characteristic polynomial
equal to x 2 (x 1) (see (3) on p. 405). The vector space of Hermitian (3 3)-
matrices with trace 1 is isomorphic to R8 and each radius intersects FlC (1, 2) at
most once. This gives an embedding of CP 2 in S 7 = R7 {} and thus in R7 .
3. The quaternionic projective plane Q = HP 2 has Wu class v4 (Q) = 0. Hence,
w 4 (T Q) = w4 (T Q) = 0. Therefore, HP 2 cannot be embedded in R12 . In
the same way, the octonionic projective plane OP 2 of dimension 16 cannot be
embedded in R24 . Improving the technique explained in the previous example
produces embeddings HP 2  R13 and OP 2  R25 (see [135, Sect. 3]).

9.9 Thoms Theorems

This section is a survey of some results of Thoms important paper [191], which,
amongst other things, was the foundation of cobordism theory. Some proofs are
almost complete and others are just sketched.

9.9.1 Representing Homology Classes by Manifolds

Theorem 9.9.1 Let X be a topological space and Hk (X ). Then, there exists a

closed smooth manifold M of dimension k and a continuous map f : M X such
that H f ([M]) = .

This theorem is due to Thom [191, Theorem III.2]. The result is wrong for inte-
gral cohomology (see [191, Theorem III.9]). This section is devoted to the proof of
Theorem 9.9.1. We start with some preliminaries.
426 9 Stiefel-Whitney Classes

Let be a vector bundle of rank r over a paracompact space Y . Let (D( ), S( ))

be the pair of the disk and sphere bundles associated to via an Euclidean structure.
The Thom space T ( ) of is defined by

T ( ) = D( )/S( ).

The homeomorphism class of T ( ) does not depend on the choice of the Euclid-
ean structure (see p. 187). Also, S( ) has a collar neighborhood in D( ). Hence,
by Lemma 3.1.39, the pair (D( ), S( )) is well cofibrant. Using Proposition 3.1.45
together with the Thom isomorphism theorem provides the following isomorphisms

H k (B)
H k+r (D( ), S( ))
H k+r (T ( )).

We now specialize to = r,N , the tautological vector bundle over the Grass-
mannian Gr(r ; R N ) (N ). (The Thom space T (r, ) is also called M O(r ) in the
literature; it is the r -th space of the Thom spectrum M O). We get some information
on the cohomology ring H (T (r,N )) using the Gysin exact sequence of the sphere
bundle q : S(r,N ) Gr(r ; R N ), whose Euler class is wr .

wr H q
H ir (Gr(r ; R N )) H i (Gr(r ; R N )) H i (S(r,N ))
wr
H ir +1 (Gr(r ; R N )) (9.9.1)

together with the exact Sequence of Corollary 3.1.48 for the pair (D(r,N ), S(r,N ))
(using that D(r,N
= Gr(r ; R N ))

H q
H i (T (r,N )) H i (Gr(r ; R N )) H i (S(r,N )) H i+1 (T (r,N )). (9.9.2)

For N = ,  wr is injective by Theorem 9.5.8. Together with Lemma 9.5.7,

Sequences (9.9.1) and (9.9.2) gives the following lemma.

Lemma 9.9.2 (1) The GrA-morphism

H p : H (T (r, )) H (Gr(r ; R ))

is injective and its image in positive degrees is the ideal generated by wr . In

particular, H p(U ) = wr .
(2) The GrA-morphism H i (T (r, )) H i (T (r,N )) generated by the inclusion
is bijective for N N r 1.
(Note that the equation H p(U ) = wr is coherent with (4.7.21)). Consider the
diagram
9.9 Thoms Theorems 427

g
Gr(r ; R ) o (RP )r
MMM
MMfMwr
p MMM f aa (9.9.3)
 M& 
fU
T (r, ) / Kr

in which the following notations are used. If X is a CW-complex and y H r (X ), then

f y : X Kr = K (Z2 , r ) denotes a map representing y. Then wr H r (Gr(r ; R ))
is the r -th Stiefel-Whitney class, U H r (T r, ) is the Thom class and 0 =
a H 1 (RP ). The map g classifies the r -th product of the tautological line bun-
dle over RP and p : Gr(r ; R ) = D(r, ) T (r, ) is the quotient map.
Diagram (9.9.3) is homotopy commutative. The commutativity of the lower triangle
comes from the already mentioned equation H p(U ) = e(r, ) = wr . For the
upper triangle (see e.g. the proof of Theorem 9.5.8).
Applying the cohomology functor to Diagram (9.9.3) provides the commutative
diagram
H g
H (Gr(r ; R )) / / H ((RP )r )
O hQQQ O
QQHQ fwr
H p QQQ H f aa . (9.9.4)
O QQ

H fU
H (T (r, )) o H (Kr )

By Corollary 8.5.9, H f aa : H i (Kr ) H i ((RP )r ) is injective for i 2r .

Therefore, H fU : H i (Kr ) H i (T (r, )) is also injective for i 2r . As indicated
in Diagram (9.9.4), H p and H g are injective. The injectivity of H p was proven
in Lemma 9.9.2 and that of H g was established in the proof of Theorem 9.5.8 or in
Theorem 9.6.1 and its proof.
The map H fU is of course not surjective. By Lemma 9.9.2, dim H r + j (T r, )
is the number of partitions of j while, for i < r , dim H r + j (Kr ) is equal, by
Lemma 8.5.14, to the number of partitions of j into integers of the form 2i 1.
Let D( j) be the number of partitions of j into integers with none of them of the form
2i 1. For each D( j) with j r , Thom constructs a class X H r + j (T r, )
represented by a map f : T r, K j . Together with fU , this gives a map


r
D ( j)
F : T r, Y = Kr Kj = Kr Kr +2 . (9.9.5)
j=1

Thom proves that H F is an isomorphism up to degree 2r . Some analogous result

is proved for the cohomology with coefficients in a field of characteristic = 2 and
both T r, and Y are simply connected. This enables Thom to prove the following
result (see [191, pp.3542]).
428 9 Stiefel-Whitney Classes

Lemma 9.9.3 If N N {} is big enough, there exists a map from the 2r -

skeleton of Y to T (r,N ) such that the restrictions of F and F to the (2r 1)-
skeleta of Y and T r,N respectively are homotopic to the identity.
As a corollary, we get the following result (see [191, Corollary II.12]).
Corollary 9.9.4 If N N {} is big enough, there exists a map from the
2r -skeleton of Kr to T r,N such that H (U ) = , the fundamental class of Kr .
We are now ready to prove Theorem 9.9.1.
Proof of Theorem 9.9.1 By Theorem 3.7.4, there is a simplicial complex K X and a
map : |K X | X such that H is an isomorphism. By Sect. 3.6, The homology
of X is isomorphic to the simplicial homology of K X . By the definition of the
simplicial homology, there is a finite simplicial subcomplex K of K X , of dimension
k, such that is in the image of Hk (|K |) Hk (X ). Now, there is a PL-embedding
: |K | R2k+1 (see e.g. [179, Theorem 3.3.9]) and the theory of smooth regular
neighborhoods [94] produces a smooth compact codimension 0 submanifold W of
R2k+1 which is a regular neighborhood of (|K |) for some C 1 -triangulation of
R2k+1 . In particular, W retracts by deformation on (K ). By general position, is
isotopic to an embedding  such that  (|K |) avoids some regular neighborhood W 
of (|K |) contained in the interior of W . The closure of W  W is homeomorphic
to M [0, 1], where M = BdW (see [104, Corollary 2.16.2, p. 74]). We can thus
construct a map  : |K | M such that the composite map


|K | M  W |K |

is isotopic to id|K | . Hence is in the image of H (M) Hk (X ). Therefore, it is

enough to prove Theorem 9.9.1 when X is a closed manifold of dimension 2k.
Let a H k (X ) be the cohomology class which is Poincar dual to . As a
smooth manifold, X admits a C 1 -triangulation by a simplicial complex of dimension
2k. There exists thus a continuous map f a : X Kk representing the class a and, by
cellular approximation, one may suppose that f a (X ) is contained in the 2k-skeleton
of K. Let f = f a : X T = T (k,N ), where is a map as provided by
Corollary 9.9.4 for N large enough. Then fU f is homotopic to f a . Note that T
is a smooth manifold except at the point [S(k,N )]. Using standard techniques of
differential topology (see [95, Sects. 2.2 and 3.2]), f is homotopic to f 1 such that f 1
is a smooth map around f 11 (Gr(r ; R N )) which is transverse to Gr(r ; R N ). Then
M = f 11 (Gr(r ; R N )) is a closed submanifold of codimension k in X with normal
bundle = f 1 k,N .
Let a  H k (X ) be the Poincar dual of the homology class generated by [M]. As
in Sect. 5.4.1, we consider the Thom class U (X, M) of as an element of H k (X, X
M) and, if j : (X, ) (X, X M) denotes the pair inclusion, one has

a  = H j (U (X, M)) by Lemma 5.4.2

= H f 1 (U )
9.9 Thoms Theorems 429

= H ( f a )(U )
= H f a H (U )
= H f a ()
= a.

As Poincar duality is an isomorphism, this proves that [M] = .

Observe that, in the proof of Theorem 9.9.1, we have established the following
result, due to Thom [191, Theorem II.1, p. 29].
Proposition 9.9.5 Let Hk (X ), where X is a closed smooth manifold of dimen-
sion k + q > 2k. Let a = PD() H q (X ) be the Poincar dual of . Then, the
following statements are equivalent.
(1) There exists a closed submanifold M in X such that [M] represents .
(2) There exists a continuous map F : X T (q, ) such that H F(U ) = a.

9.9.2 Cobordism and Stiefel-Whitney Numbers

Let M be a (smooth, possibly disconnected) manifold of dimension n. For a polyno-

mial P Z2 [X 1 , . . . , X n ], we set PM = P(w1 (T M), . . . , wn (T M)) H (M). If
M is closed, the number mod 2

PM , [M] Z2

is called the Stiefel-Whitney number of M associated to P. We use the convention

that a,  = 0 if a H r (X ) and H s (X ) with r = s.
Two closed manifolds of the same dimension are called cobordant if their disjoint
union is the boundary of a compact manifold. One fundamental result of Thom [191,
Theorema IV.3 and IV.10] is the following theorem, generalizing Corollary 5.3.10.
Theorem 9.9.6 Two closed manifolds of the same dimension are cobordant if and
only if their Stiefel-Whitney numbers coincide.
Example 9.9.7 Let M be a closed 3-dimensional manifold. Its Wu class is v(M) =
1+v1 (M) = 1+w1 (T M). By Wus formula, w(T M) = Sq(v(M)) = 1+w1 (T M)+
w1 (T M)2 , so w2 (T M) = w1 (T M)2 . The only possible non-zero Stiefel-Whitney
number is then w1 (T M)3 , [M]. But

w1 (T M)3 = w1 (T M)2 v1 (M) since w1 (T M) = v1 (M)

= Sq1 (w1 (T M)2 ) by definition of v1 , since dim M = 3
=0 by the Cartan formula.

Therefore, M is the boundary of a compact manifold. Note that, if M is orientable,

the vanishing of its Stiefel-Whitney numbers follows from Proposition 9.8.13.
430 9 Stiefel-Whitney Classes

Example 9.9.8 The complex projective space CP 2 and the manifold RP 2 RP 2

have the same Stiefel-Whtney numbers by Proposition 9.8.10 and Remark 9.8.11.
Therefore, they are cobordant.

Example 9.9.9 Let M be a closed orientable 4-dimensional manifold. Then,

w1 (T M) = 0 and w4 (T M) = w2 (T M)2 (see Corollary 9.8.8). Its only possible
non-vanishing Stiefel-Whitney number is thus

w4 (T M), [M] = e(T M), [M] = (M) mod 2

(using Corollary 5.4.16). Therefore, M a boundary of a (possibly non-orientable)

compact 5-manifold if and only if its Euler characteristic is even.

Proof of Theorem 9.9.6 Let M1 and M2 be two closed manifolds of the same dimen-
2 . For any P Z2 [X 1 , . . . , X n ], one has
sion and let M = M1 M

PM , [M] = PM1 , [M1 ] + PM2 , [M2 ].

Hence, Theorem 9.9.6 is equivalent to the following statement: a closed manifold M

bounds if and only if its Stiefel-Whitney numbers vanishes.
Suppose that M = Bd W for some compact manifold W . Then T M T W|M
where is the trivial bundle of rank 1 over M. If j : M W denotes the inclusion,
one has

PM , [M] = H j (PW ), [M] = PW , H j ([M]) = 0,

since H j ([M]) = 0 (see Eq. (5.3.6) and the end of the proof of Theorem 5.3.7).
For the converse, we shall prove that if a closed manifold M of dimension n
does not bound, then at least one of its Stiefel-Whitney numbers is not zero. Let us
embed M into Rn+r for r large, with normal bundle . Let f : M Gr(r ; R )
be a map such that f r, . The map f induces a map T f : T T r, .
A closed tubular neighbourhood N of M is diffeomorphic to D(). We consider
Rn+r S n+r . The projection N D() T () extends to a continuous map
: S n+r T by sending the complement of N onto the point [S()]. This gives
a map f = T f : S n+r T r, (called the Pontryagin-Thom construction).
By an argument based on transversality, one can prove that, for r large enough, M
bounds if and only if f is homotopic to a constant map [191, Lemma IV.7 and its
proof].
Let us compose f with the map F : T r, Y of (9.9.5). By Lemma 9.9.3,

f is not homotopic to a constant map if and only if F f is not homotopic to a
constant map. As Y is a product of Eilenberg-MacLane spaces, F f is not homotopic
to a constant map if and only if H (F f) = 0. The latter implies that H T f :
H n+r (T r, ) H n+r (T ) does not vanish. Using the Thom isomorphisms, this
implies that H f : H n (Gr(r ; R )) H n (M) does not vanish. This implies that
there is a polynomial P in the Stiefel-Whitney classes of such that  P,
[M] = 0.
9.9 Thoms Theorems 431

These classes w i are the normal Stiefel-Whitney classes of M and, by Lemma 9.5.19,
the Stiefel-Whitney classes w j = w j (T M) have polynomial expressions in the w i .
Therefore, there is a polynomial P in w j such that P, [M] = 0, producing a
non-zero Stiefel-Whitney number for M.

Corollary 9.9.10 Let M and M  be two closed smooth manifolds of the same dimen-
sion. Suppose that there exists a map f : M  M such that H f is an isomorphism.
Then, M and M  are cobordant.

As a consequence of Corollary 9.9.10, a Z2 -homology sphere bounds.

Proof As H f is an isomorphism, 0 f is a bijection and then H f ([M  ]) = [M].

Let P Z2 [X 1 , . . . , X n ]. By Corollary 9.8.3, H f (w(T M)) = w(T M  ) and then
H f (PM  ) = PM . Therefore,

PM  , [M  ] = PM  , H f ([M])

= H f (PM  ), [M]
= PM , [M].

Hence, M and M  have the same Stiefel-Whitney numbers. By Theorem 9.9.6, they
are cobordant. 

For closed manifolds of dimension n, being cobordant is an equivalence relation.

The set of equivalence classes (cobordism classes) is denoted by Nn . The disjoint
union endows Nn with an abelian group structure, actually a Z2 -vector space structure
since 2M = M M is diffeomorphic to the boundary of M [0, 1]. The Cartesian

product of manifolds makes N = n Nn a Z2 -algebra, called the cobordism ring.
A development of the results of this section and the previous one enabled Thom to
compute the cobordism ring N [191, Sect. IV]; the results are summed up in the
following theorem.

Theorem 9.9.11 (1) Nn is isomorphic to lim n+k (T (k, )).

k
(2) dim Nn is the number of partitions of n into integers with none of them of the
form 2i 1.
(3) N is GrA-isomorphic to a polynomial algebra Z2 [X 2 , X 4 , X 5 , X 6 , X 8 , X 9 , . . . ]
with one generator X k for each integer k not of the form 2i 1.

A representative for X 2k is given by the cobordism class of RP 2k [191, p. 81].

Odd dimensional generator of dimension = 2i 1 were first constructed by Dold
[41]. For details and proofs (see [191] or [186, Chap. VI]).
For example, N3 = 0, confirming Example 9.9.7. Another simple consequence
of Theorem 9.9.11 is the following corollary.

Corollary 9.9.12 Let M and N be closed manifolds which are not boundaries. Then
M N is not a boundary.
432 9 Stiefel-Whitney Classes

9.10 Exercises for Chapter 9

9.1. Let be a vector bundle. Prove that is orientable. If is orientable, prove

that admits a spin structure.
9.2. Let p : X X be a smooth covering of a smooth manifold X , with an odd
number of sheets. Show that X is orientable if and only if X is orientable.
Show that X admits a spin structure if and only if X does.
9.3. Let p : M M be a 2-fold covering of a smooth connected manifold M,
with characteristic class H 1 (M). Suppose that M is orientable and that
M is not orientable. Prove that = w1 (T M).
9.4. Let M1 and M2 be two closed connected manifold of the same dimension and
let M be one of their connected sums (see p. 135). What is the total Stiefel-
Whitney class w(T (M))?
9.5. Prove that T RP 4 and T CP 2 are indecomposable as a Whitney sum of vector
bundles of smaller ranks.
9.6. Let be a vector bundle over a space X , with w( ) = 1. Show that the smallest
integer i > 0 such that wi ( ) = 0 is a power of 2.
9.7. List the critical points with their index for the weighted trace on Gr(2; R4 )
and Gr(2; R6 ). Using also Example 9.5.3, verify the statement of Lemma 9.5.7
and the second formula of Corollary 9.5.15.
9.8. Same exercise as the previous one for Fl(1, 1, 1).
9.9. Let j : B O(n) BU (n) denote the inclusion. Prove that H j (ci ) = wi2 .
9.10. Write the details for Remark 9.5.16.
9.11. Like in Example 9.5.21, find an additive basis of H (Gr(2; R5 )) in terms of
products of Stiefel-Whitney classes. Express each of these elements in terms
of Schubert symbols.
9.12. Let M be an orientable smooth closed manifold of dimension 6 or 10. Prove
that (M) is even.
9.13. Let f : P Q be a continuous map between n-dimensional connected closed
smooth manifolds. Suppose that one of the following conditions is satisfied:
(a) P is orientable while Q is non-orientable
(b) P is spin while Q is non-spin (w2 (T Q) = 0).
Then H n f : H n (Q) H n (P) is trivial.
9.14. Let f : S n+k S k be a smooth map. Let x S k be a regular value. Show that
the closed manifold f 1 ({x}) is the boundary of a (possibly non-orientable)
compact manifold.
9.15. Prove that RP 2 RP 2 and RP 4 are not cobordant.
9.16. Let M be a closed n-dimensional manifold whose cohomology ring is isomor-
phic to that of RP n . Prove that M and RP n are cobordant.
Chapter 10
Miscellaneous Applications and
Developments

This chapter, contains various applications and developments of the techniques of

mod 2 (co)homology. Most of them are somewhat original.

10.1 Actions with Scattered or Discrete Fixed Point Sets

Let X be a finite dimensional G-complex (G = {id, }) with b(X) < . By

Smith theory (Proposition 7.3.7), we know that b(X G ) b(X), which implies that
(0 (X G )) b(X). Inspired by the work of Puppe [165], we study in this section
the extremal case (0 (X G )) = b(X) (scattered fixed point set). Analogous results
for S 1 -actions are presented at the end of this section.

Proposition 10.1.1 Let X be a finite dimensional G-complex with b(X) < . Sup-
pose that (0 (X G )) = b(X). Let a HGk (X). Then Sqi (a) = (ki ) ui a.

Proof By Proposition 7.3.7, H >0 (X G ) = 0 and X is equivariantly formal. Therefore,

X G has the cohomology of b(X) points and (X G )G BG X G is homotopy equiv-
alent to a disjoint union of b(X) copies of RP . By (8.2.2), any class b HGk (X G )
satisfies Sqi (b) = (ki )ui b. As the restriction homomorphism HG (X) HG (X G ) is
injective by Proposition 7.3.9, this proves the assertion.

As seen in the above proof, the G-space X of Proposition 10.1.1 is equivariantly

formal. Thus, : HG (X) H (X) is surjective. As ker = u HG (X) by (7.1.7),
Proposition 10.1.1 has the following corollary (compare [165, Corollary 1]).

Corollary 10.1.2 Let X be a finite dimensional G-complex with b(X) < . Suppose
that (0 (X G )) = b(X). Then, any a H (X) satisfies Sq(a) = a (i.e. Sqi = 0 for
i > 0). In particular, a  a = 0 if a H >0 (X).

Let us restrict the above results to the case where X is a smooth closed G-manifold.
Then, X G is a union of closed manifolds (see, e.g. [12, Corollary 2.2.2]). We have seen
Springer International Publishing Switzerland 2014 433
J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_10
434 10 Miscellaneous Applications and Developments

in the proof of Proposition 10.1.1 that each component of X G has the cohomology of
a point. Hence X G is a discrete set of b(X) points (the smooth involution is called
an m-involution in [165]). Examples include linear spheres S0n ; if X1 and X2 are such
G-manifolds, so is X1 X2 with the diagonal involution; if dim X1 = dim X2 , the
G-equivariant connected sum X1 X2 around fixed points carries an m-involution.
Thus, an orientable surface carries an m-involution. Also, if X admits a G-invariant
Morse function, then is an m-involution by Theorem 7.6.6. Corollary 10.1.2 has
the following consequence for the Stiefel-Whitney class w(TX) of a manifold X
admitting an m-involution.

Corollary 10.1.3 Let X be a smooth closed G-manifold such that (0 (X G )) =

b(X). Then w(TX) = 1.

In consequence, a closed manifold X carrying an m-involution is orientable and

admits a spin structure. Also, X is the boundary of a (possibly non-orientable) com-
pact manifold by Thoms Theorem 9.9.6.

Proof By Corollary 10.1.2, Sqi = 0 for i > 0. Hence, the Wu class V (X) is equal
to 1. Therefore, using Wus formula 9.8.2, w(TX) = Sq(V (X)) = 1. 

We now generalize to smooth closed G-manifolds with X G discrete (without ask-

ing that XG = b(X)). This will lead us toward the link between closed G-manifolds
with discrete fixed point set and coding theory; such a link was initiated in [165] and
further developed in [121]. We start with the following lemma.

Lemma 10.1.4 Let X 2k+1 be a smooth closed G-manifold such that X G is discrete.
Then, X G is even.

Proof Let r = X G . Let W = X int D where D is a closed G-invariant tubular

neighborhood of X G . Then W is a compact free G-manifold with boundary V . The
orbit space W = W /G is a compact manifold whose boundary V = V /G is a disjoint
union of r copies of RP2k . By Proposition 5.3.9, the image B of H k (W ) H k (V )
satisfies

2 dim B = dim H k (Bd W ) = r

which shows that r is even. 

Remark 10.1.5 If X is a finite dimensional G-CW-complex with b(X) < , it is

known that b(X) b(XG ) mod 2 [9, Corollary 1.3.8]. If X is an odd dimensional
closed manifold, then b(X) is even by Poincar duality. This provides another proof
of Lemma 10.1.4

We use the notation of the proof of Lemma 10.1.4. Let ,

be the bilinear
form on H k (V ) given by a, b

= a  b, [V ]
. By Proposition 5.3.9 and its
proof, one has B, B

= 0 and r = 2 dim B. Labeling the r points of X G produces

an isomorphism H k (V ) Zr2 intertwining ,

with the standard bilinear form

10.1 Actions with Scattered or Discrete Fixed Point Sets 435

on Zr2 . Hence, in terms of coding theory (see, e.g. [46]), B is a binary self-dual linear
code on Zr2 . Choosing another labeling for the points of X G changes B by an isometry
of Zr2 obtained by coordinate permutations. The class of B modulo these isometries
thus provides an invariant of the G-action.
The self-dual code B has other descriptions. For instance, the diagram of inclusions

V
i / D

j
 
W / X

gives rise to the commutative diagram

HGk (X) / H k (W ) H k (D) / H k (V )

G G G

= (10.1.1)
  
HGk (X) / H k (W ) H k (X G ) / H k (V )
G

whose row are the Mayer-Vietoris exact sequences. Lemma 7.1.4 guarantees that
the vertical maps are isomorphisms and, together with Corollary 3.8.4, implies that
the map V  VG (X G )G is, on each component, homotopy equivalent to the
inclusion RP2k  RP . Therefore, the homomorphism HGk (X G ) H k (V ) is an
isomorphism. Hence, diagram-chasing in (10.1.1) shows that B = image(H k (W )
H k (V )) coincides with
image(HGk (X) HGk (X G ))

(pushed into H k (V )). For other descriptions of B, see [165, Sect. 2]. The following
theorem is proved in [121, Theorem 3].
Theorem 10.1.6 Every binary self-dual linear code may be obtained from a closed
smooth 3-dimensional G-manifold X with scattered fixed point set.
As in Sect. 7.3, the above results have analogues for S 1 -actions. The proofs of
Proposition 10.1.7 and Corollary 10.1.8 below are the same as for Proposition 10.1.1
and Corollary 10.1.2, replacing Propositions 7.3.7 and 7.3.9 by Propositions 7.3.12
and 7.3.14, etc. Recall that HS1 (pt) Z2 [v] with v of degree 2.

Proposition 10.1.7 Let X be a finite dimensional S 1 -complex with b(X) < and
1 0 1
X S = X S . Suppose that (0 (X S )) = b(X). Then HSodd
1 (X) = 0 and, if a HS 1 (X),
2k

then Sq2i (a) = (ki ) v i a.

Corollary 10.1.8 Let X be a finite dimensional S 1 -complex with b(X) < and
1 0 1
X S = X S . Suppose that (0 (X S )) = b(X). Then, H odd (X) = 0 and any a
H (X) satisfies Sq(a) = a.
436 10 Miscellaneous Applications and Developments

Analogously to Corollary 10.1.3, one has the following result, with the same
proof.
1 0
Corollary 10.1.9 Let X be a smooth closed S 1 -manifold such that X S = X S and
1
(0 (X S )) = b(X). Then, H odd (X) = 0 and w(TX) = 1.

In particular, the manifold X of Corollary 10.1.9 is even-dimensional. Note that

this is necessary for an S 1 -action admitting an isolated fixed point (the action,
restricted to an invariant sphere around the fixed point has discrete stabilizers, so
the sphere is odd-dimensional). The analogue of Lemma 10.1.4 is Lemma 10.1.10
below. To simplify, we restrict ourselves to semi-free actions (A -action is called
semi-free if the stabilizer of any point is either {id} or ). Incidentally, the hypothesis
1 0
X S = X S is not required.
1
Lemma 10.1.10 Let X be a smooth closed S 1 -manifold such that X S is discrete.
1
Suppose that the action is semi-free. Then, X S is even.
1
Proof As seen above, X is even-dimensional, say dim X = 2k + 2. Let r = X S . Let
W = X int D where D is a closed S 1 -invariant tubular neighborhood of X S . Then
1

W is a compact free S 1 -manifold with boundary V . The orbit space W = W /G is

then a compact manifold of dimension 2k +1 whose boundary V = V /G is a disjoint
union of r copies of CPk . By Proposition 5.3.9, the image B of H k (W ) H k (V )
satisfies

2 dim B = dim H k (Bd W ) = r

which shows that r is even. 

As for the case of an involution, Lemma 10.1.10 permits us to associate, to a

1
closed smooth semi-free S 1 -manifold with X S discrete, the self-dual linear code
1
B H k (V ) Zr2 . One can also see B as the image of HSk1 (X) in HSk1 (X S ).

10.2 Conjugation Spaces

Introduced in [87], conjugation spaces are equivariantly formal G-spaces (G =

{id, }) quite different from those with scattered fixed point sets studied in Sect. 10.1.
Here, the cohomology ring of the fixed point set most resembles that of the total
space. This similarity should be given by a cohomology frame, a notion which
we now define. We use the notations of Sect. 7.1 for a G-space X, for example
the forgetful homomorphism : HG (X) H (X) and the GrA[u]-morphism
r: HG (X) HG (X G ) H (X G )[u] induced by the inclusion X G X.
10.2 Conjugation Spaces 437

Let (X, Y ) be a G-pair. A cohomology frame or an H -frame for (X, Y ) is a pair

(, ), where
(a) : H 2 (X, Y ) H (X G , Y G ) is an additive isomorphism dividing the degrees
in half; and
(b) : H 2 (X, Y ) HG2 (X, Y ) is an additive section of the natural homomorphism
: HG (X, Y ) H (X, Y ),

satisfying, in H (X G ) H (X G )[u], the conjugation equation

r (a) = (a)um + ltm (10.2.1)

for all a H 2m (X, Y ) and all m N; in (10.2.1), ltm denotes any element in
H (X, Y )[u] which is of degree less than n in the variable u. An involution admitting
an H -frame is called a conjugation. An even cohomology pair (i.e. H odd (X, Y ) = 0)
together with a conjugation is called a conjugation pair. A G-space X is a conjugation
space if the pair (X, ) is a conjugation pair. The existence of the section implies
that a conjugation space is equivariantly formal. Note that there are examples of
G-spaces which admit pairs (, ) satisfying (a) and (b) above but none of them
satisfying the conjugation equation (see [63, Example 1]). For simplicity, we shall
mostly restrict this survey to conjugation spaces; the corresponding statements for
conjugation pairs may be found in [87].
Any space X such that H (X) = H (pt) = H (X G ) is a conjugation space. For
instance, a finite dimensional G-CW-complex X satisfying H (X) = H (pt) and
X G = is a conjugation space by Corollary 7.3.8. Another easy example is the
G-sphere Sm 2m of Example 7.1.14. Indeed, one has the following lemma.

Lemma 10.2.1 Let X be a finite dimensional G-CW-complex. Suppose that

H (X) H (S 2n ) and H n (X G ) = 0. Then X is a conjugation space.

Proof By Corollary 7.3.8, H (X G ) H (S n ). By Proposition 7.3.7, X is equi-

variantly formal and, by Proposition 7.3.9, r: HG (X) HG (X G ) is injective.
The proof of the existence of an H -frame then proceeds as in the proof of
Corollary 7.1.17. 

An other important example is the complex projective space.

Example 10.2.2 Let a H 2 (CPm ) and b H 1 (RPm ) (m ). Then, the section

: H (CPm ) HG (CPm ) of Proposition 7.1.18, together with the isomorphism
: H 2 (CPm ) H (RPm ) sending a to b makes an H -frame for the complex
conjugation on CPm . By Proposition 7.1.18, the conjugation equation takes the form

r (ak ) = ((a)u + b2 )k = (ak )uk + ltk . (10.2.2)

The same treatment may be done for HPm or OP2 (see Remark 7.1.21).
438 10 Miscellaneous Applications and Developments

These examples are actually spherical conjugation complexes, i.e. G-spaces

obtained from the empty set by countably many successive adjunctions of collec-
tions of conjugation cells. A conjugation cell (of dimension 2k) is a G-space which
2k1
is G-homeomorphic to the cone over Sk1 , i.e. to the closed disk of radius 1 in R2k ,
equipped with a linear involution with exactly k eigenvalues equal to 1. At each
step, the collection of conjugation cells consists of cells of the same dimension but,
as in [74], the adjective spherical is a warning that these dimensions do not need
to be increasing. For less standard examples of spherical conjugation complexes,
see [87, 5.3.3, p. 944].
It is proven in [87, Proposition 5.2] that a spherical conjugation complex is a
conjugation space. For example, complex flag manifolds (with being the complex
conjugation) are conjugation spaces because the Schubert cells (see Sect. 9.5.3)
are conjugation cells. This example generalizes to co-adjoint orbits of compact Lie
groups for the Chevalley involution (see [87, Sect. 8.3]) and more examples coming
from Hamiltonian geometry (see [87, Sects. 8.2 and 8.4] and [86]).
Other examples may be obtained from the previous ones since the class of conju-
gation spaces is closed under many construction, such as
direct products, with the diagonal G-action, when one of the factor is of finite
cohomology type (see [87, Proposition 4.5]).
inductive limits (see [87, Proposition 4.6]).
if (X, Y , Z) is a G-triple so that (X, Y ) and (Y , Z) are conjugation pairs, then (X, Z)
is a conjugation pair. A direct proof using H -frames is given in
[87, Proposition 4.1]; a shorter proof using the conjugation criterion of [158,
Theorem 2.3] is provided in [157, Proposition 2.2.1].
if F E B be G-equivariant bundle (with a compact Lie group as structure
group) such that F is a conjugation space and B is a spherical conjugation complex,
then E is a conjugation space (see [87, Proposition 5.3]).
We now show the naturality of H -frames under G-equivariant maps, as proven
in [87, Proposition 3.11]. Let X and Y be two conjugation spaces, with H -frames
(X , X ) and (Y , Y ). Let f : Y X be a G-equivariant map. We denote by
f G : Y G X G the restriction of f to Y G .

Proposition 10.2.3 (Naturality of H -frames) The equations HG f X = Y H f

and H f G X = Y H f hold true.

Proof Let a H 2k (X). As X and Y are sections of X : HG (X) H (X) and

Y : HG (Y ) H (Y ) respectively, one has

Y HG f X (a) = H f X X (a) = H f (a) = Y Y H f (a) . (10.2.3)

This implies that

 
HG f X (a) Y H f (a) mod ker 2k
Y : HG (Y ) H (Y ) .
2k 2k
10.2 Conjugation Spaces 439

The section produces an isomorphism HG (X) H (X)[u] and ker Y is the ideal
generated by u (see Remark 7.1.7). As H odd (Y ) = 0, we deduce that there exists
di H i (Y ) such that

HG f X (a) = Y H f (a) + Y (d2k2 ) u2 + + Y (d0 ) u2k . (10.2.4)

Let us apply rY to both sides of (10.2.4). For the left hand side, we get

rY HG f X (a) = HG f G rX X (a)
= HG f G (X (a) uk + ltk ) by the conjugation equation
= HG f G (X (a)) uk + ltk HG f G being a GrA[u]-morphism.
(10.2.5)

But, using the right hand side of (10.2.4), we get

rY HG f X (a) = Y (d0 )u2k + lt2k . (10.2.6)

Comparing (10.2.5) with (10.2.6) and using that Y is injective implies that d0 = 0.
Then, (10.2.6) may be replaced by

rY HG f X (a) = Y (d2 ) u2k2 + lt2k2 . (10.2.7)

Again, the comparison with (10.2.5) implies that d2 = 0. This process may be
continued, eventually giving that HG f X (a) = Y H f (a). Applying rY to the
right hand member of this equation gives

rY Y H f (a) = Y H f (a) uk + ltk (10.2.8)

by the conjugation equation. Comparing the leading terms of (10.2.8) and (10.2.5)
gives that H f G X (a) = Y H f (a). 

Applying Proposition 10.2.3 to X = Y and f = id, we get the following corollary.

Corollary 10.2.4 (Uniqueness of H -frames) Let (, ) and ( ,  ) be two H -

frames for the conjugation space X. Then (, ) = ( ,  ).

We can thus speak about the H -frame of a conjugation space.

Proposition 10.2.5 Let (, ) be the H -frame of a conjugation space X. Then
and are multiplicative.

Proof Let a H 2m (X) and b H 2n (X) One has a  b = (a  b) and

((a)  (b)) = ((a))  ((b)) = a  b. Hence, (a)  (b) is congruent
to (a  b) modulo ker . The same proof as for Proposition 10.2.3 then proves the
proposition (details may be found in [87, Theorem 3.3]). 
440 10 Miscellaneous Applications and Developments

Much more difficult to prove, the following result was established in

[63, Theorem 1.3].
Proposition 10.2.6 Let (, ) be the H -frame of a conjugation space X. Then
Sqi = Sq2i for all integer i.
Remark 10.2.7 It is not true in general that Sq = Sq . For example, consider
the conjugation space CPm for 1 m , with the notations of Example 10.2.2.
Of course, Sq1 (a) = 0 and then Sq1 (a) = 0. On the other hand,

r Sq1 (a) = Sq1  r (a) 

= Sq1 (bu + b2 ) by the conjugation Eq. (10.2.2)
= Sq1 (b)u + bSq1 (u) by the Cartan formula
= b2 u + bu2
= r((a) u) .

Since r is injective (see Lemma 10.2.8 below), this proves that

Sq1 ((a)) = (a) u .

The following lemma is recopied with its short proof from [87, Lemma 3.8].
Lemma 10.2.8 Let X be a conjugation space. Then the restriction homomorphism
r: HG (X) HG (X G ) is injective.

Proof Suppose that r is not injective. Let 0 = x = (y)uk + ltk HG2n+k (X) be an
element in ker r. The conjugation equation guarantees that k = 0. We may assume
that k is minimal. By the conjugation equation again, we have 0 = r(x) = (y)un+k +
ltn+k . Since is an isomorphism, we get y = 0, which is a contradiction. 
The H -frame of a conjugation space behaves well with respect to the character-
istic classes of G-conjugate-equivariant bundles. A G -conjugate-equivariant bundle
p
over a G-space X (with an involution ) is a complex vector bundle = (E X),
together with an involution on E such that p = p and is conjugate-linear
on each fiber: ( x) = (x) for all C and x E. This was called a real
p
bundle by Atiyah [11]. Note that G = (E G X G ) is a real vector bundle and
rank R G = rank C . The following result is proven in [87, Proposition 6.8].
Proposition 10.2.9 Let be a G-conjugate-equivariant bundle over a spherical
conjugation complex X. Then (c()) = w( ).
A theory of (integral) equivariant Chern classes for G-conjugate-equivariant bun-
dles over a conjugation space is developed in [160].
Another relationship between conjugation spaces and the Steenrod squares was
discovered by Franz and Puppe in [63]. It is illustrated by the case of CPm , with the
notations of Example 10.2.2, where the conjugation equation (10.2.2) for CPm may
be written as follows.
10.2 Conjugation Spaces 441


k 
k
r (ak ) = ((a)u + (a)2 )k = (kj )(a)k+j ukj = Sqj ((ak ))ukj .
j=0 j=0
(10.2.9)

It was proven in [63, Theorem 1.1] that (10.2.9) holds true in general, leading to the
following universal conjugation equation.
Theorem 10.2.10 Let X be a conjugation space, with H -frame (, ). Then, for
x H 2k (X), one has


k
r (x) = Sqj ((x))ukj . (10.2.10)
j=0

Note the resemblance between the right member of (10.2.10) and that of (8.3.9).
For other occurrences of such an expression, see [125, Sect. 2.4].
Let r : H (X) H (X G ) be the restriction homomorphism in non-equivariant
cohomology. The following corollary was observed in [63, Corollary 1.2].
Corollary 10.2.11 For x H (X), one has r (x) = (x)  (x).
Proof Suppose that x H 2k (X). Denote by G : HG (X G ) H (X G ) the forgetful
homomorphism for X G . Then

r (x) = r (x) since = id

= G r (X) using (7.1.5)
 k 
= G j=0 Sq ((x))u
j kj by Theorem 10.2.10
= Sqk ((x)) since G = evu=0 , see (7.1.6)
= (x)  (x) . 

Another consequence of Theorem 10.2.10 is the commutativity of the diagram.

/ H (X) r / H (X G ) o / H (X G )[u]
H (X) G G

ev1 (10.2.11)
 
Sq
H (X) / H (X G )

Theorem 10.2.10 has also consequences for conjugation manifolds, i.e. closed
manifolds X with a smooth conjugation . Then, X G is a closed manifold (see, e.g.

[12, Corollary 2.2.2]) whose dimension, because of the isomorphism : H (X)
H (X G ), is half of the dimension of X. By looking at the derivative of around a fixed
point, one checks that preserves the orientation if and only if dim X 0 mod 4.
For various properties of conjugation manifolds, see [80, Sect. 2.7], from which we
extract the following results (see also [160, Appendix A]).
442 10 Miscellaneous Applications and Developments

Proposition 10.2.12 Let X be a smooth conjugation manifold of dimension 2n, with

H -frame (, ). Then preserves the Wu and Stiefel-Whitney classes:

(v(X)) = v(X G ) and (w(TX)) = w(TX G ) .

Proof The Wu class v2i (X) is characterized by the equation

v2i (X)  a = Sq2i (a) for all a H 2n2i (X) . (10.2.12)

Applying the ring isomorphism to (10.2.12) and using Proposition 10.2.6 gives

(v2i (X))  (a) = Sqi ((a)) for all a H 2n2i (X) . (10.2.13)

As is bijective, (10.2.13) implies that

(v2i (X))  b = Sqi (b) for all b H ni (X G ) ,

which implies that (v2i (X)) = vi (X G ), and, as H odd (X) = 0, that (v(X)) = v(X G ).
Using this and the Wu formula, one gets

(w(TX)) = Sq(v(X)) by the Wu formula

= Sq (v(X)) Proposition 10.2.6
= Sq(v(X G ))) as already seen
= w(TX G ) by the Wu formula. 

In particular, X admits a spin structure if and only if X G is orientable. Also, the

Stiefel-Whitney numbers of X all vanish if and only if those of X G do so. By Thoms
Theorem 9.9.6, this gives the following
Corollary 10.2.13 Let X be a conjugation manifold. Then X bounds a compact
manifold if and only if X G does so.
Two natural problems occur for conjugation manifolds.
(i) Given a closed connected smooth manifold M n , does there exist a conjugation
2n -manifold X with X G diffeomorphic to M ?
(ii) Classify, up to G -diffeomorphism, conjugation manifolds with a given fixed point
set.
The circle is the fixed point set of a unique conjugation 2-manifold, namely
S12 ; the uniqueness may be proved using the Schoenflies theorem (compare [34,
Theorem 4.1]).
For n = 2, recall that RP2 is the fixed point set of the conjugation manifold
CP2 and S 1 S 1 is that of S 2 S 2 . The equivariant connected sum (around a
fixed point) of conjugation manifolds being again a conjugation manifold (see [87,
Proposition 4.7]), any closed surface is the fixed point set of some conjugation 4-
manifold (of course, S 2 = (S24 )G ). Answering Question (ii) is the main object of [80],
10.2 Conjugation Spaces 443

using the following idea. For a smooth G-action on a manifold X with X G being of
codimension 2, the quotient space X/G inherits a canonical smooth structure. If X
is a conjugation 4-manifold, then H (X/G) H (S 4 ). Conversely, let (Y , ) be a
manifold pair such that Y is a 4-dimensional Z2 -homology sphere containing  as
a codimension 2 closed submanifold. By Alexander duality (Theorem 5.3.14), one
has H 1 (Y ) = Z2 . Thus, Y  admits a unique non-trivial 2-fold covering (see
Sect. 4.3); the latter extends to a unique branched covering X Y , with branched
locus , and X turns out to be a conjugation 4-manifold with X G = M. The final
statement is thus the following ([80, Theorem A]).
Theorem 10.2.14 The correspondence X  (X/G, X G ) defines a bijection
between
(a) the orientation-preserving G-diffeomorphism classes of oriented connected con-
jugation 4-manifolds, and
(b) the orientation-preserving G-diffeomorphism classes of smooth manifold pairs
(Y , ), where Y is an oriented 4-dimensional homology sphere and  is a closed
connected surface embedded in M.
The conjugation sphere S24 corresponds to the trivial knot S 2 S 4 . Under the
bijection of Theorem A, any knot S 2  S 4 corresponds to a conjugation 4-manifold
X with X G S 2 . In general, X is not simply connected. On the other hand, Gor-
don [69, 70] and Sumners [189] found infinitely many topologically distinct knots
in S 4 which are the fixed point set of smooth involutions. These examples pro-
duce infinitely many topologically inequivalent smooth conjugations on S 4 (see [80,
Proposition 5.12]).
If X is a simply-connected conjugation 4-manifold, it is known that X/G is at least
homeomorphic to S 4 (see [80, Proposition 5.3]). In addition, X is homeomorphic (not
2
necessarily equivariantly) to a connected sum of copies of S 2 S 2 , CP2 , and CP
(see [80, Proposition 2.17]). These are severe restrictions on a simply-connected
closed smooth 4-manifold to carry a smooth conjugation.
Olbermann, in his thesis [157], was the first to address Question (i); he proved
the following result (see [158, Theorem 1.2]).
Theorem 10.2.15 Any closed smooth orientable 3-manifold is diffeomorphic to the
fixed point-set of a conjugation 6-manifold.
The case of non-orientable 3-manifolds is not known. Any 3-dimensional Z2 -
homology sphere is the fixed point of infinitely many inequivalent conjugations
on S 6 , [159]; this gives a partial answer to Question (ii) in this case.
Remark 10.2.16 As observed by W. Pitsch and J. Scherer, the answer to Question (i)
is not always positive. For example, the octonionic projective plane OP2 , which
is a smooth closed 16-manifold (see Remark 6.1.8), is not the fixed point set of
any conjugation space. Indeed, H (OP2 ) Z2 [x]/(x 3 ) by Proposition 6.1.7, with
degree (x) = 8, but, by Theorem 8.6.5, Z2 [x]/(x 3 ) is not the cohomology ring of a
topological space if degree (x) > 8.
444 10 Miscellaneous Applications and Developments

10.3 Chain and Polygon Spaces

Chain and polygon spaces are examples of configuration spaces, a main concept of
classical mechanics. In recent decades, starting in [83, 202] (inspired by talks of
Thurston on linkages [195], new interests arose for polygon spaces, in connections
with Hamiltonian geometry (see e.g. [88, 91, 111, 118]), mathematical robotics [56,
142] and statistical shape theory [90]. This section contains original results on the
equivariant cohomology of chain spaces, giving new proofs for known statements
about their ordinary cohomology.
We use the notations of [59, 85, 90], inspired by those of statistical shape theory
[112]. In order to make some formulae more readable, we may write |J| for the
cardinality J of a finite set J.

10.3.1 Definitions and Basic Properties

Let  = (1 , . . . , n ) Rn>0 and let d be an integer. We define the subspace Cdn ()
of (S d1 )n1 by

 
n1

Cdn () = z = (z1 , . . . , zn1 ) (S d1 )n1 | i zi = n e1 ,
i=1

where e1 = (1, 0, . . . , 0) is the first vector of the standard basis e1 , . . . , ed of Rd .

An element of Cdn (), called a chain, may be visualized as a configuration of (n 1)
successive segments in Rd , of length 1 , . . . , n1 , joining the origin to n e1 . The
vector  is called the length vector. The chain space Cdn () is contained in the big
chain space BCdn () defines as follows:

 n1

BCdn () = z = (z1 , . . . , zn1 ) (S d1 )n1 | i zi , e1
= n ,
i=1

(successions of (n 1) segments in Rd , of length 1 , . . . , n1 , joining the origin to

the affine hyperplane with first coordinate n ).
The group O(d 1), viewed as the subgroup of O(d) stabilizing the first  axis, acts
naturally (on the left) upon the pair (BCdn (), Cdn ()). The quotient Cdn () SO(d 1)
is the polygon space Ndn , also defined as

Ndn () = N dn () SO(d) ,
10.3 Chain and Polygon Spaces 445

where

 

n
d1 n

Nd () = z (S )

n
i zi = 0
i=1

is the free polygon space (called space of polygons in [57, 65]).

The map from SO(d) Cdn () to N dn () given by

(A, (z1 , . . . , zn1 ))  (Az1 , . . . , Azn1 , Ae1 )

descends to an SO(d)-homeomorphism

N dn () .
SO(d) SO(d1) Cdn () (10.3.1)

Recall that the map SO(d) S d1 given by A  Ae1 is the orthonormal oriented
frame bundle for the tangent bundle to S d1 (see p. 362). Thus, by (10.3.1), we get
a locally trivial bundle

Cdn () N dn () S d1 . (10.3.2)

When d = 2 the space of chains C2n () coincides with the polygon space N2n ().
The axial involution on Rd = R Rd1 given by (t, y) = (t, y) induces
an involution, still called , on the pair (BCdn (), Cdn ()) and on (S d1 )n1 . As
commutes with the O(d 1)-action on Cdn (), it descends to a G-action on Ndn (),
where G = {id, }. A bar above a G-space denotes its orbit space:

BC d () = BCdn ()/G , Cdn () = Cdn ()/G , N dn () = Ndn ()/G .
n

We shall compute the G-equivariant cohomology of BCdn () and Cdn (), as algebras
over HG (pt) = Z2 [u] (u of degree 1). This uses some G-invariant Morse theory on
M = (S d1 )n1 . We start with the robot arm map F : M Rd defined by


n1
F (z) = i zi , z = (z1 , . . . , zn1 ) . (10.3.3)
i=1

Consider Rd as the product R Rd1 , which defines the projections p1 : Rd R

and pd1 : Rd Rd1 . Define f : M R by


n1
f (z) = p1 (F (z)) = i zi , e1
.
i=1
446 10 Miscellaneous Applications and Developments

Note that F is O(d)-equivariant and f is O(d 1)-invariant. For n = 2, it is clear

that f is Morse function on S d1 , with two critical points, namely e1 of index 0 and
e1 of index d 1. The following lemma follows easily.

Lemma 10.3.1 The function f : (S d1 )n1 R defined by


n1
f (z1 , . . . , zn1 ) = i zi , e1

i=1

is a G-invariant Morse function, with one critical point PJ for each J {1, . . . , n1},
where PJ = (z1 , . . . , zn1 ) with zi equal to e1 if i J and e1 otherwise (a collinear
chain). The index of PJ is (d 1) |J|.

A length vector  Rn>0 is generic if C1n () = , that is to say there are no
collinear chains or polygons. In this section, we shall only deal with generic length
vectors.

Corollary 10.3.2 If  is a generic length vector, then BCdn (), Cdn () and N dn () are
smooth closed orientable manifolds of dimension

dim BCdn () = dim N dn () = (n1)(d 1)1 and dim Cdn () = (n2)(d 1)1 .

Proof If  is generic, then  is a regular value of f . Indeed, if pd1 (F (z)) = 0,

this follows from the O(d)-equivariance of F . If pd1 (F (z)) = 0, then, as  is
generic, z is not a critical point of F (these are the collinear configurations zi = zj :
see [83, Theorem 3.1]). Since BCdn () = f1 (n ), this proves the assertion on
BCdn () (which is orientable, having trivial normal bundle in the orientable manifold
(S d1 )n1 ).
Define P: BCdn () Rd1 by P(z) = pd1 (F (z)). As seen above, as  is generic,
1
P (0) contains no critical points of F . Therefore, P is transversal to 0 and thus
Cdn () = P1 (0) is a closed submanifold of codimension d 1 of BCdn (), with trivial
normal bundle.
When  is generic, the O(d 1)-action on Cdn () is smooth. Hence, the bun-
dle (10.3.2) is a smooth bundle and the assertion on N dn () follows form (10.3.1).
For another proof that N dn () is a manifold, see [57, Proposition 3.1]. 

We now see how chain and polygon spaces are determined by some combinatorics
of their length vector
  = (1 , . . . , n ). A subset J of {1, . . . , n} is called -short
(or just short) if iJ i < iJ / i . The complement of a short subset is called long.
If  is generic, subsets are either short or long. Short subsets form, with the inclusion,
a poset Sh(). Define Shn () = {J {1, . . . , n 1} | J {n} Sh()} .
For J {1, . . . , n}, let HJ be the hyperplane (wall) of Rn defined by

  

HJ := (1 , . . . , n ) Rn
i = i .
iJ iJ
/
10.3 Chain and Polygon Spaces 447

The union H(Rn ) of all these walls determines a set of open chambers in (R>0 )n
whose union is the set of generic length vectors (a chamber is a connected component
of (R>0 )n H(Rn )). We denote by Ch() the chamber of a generic length vector .
Note that Ch() = Ch( ) if and only if Sh() = Sh( ).
Let Symn be the group of bijections of {1, . . . , n}; we see Symn1 as the sub-
group of Symn formed by those bijections fixing n. If X is a set, the group Symn
acts on the Cartesian product X n by

(x1 , . . . , xn ) = (x(1) , . . . , x(n) ) .

The notation emphasizes that this action is on the right: an element x X n is formally
a map x : {1, . . . , n} X (xi = x(i)) and Symn acts by pre-composition, i.e.
x = x . We shall use this action on various n-tuples, in particular on length vectors.

Lemma 10.3.3 Let  = (1 , . . . , n ) and  = (1 , . . . , n ) be two generic length
vectors. Then, the following conditions are equivalent
(1) Shn () and Shn ( ) are poset isomorphic.
(2) Sh() and Sh( ) are poset isomorphic via a bijection Symn1 .
(3) Ch( ) = Ch( ) for some Symn1 .
Moreover, if one of the above conditions is satisfied, there are O(d 1)-diffeomor-
phisms of manifolds pairs

(BCdn ( ), Cdn ( ))
h: (BCdn (), Cdn ())

and

(BC nd (), Cn ())
h: (BC d ( ), Cdn ( )) .
n
d

Proof Implications (1) (2) (3) are obvious. Let us prove that (1) (2). Let
Symn1 be the permutation giving the poset isomorphism Shn () Shn ( ).
Replacing  by  , we may assume that Shn () = Shn ( ). We now observe that
Shn () determines Sh(). Indeed, let J {1, . . . , n}. Then, either n J or n J,

and thus Shn () tells us whether J Sh() (or J Sh()).
It remains to prove that (3) implies the existence of the O(d1)-diffeomorphisms h
As the G action commutes with the O(d 1)-action (G is naturally in the center
and h.
If Sym
of O(d 1)), it suffices to construct h, which will induce h. n1 , then the

correspondence z z defines an O(d 1)-diffeomorphism (BCdn (), Cdn ())
(BCdn ( ), Cdn ( )). Replacing  by  , we may thus assume that Ch() = Ch( )
and = id.
Consider the smooth map L: (Rd {0})n1 (R>0 )n given, for x =
(x1 , . . . , xn1 ) by
 

L(x) = |x1 |, . . . , |xn1 |, F(x), e1
,
448 10 Miscellaneous Applications and Developments

n1

where F(x) = xi . Observe that the map (S d1 )n1 (Rd {0})n1 sending
i=1

(z1 , . . . , zn1 ) to (1 z1 , . . . , n1 zn1 ) induces a diffeomorphism  : BCdn ()
L 1 () such that F = F, the robot arm map of (10.3.3). If  is generic, then
 is a regular value of L. Indeed, let x = (x1 , . . . , xn1 ) L 1 (). For each i =
1, . . . , n 1, one can construct a path x i (t) = (x1i (t), . . . , xn1
i (t)) (Rd {0})n1
with x(0) = x such that L(x (t)) = (1 (t), . . . , n (t)) satisfies j (t) = j for j = i and
i i i i

ii (t) = i +t with = 0. For i = n, this follows from the proof of Corollary  10.3.2.
Suppose that i n 1. If x is not a lined configuration, then xi and j=i xj are
linearly independent and generate a 2-dimensional plane , containing F(x). There
are rotations it and t of , depending smoothly on t, such that it ((1 + t)xi ) +


t ( j=i xj ) = F(x). Hence, X i (t) may be defined as

it ((1 + t)xi ) if j = i
xji (t) =
t (xj ) if j = i .

Finally, if x is a lined configuration, then F(x) and e1 are linearly independent (since
 is generic). They thus generate a 2-dimensional plane . Let x i (t) defined by
x ii (t) = (1 + t)xi and x ji (t) = xj when j = i. If t is small enough, there is a unique
rotation rt of such that rt (F( x i (t)), e1
= n . We can thus define x i (t) = t (x i (t)).
j j
As Ch() is convex, it contains the segment [,  ] consisting of only generic length
vectors. What has been done above shows that the map L is transversal to [,  ].
Therefore, X = L 1 ([,  ]) is O(d)-cobordism between BCdn () and BCdn ( ). Let
pd1 : Rd Rd1 be the projection onto the last d 1 coordinates. As in the
proof of Corollary 10.3.2, the map P: X Rd1 defined by P(x) = pd1 (F(x)) is
transversal to 0. Thus, Y = P1 (0) is a submanifold of X of codimension n1 and the
pair (X, Y ) is a cobordism of pairs between (BCdn (), Cdn ()) and (BCdn ( ), Cdn ( )).
The map L: X [,  ] has no critical point. The standard Riemannian metric
on (Rd )n induces an O(d 1)-invariant Riemannian metric on (X, Y ). Following
the gradient lines of for this metric provides the required O(d 1)-equivariant
diffeomorphism h. 

For n 9, a list of all chambers (modulo the action of Symn ) was obtained
in [90]. Their numbers are as follows (for n = 10, it was computed independently
by Minfeng Wang and Dirk Schtz: see the Web complement of [90]).

n 345 6 7 8 9 10
Nb of chambers 2 3 7 21 135 2470 175428 52980624

Geometric descriptions of several chain and polygon spaces for -generic are
provided in [85], as well as some general constructions. Among them, the operation
of adding a tiny edge, which we now describe. Let  = (2 , . . . , n ) be a generic
10.3 Chain and Polygon Spaces 449

length vector. If > 0 is small enough, the n-tuple + := (, 2 , . . . , n ) is a generic

length vector for 0 < .
Lemma 10.3.4 There are O(d 1)-equivariant diffeomorphisms

B
: BCdn (+ ) S d1 BCdn1 () and
: Cdn (+ ) S d1 Cdn1 () ,

where S d1 BCd1
m () and S d1 C m () are equipped with the diagonal O(d1) -
d1
action.
Proof The diffeomorphism
is constructed in [85, Proposition 2.1]. The construc-
tion can be easily adapted to give B
. 

10.3.2 Equivariant Cohomology

Let M = (S d1 )n1 . The G-invariant Morse function f = f : M R of

Lemma 10.3.1 satisfies the hypotheses of Proposition 7.6.13, i.e. M G = Crit f .
Therefore, M is G-equivariantly formal and the restriction morphism

r: HG (M) HG (M G ) HG (PJ ) Z2 [uJ ] (10.3.4)
J J

is injective (this also follows from Lemma 7.3.6 and Proposition 7.3.9). The variables

one and the Z2 [u]-module structure on HG (M ) is given by the
uJ are of degree G

inclusion u  J uJ .
 xi (i N) are formal variables in a
In the remainder of this section, whenever
polynomial ring and J N, we set xJ = jJ xj . In particular, x = 1.
Proposition 10.3.5 For n 2, there is a GrA[u]-isomorphism
 d1 n1
Z2 [u, A1 . . . , An1 , B1 , . . . , Bn1 ] I
HG ((S ) ) (10.3.5)

where the variables Ai and Bi are of degree d 1 and I is the ideal generated by
the families of relators
(a) Ai + Bi + ud1 i = 1, . . . , n 1
(b) A2i + Ai ud1 i = 1, . . . , n 1

Moreover, using (10.3.4), one has for J {1, 2, . . . , n 1}:

 |J|(d1)
r(AJ ) = uK
JK
 |J|(d1)
(10.3.6)
r(BJ ) = uK
JK=
450 10 Miscellaneous Applications and Developments

Proposition 10.3.5 generalizes Examples 7.6.9 and 7.4.5, with the slightly different
notations of (10.3.4) for the equivariant cohomology of the fixed point set.

Proof The proof proceeds by induction on n. It starts with n = 2, using Example

M0 , where M
7.6.9. For the induction step, set M = (S d1 )n1 = M = (S d1 )n2
and M0 = S . The induction hypothesis implies that
d1


Z2 [u, A 1 . . . , A n2 , B 1 , . . . , B n2 ] I
HG (M)

where I is the ideal generated by the families A i + B i + u d1 and A 2i + A i u d1

(i = 1, . . . , n 2). The Z2 [u]-module structure is obtained by identifying u with u .
Also,

HG (M0 ) Z2 [u0 , A, B] (A + B + u0d1 , A2 = u0d1 A)

and the Z2 [u]-module structure is obtained by identifying u with u0 . By Theorem

7.4.3, the strong equivariant cross product provides an isomorphism

Z [u] HG (M0 )
G : HG (M) HG (M) .
2

Setting Ai = A i G 1, Bi = B i
G 1 (i = 1, . . . , n 2), An1 = 1
G A and

Bn1 = 1 G B gives the induction step for the isomorphism (10.3.5).
We now prove the induction step for (10.3.6). The fixed points of M are denoted

by PJ , indexed by J {1, . . . , n 2}. We denote the fixed point of M0 = S d1
R Rd1 by min = (1, 0) and max = (1, 0) (corresponding to the extrema
of the Morse function (t, x)  t). Set HG (M0G ) Z2 [umin ] Z2 [umax ]. For
J {1, . . . , n 2}, then PJ = P J min and PJ{n1} = P J max . Hence, for
i = 1, . . . , n 2, one has, using the obvious notations, that

r(Ai ) = r(A i G 1)

= r (Ai ) G r0 (1) G
by naturality of
  |J|(d1) 
= uJ G [1min + 1max ] by induction hypothesis

J{1,...,n2}
 
iJ
|J|(d1) |J|(d1)
= uJ + uJ (J {1, . . . , n 2})
J J{n1}

iJ iJ
|J|(d1)
= uJ .
J{1,...,n1}
iJ
10.3 Chain and Polygon Spaces 451

As for i = n 1, one has

r(An1 ) = r(1 G A)
= r (1) G r0 (A)
  
= 1J G umax
d1

J{1,...,n2}

d1
= uJ{n1}
J{1,...,n2}
 |J|(d1)
= uJ .
J{1,...,n1}
n1J

This proves (10.3.6) for r(Ai ), i = 1, . . . , n 1. The formula for r(Bi ) are
deduced using relators (a). The formulae for r(AJ ) and r(BJ ) follow since r is mul-
tiplicative. 

We are now ready to compute the G-equivariant cohomology of BCdn ().

Theorem 10.3.6 Let  = (1 , . . . , n ) be a generic length vector. There is a GrA[u]-

isomorphism
 n
HG (BCdn ()) H (BC d ())
Z2 [u, A1 . . . , An1 , B1 , . . . , Bn1 ] I

where the variables Ai and Bi are of degree d 1 and I is the ideal generated by
the families of relators

(a) Ai + Bi + ud1 i = 1, . . . , n 1
(b) A2i + Ai ud1 i = 1, . . . , n 1
(c) AJ J {1, . . . , n 1} and J {n} is long
(d) BJ J {1, . . . , n 1} and J is long.

Proof Let M = (S d1 )n1 , M = f 1 ((, n ]) and M+ = f 1 ([n , ]),

with the inclusions j : M M (f = f , the Morse function of Lemma 10.3.1).
One has M M+ = B = BCdn () = f 1 (n ). The G-invariant Morse function
f : M R satisfies the hypotheses of Proposition 7.6.13. The latter implies that the
morphism HG (M) HG (B) induced by the inclusion is surjective with kernel equal
to ker HG j + ker HG j+ . By Proposition  10.3.5, HG (M) is GrA[u]-isomorphic to
Z2 [u, A1 . . . , An1 , B1 , . . . , Bn1 ] I where I is the ideal generated by families (a)
and (b). We shall prove that ker HG j is the ideal generated by relators (c) and that
ker HG j+ is the ideal generated by relators (d).
The critical point PJ satisfies
 
f (PJ ) = i i .
iJ iJ
/
452 10 Miscellaneous Applications and Developments

Therefore,

PJ M f (PJ ) < n J {n}is short . (10.3.7)

Therefore, one has a commutative diagram

HG (M) / r / H (M G ) / Z2 [uJ ]
G
J{1,...,n1}

HG j HG j
G pr (10.3.8)

 
r Z2 [uJ ]
HG (M ) / / H (M G )
G
/
J{1,...,n1}
J{n} short

That r and r are injective follows from Theorem 7.6.6. Hence, for x HG (M),
H j (AJ ) = 0 if and only if pr r(x) = 0. Since M is equivariantly formal (by
Theorem 7.6.6 again), Theorem 10.3.6 implies that

H (M) HG (M)/(u) Z2 [A1 , . . . , An1 ]/(A2i ) .

By the Leray-Hirsch theorem, HG (M)/(u) is then isomorphic to the free Z2 [u]-

module with basis {Aj | J {1, . . . , n 1}} (or {Bj | J {1, . . . , n 1}}). Thus,
x HG (M) may be uniquely written as x = J{1,...,n1} J AJ , with J Z2 [u].
Let J0 {1, . . . , n 1} minimal (for the inclusion) such that J0 = 0. By (10.3.7),
one has

r(x) = J0 uJ0 mod Z2 [uJ ] .
J{1,...,n1}
J =J0

Hence, if x ker HG j , we deduce using Diagram (10.3.8) that J0 {n} is long.

Therefore, J0 AJ0 ker HG j and x + J0 AJ0 ker HG j . Repeating the above
argument with x + J0 and so on proves that

x= J AJ ker HG j J = 0 whenever J {n} is short .
J{1,...,n1}

This proves that ker HG j is the Z2 [u]-module generated by relators (c) (since
AJ AK = AJK , this is an ideal).
In the same way, we prove that ker HG j+ is the Z2 [u]-module generated by relators
(d). Details are left to the reader. 
10.3 Chain and Polygon Spaces 453

Corollary 10.3.7 For a generic length vector  = (1 , . . . , n ), there is a GrA[u]-

isomorphism

Z2 [u, A1 . . . , An1 ] I 
HG (BCdn ())

where the variables Ai are of degree d 1 and I  is the ideal generated by the
families of relators

(1) A2i + Ai ud1 i = 1, . . . , n 1

(2) AJ  J {1, . . . , n 1} and J {n} is long
(3) ud1 AK u(|JK|1)(d1) J {1, . . . , n 1} and J is long
KJ

Note that, by (2), only the sets K J with K {n} being short occur in the sum
of Relators (3).

Proof This presentation of HG (BCdn ()) is algebraically deduced from that of

Theorem 10.3.6. The generators Bi are eliminated using relators (a). Realtors (b)
and (c) become respectively (1) and (2). Relators (d) become relators (3). Indeed,

BJ = iJ Bi
= iJ (Ai + ud1 ) using (a)
= KJ AK u|JK|(d1) plain extension

=u d1 AK u (|JK|1)(d1) as AJ = 0 since J is long. 
KJ

Example 10.3.8 Elementary geometry easily shows that BCdn () = if and only if
{n} is long (see also Example 10.3.23 below). In Corollary 10.3.7, we see that if {n}
is long, then relator (2) for J = implies that 1 I  and thus HG (BCdn ()) = 0.
Compare Example 10.3.21.

Example 10.3.9 Suppose that  is generic and that n = + n1 i=1 i , with > 0
small enough such that J {n} is short only for J = . Hence, relators (2) imply that
Ai = 0 for i = 1, . . . , n 1. The only subset J of {1, . . . , n 1} which is long is
{1, . . . , n 1} itself. Thus, the family of relators (3) contains one element, in which
the only non-zero term in the sum occurs for K = . This relator has thus the form
u(n1)(d1) and we get

HG (BCdn ()) Z2 [u] (u(n1)(d1) ) . (10.3.9)

Notice that, with our hypothesis, n is a regular value of f which is just above a
minimum. Thus, BCdn () = f 1 () is G-diffeomorphic to the sphere S (n1)(d1)1
endowed with the antipodal involution (the isotropy representation of G on the tangent
space to M at the minimum P of f ). As the G-action on BCdn () is free, one has
454 10 Miscellaneous Applications and Developments

n
HG (BCdn ()) H (BC d ()/G) H (RP(n1)(d1)1 )

which is coherent with (10.3.9).

Example 10.3.9 will help us to compute HG (Cdn ()), after introducing some pre-
liminary material. We use the robot arm map F : M = (S d1 )n1 Rd = RRd1
defined in (10.3.3). Let N = F1 (R 0). If  = (1 , . . . , n1 ) is itself generic,
then N is a closed submanifold of codimension d 1 in M. Indeed, except at F1 (0),
the robot arm map is clearly transverse to R 0 (use that F is SO(d)-equivariant).
If  is generic, then 0 is a regular value of F (see the proof of Corollary 10.3.2).
Hence, F is everywhere transversal to R 0.
A slight change of e.g. 1 (which does not change the G-diffeomorphism type
of the pair (BCdn (), Cdn ()) by Lemma 10.3.3) will make  is generic. Hence, with-
out loss of generality, one may assume that N is a closed G-invariant submanifold
of M. One has Cdn () = N N+ where N = N M (notation of the proof of
Corollary 10.3.7).
There is a G-equivariant map t : M M such that 0 = id and 1 (M ) =
N . Indeed, for z M , denote by z the 2-plane in Rd generated by e1 and F (z).
Define t (z) SO(d) to be the rotation of angle cos1 (t|F (z)|/|f (z)|) on z , the
identity on z and such that t F (z), e1
f (z). The retraction by deformation t
is defined by t (z1 , . . . , zn1 ) = (t (z1 ), . . . , t (zn1 )). The existence of t implies
that

HG (M ) HG (N ) . (10.3.10)

The restriction of f,N : N R of f is also a G-invariant Morse function on M,

with Crit f,N = Crit f (see [83, Sect. 3]; the index of a critical point P is different
for f,N and f when f (P) < 0).

Proposition 10.3.10 Let  = (1 , . . . , n ) be a generic length vector and let

i: Cdn () BCdn () be the inclusion. Then, HG i: HG (BCdn ()) HG (Cdn ()) is
surjective, with kernel equal to Ann (ud1 ), the annihilator of ud1 .

Proof Let (B, C) = (BCdn (), Cdn ()). Consider the commutative diagram

HG (M ) / / H (B)
G

HG j HG i
 
HG (N ) / / H (C)
G

where all the arrows are induced by the inclusions. The horizontal maps are indicated
to be surjective: this follows from Proposition 7.6.13 since Crit f = M G = N G . That
HG j is an isomorphism was noticed in (10.3.10). Hence, HG i is surjective.
10.3 Chain and Polygon Spaces 455

Let B = B/G and C = C/G. As the G-action on (B, C) is free, the vertical maps
in the diagram

iG
CG / BG

 i

C / B

are homotopy equivalences (see Lemma 7.1.4). As B and C are smooth closed man-
ifolds, Proposition 5.4.5 implies that ker HG i is the annihilator of the Poincar dual
PD(C) H d1 (B) H d1 (B). It thus remains to show that PD(C) = ud1 .
G
Let pd1 : R = R R
d d1 R d1 be the projection onto the second factor.
The map : B (Rd1 )n1 {0} defined by

(z1 , . . . , zn1 ) = (pd1 (z1 ), . . . , pd1 (zn1 ))

is smooth, G-equivariant (for the involution x  x on (Rd1 )n1 ) and satisfies C =

1 ((Rd1 )n2 {0}). It thus descends to a smooth map : B RP(n1)(d1)1 ,
1
such that C = (RP (n2)(d1)1 ). As in the proof of Corollary 10.3.2, one shows
that is transversal to RP(n2)(d1)1 . By Proposition 5.4.5, one has

= H (PD(RP
PD(C) (n2)(d1)1
)) = H (u
d1 ) = ud1 .

The two occurrences of the letter u in the above formulae is a slight abuse of language,
permitted by the considerations of Lemma 7.1.4: the G-action under consideration
are all free and in the commutative diagram

/ H ((Rd1 )n1 {0})

H (RP(n1)(d1)1 ) G

H HG
 
/ H (B)

H (B) G

the generator of H 1 (RP(n1)(d1)1 ) is sent to u. 

We are now ready to compute HG (Cdn ()).
Theorem 10.3.11 For a generic length vector  = (1 . . . , n ), there is a GrA[u]-
isomorphism
 n
Z2 [u, A1 . . . , An1 ] I 
HG (Cd ()) H (Cdn ())

where the variables Ai are of degree d 1 and I  is the ideal generated by the
families of relators
456 10 Miscellaneous Applications and Developments

(1) A2i + Ai ud1 i = 1, . . . , n 1

(2) A 
J J {1, . . . , n 1} and J {n} is long
(3 ) AK u(|JK|1)(d1) J {1, . . . , n 1} and J is long
KJ

Proof We use the notations of the proof Theorem 10.3.6, with M = (S d1 )n1 , etc.
Recall from Proposition 10.3.5 that

HG (M) Z2 [u, A1 . . . , An1 ] I1

where I1 is the ideal generated by relators (1). Denote by I2 , I3 and I3 the ideals
of H (M) generated by, respectively, relators (2), (3) and relators (3) of Corol-
lary 10.3.7. It was shown in Theorem 10.3.6 and Corollary 10.3.7 that

J = ker(HG (M) HG (BCdn ()) = I2 + I3 . (10.3.11)

In view of Proposition 10.3.10, we have to prove that the quotient ideal

J = {x HG (M) | ud1 x J }

is equal to I2 + I3 .
That I2 + I3 J is obvious. For the reverse inclusion, let x J . By (10.3.11),
one has ud1 x = y2 + y3 for some y2 I2 and y3 I3 . As I3 = ud1 I3 , we can
write y3 = ud1 y3 with y3 I3 . Let z = y + y3 . Then ud1 z I2 . We shall prove
that z I2 .
As noticed in the proof of Theorem 10.3.6, HG (M) is the free Z[u]-module gen-
erated by AJ (J {1, . . . , n 1}). Thus, z admits a unique expression

z= J AJ ,
J{1,...,n1}

with J Z2 [u]. Hence,


ud1 z = (ud1 J )AJ . (10.3.12)
J{1,...,n1}

But, as ud1 z I2 , one has


ud1 z = J AJ . (10.3.13)
J{1,...,n1}
J{n}long

As, HG (M) is the free Z[u]-module generated by the classes AJ , one deduces
from (10.3.12) and (10.3.13) that J = 0 if J {n} is short. Thus, z I2 .
10.3 Chain and Polygon Spaces 457

The equality J = I2 +I3 may be also obtained using a partial Groebner calculus
with respect to the variable u, as presented in [89, Sect. 6]. 

Remark 10.3.12 In the case d = 2, where C2n () = N 2n (), the presentation of
H (C2n ()) of Theorem 10.3.11 was obtained in [89, Corollary 9.2], using techniques
of toric manifolds.

Example 10.3.13 It is easy to see that Cdn () = if and only {k} is long for some
k {1, . . . , n}. If k = n then relator (2) for J = implies that 1 I  and thus
HG (Cdn ()) = 0. If k < n, it is relator (3) for J = {k} which implies that 1 I  .

Example 10.3.14 Let  = (1, 1, 1, ), with < 1. The presentation of HG (BCd4 ())
given by Corollary 10.3.7 takes the form

HG (BCdn ()) Z2 [u, A1 , A2 , A3 ] I 

with I  being the ideal generated by A2i + Ai ud1 (i = 1, 2, 3), AJ for |J| = 2, and
relators (3) for J = {1, 2}, {1, 3} and {2, 3}, which are

ud1 (ud1 + A1 + A2 )
ud1 (ud1 + A2 + A3 )
ud1 (ud1 + A1 + A3 ) .

The sum of these relators equals u2(d1) which thus belongs to I  . Relator (3) for
J = {1, 2, 3} does not bring new generators for I  .
The presentation of HG (Cd4 ()) given by Theorem 10.3.11 is similar, with relators
(3) replaced by relators (3):
ud1 + A1 + A2
ud1 + A2 + A3
ud1 + A1 + A3 .

The sum of these relators being equal to ud1 , we get that the three classes Ai
HGd1 (BCd4 ()) are mapped to the same class A HGd1 (Cd4 ()). Therefore,

HG (Cd4 ()) Z2 [u, A] (ud1 , A2 ) . (10.3.14)

Note that Cd4 () is G-diffeomorphic to the unit tangent space T 1 S d1 (by
orthonormalizing (z1 , z2 )). Thus, (10.3.14) is a presentation of H (Cd4 ())
H ((T 1 S d1 )/G).

In the presentations of HG (BCdn ()) and HG (Cdn ()) given in Corollary 10.3.7 and
Theorem 10.3.11, the integer d is only used to fix the degree of the variables Ai . Here
is an application of that.
458 10 Miscellaneous Applications and Developments

Lemma 10.3.15 Let  = (1 , . . . , n ) be a generic length vector. For d 2, there

is an isomorphism of graded rings

(d1) (d1)
dBC : HG (BC2n ())
HG (BCdn ()) and dC : HG (C2n ())
HG (Cdn ())

which multiply the degrees by d 1.

Proof For an integer a 2, set Ma = (S a1 )n1 and BCa = BCan (). As Ma is
equivariantly formal, Corollary 4.7.20 applied to the bundle Ma (Ma )G RP
implies that

(1 + t a1 )n1
Pt (HG (Ma )) = Pt (Ma ) Pt (RP ) = .
1t

Hence

(d1) (1 + t d1 )n1
Pt (HG (Md )) = = Pt d1 (HG (M2 ))
1 t d1

which implies that

p p(d1)
dim HG (M2 ) = dim HG (Md ) (10.3.15)

for all p N.
By eliminating the variables Bi in the presentation of HG (Ma ) given in
Proposition 10.3.5, we get the presentation
 a 2 
HG (Ma ) Z2 [ua , Aa1 . . . , Aan1 ] (Ai ) = uaa1 Aai ,

where Aai is of degree a 1 and ua is of degree 1. Therefore, the correspon-

dences u2  udd1 and A2i  Aai define a homomorphism of graded rings
 d : H (M2 ) H (d1) (Md ), multiplying the degrees by d 1, which is clearly
G G
surjective. By (10.3.15),  d is an isomorphism.
By Corollary 10.3.7, HG (BCa ) is the quotient of HG (Ma ) by I2a + I3a , where
Ij is the ideal generated by relators (j) of Corollary 10.3.7. As 
a d (I 2 ) = I d
j j
(d1) d descends to the required isomorphism  BC . In
HG (Md ), the isomorphism  d
the same way, we construct dC using Theorem 10.3.11. 

10.3.3 Non-equivariant Cohomology

The G-cohomology computations of Corollary 10.3.7 and Theorem 10.3.11 give

some information on the non-equivariant cohomology of BCdn () and Cdn (). We start
with the big chain space.
10.3 Chain and Polygon Spaces 459

Theorem 10.3.16 Let  = (1 , . . . , n ) be a generic length vector. The Poincar

polynomial of BCdn () is
 
Pt (BCdn ()) = t |J|(d1) + t |K|(d1)1 . (10.3.16)
J{1,...,n1} K{1,...,n1}
J{n}short K long

The proof of Theorem 10.3.16 makes use the simplicial complex Sh

n () whose
simplexes are the non-empty subsets of the poset Shn ().

Proof Let BCd = BCdn (). By (7.1.8), one has a short exact sequence

tr
0 HG (BCd )/(u)
H (BCd ) Ann (u) 0 , (10.3.17)

whence

Pt (BCd ) = Pt (HG (BCd )/(u)) + Pt (Ann (u)) . (10.3.18)

The presentation of HG (BCd ) given in Corollary 10.3.7 implies that


HG (BCd )/(u) Z2 [A1 , . . . , An1 ] J (10.3.19)

where J is the ideal generated by the squares A2i of the variables and the monomials
AJ when J {n} is long. Therefore,

HG (BCd )/(u) d1 (Sh

n ()) , (10.3.20)

the face exterior algebra of the simplicial complex Sh

n () (see Sect. 4.7.8). Then,
by Corollary 4.7.52,
 

Pt (HG (BCd )/(u)) = Pt (d1 (Sh
n ())) = t (dim +1)(d1) = t |J|(d1) .
S (Sh
n ())
JShn ()
(10.3.21)

Let us assume that d 3. The graded algebra HG (BCd )/(u) is concentrated in

degrees (d 1). We claim that Ann (u) is concentrated in degrees (d 1)
1. Indeed, let us write HG (BCd ) as the quotient H (M)/J as in the proof of
Theorem 10.3.11. A class 0 = z HG (BCd ) is the image of z HG (M). As M
p

is equivariantly formal, one has uz = 0. Hence, if z Ann (u), one has 0 = uz J .

As the ideal J is concentrated in degrees (d 1), we deduce that p = q(d 1) 1.
Together with (10.3.17), this implies that

H (d1) (BCd ) HG (BC)/(u) , H (d1)1 (BCd ) Ann (u) (10.3.22)

460 10 Miscellaneous Applications and Developments

and H (BCd ) vanishes in other degrees. Since dim BCd = (n 1)(d 1) 1,

Poincar duality gives the formula
 
Pt (BCd ) = t |J|(d1) + t (n1|J|)(d1)1 . (10.3.23)
JShn () JShn ()

Using (10.3.17), we thus get, for d 3, that

 
Pt (Ann (u)) = t ((n1|J|)(d1)1)(d1) = t |K|(d1)1 ,
JShn () K{1,...,n1}
K long
(10.3.24)
where the last equality is obtained by re-indexing the sum with K = {1, . . . , n1}J.
It remains to prove that (10.3.24) is also valid when d = 2.
Let us fix some integer d 3. By Lemma 10.3.15 and its proof, there is an

isomorphism of graded rings dBC : H (BC2 ) H (d1) (BCd ) such that
 
dBC Ann (u; H (BC2 ) = Ann (ud1 ; H (d1) (BCd )) , (10.3.25)

where the second argument in Ann( ) specifies the ring in which the first argument is
considered. As the relators of the presentation of H (BCd ) given in Corollary 10.3.7
are in degree (d 1), the correspondence x  ud2 x provides, for every p 0,
an isomorphism of Z2 -vector spaces

H (p+1)(d1)1 (BCd ) .

d : H p(d1) (BCd )

We thus get an isomorphism of Z2 -vector spaces

d : H (d1) (BCd )
H (d1)1 (BCd )

multiplying the degrees by d 2 and satisfying

 
d Ann (ud1 ; H (d1) (BCd )) = Ann (u; H (BCd )) . (10.3.26)

From (10.3.25) and (10.3.26), we get

 
t d2 Pt d1 Ann (u; H (BC2 )) = Pt (Ann (u; H (BCd ))) . (10.3.27)

The right hand of (10.3.27) being given by (10.3.24), we checks that


Pt (Ann (u; H (BC2 ))) = t |K|1
K{1,...,n1}
K long

is the unique solution of Equation (10.3.27).

10.3 Chain and Polygon Spaces 461

We have thus proven that (10.3.24) is valid for all d 2. Together with (10.3.21)
and (10.3.18), this establishes the proposition. 

Below is the counterpart of Theorem 10.3.16 for chain spaces. It requires the
length vector  be dominated, i.e. satisfying n i for i n.

Theorem 10.3.17 Let  be a generic length vector which is dominated. Then, the
Poincar polynomial of Cdn () is
 
Pt (Cdn ()) = t |J|(d1) + t (|K|1)(d1)1 . (10.3.28)
J{1,...,n1} K{1,...,n1}
J{n}short K long

Theorem 10.3.17 above reproves the computations of the Betti numbers of Cdn ()
obtained by other methods in [60, Theorem 1] and [58, Theorem 2.1].

Proof The proof is the same as that of Theorem 10.3.16, using the isomorphism dC
of Lemma 10.3.15, instead of dBC . The hypothesis that  is dominated is used to
obtain the analogue of Equation (10.3.20), namely

HG (Cdn ()) (u) d1 (Sh
n ()) . (10.3.29)

Indeed, let J {1, . . . , n 1} be a long subset and k J. As  is dominated,

the set (J {k}) {n} is long. Therefore, the constant terms in relators (3) of
Theorem 10.3.11 vanish and these relators are all multiples of ud1 . Equation (10.3.29)
thus follows from Theorem 10.3.11. 

Example 10.3.18 The length vector  = (1, 1, . . . , 1) is dominated and is generic if

n = 2r + 1. A subset J of {1, . . . , n} is short if and only if |J| r. Hence, for d = 2,
Eq. (10.3.28) gives
 n1  n1
Pt (C22r+1 (1, . . . , 1)) = Pt (N22r+1 (1, . . . , 1)) = k tk + k+2 tk .
kr1 kr1

This formula was first proven in [110, Theorem C].

Remark 10.3.19 The hypothesis that  is dominated is necessary (for any d) in

Theorem 10.3.17, as shown by the example C = Cd4 () for  = (1, 1, 1, ) (see
Example 10.3.14). As C is diffeomorphic to the unit tangent space T 1 S d1 , one
has Pt (C) = 1 + t d2 + t d1 + t 2(d1)1 , as seen in Example 5.4.4, while The-
orem 10.3.17 would give 1 + 3t d2 + 3t d1 + t 2(d1)1 . What goes wrong is
Formula (10.3.29). Using the presentation of HG (C) given in Theorem 10.3.11, one
d1 (Sh
gets that H (d1) (C) is the quotient of n ()) by the constant terms of rela-
tors (3) in Theorem 10.3.11, namely jJ AJ{j} for all J {1, . . . , n 1} which
are long.
462 10 Miscellaneous Applications and Developments

Remark 10.3.20 Let C = Cdn () with  generic and dominated. As observed in [65,
Proposition A.2.4], H (C) is determined by H (d1) (C) when d > 3, using Poincar
duality. Indeed, by Theorem 10.3.11 and Eq. (10.3.29), Z = {ZJ = (AJ ) | J
Shn ()} is a Z2 -basis of H (d1) (C) (Z = 1). The bilinear map H (d1) (C)
H (d1)1 (C) Z2 given by

(x, y)  x  y, [C]

is non degenerate (see Theorem 5.3.12) and thus identifies H (d1)1 (C) with
H (d1) (C) . Let Y = {YJ | J Shn ()} be the Z2 -basis of H (d1)1 (C) which
is dual to Z under this identification. In particular, Y = [C], the generator of
H (n2)(d1)1 (C) = Z2 (we say that Y is the Poincar dual basis to the basis Z).
One has then the multiplication table.

ZJK if J K = and J K Shn ()
ZJ  ZK = (10.3.30)
0 otherwise,

YJK if K J
ZJ  YK = (10.3.31)
0 otherwise

and

YJ  YK = 0 . (10.3.32)

Indeed, (10.3.30) comes from the corresponding relation amongst the classes AJ .
Formula (10.3.32) is true for dimensional reasons, since d > 3. For (10.3.31), note
that, ZJ  YK H (n2(|K||J|)(d1)1 (C) and hence may be uniquely written as a
linear combination

ZJ  YK = L YL ,
LL

where L is the set L Shn () with |L| = |K| |J|. If I L, one has on one hand

ZI  L YL , [C]
= I
LL

and on the other hand


1 if J K = and I J = k
ZI  (ZJ  YK ), [C]
= ZIJ  YK , [C]
=
0 otherwise.

This shows that L = 1 if and only if L = K J.

10.3 Chain and Polygon Spaces 463

We finish this subsection with some illustrations and applications of Theorems

10.3.16 and 10.3.17. The lopsidedness lops () of a length vector  = (1 , . . . , n )
is defined by

lops () = inf{k | J {1, . . . , n 1} with J long and |J| = k} ,

with the convention that inf = 0. The terminology is inspired by that of [91]. If
lops () > 0, one has

dim Sh
n () = n 2 lops () . (10.3.33)

Example 10.3.21 For a generic length vector  = (1 , . . . , n ), the condition

lops () = 0 is equivalent to {n} being long. By Theorem 10.3.16 this is equiva-
lent to BCdn () = : otherwise J = produces a non-zero summand in the first sum
of (10.3.20) (Compare Example 10.3.8). The chamber of  is unique, represented by
e.g. 0 = (, . . . , , 1) with < 1/(n 1).

Example 10.3.22 Let  = (1 , . . . , n ) be a generic length vector with lops () = 1.
From the second sum of (10.3.20), we see that this is equivalent to H d2 (BCdn ()) = 0
(the reduced cohomology is relevant for the cases d = 2, where it says that BC2n ()
is not connected). We check that Shn () is poset isomorphic to Shn (0 ) where 0 =
(, . . . , , 2, 1), with < 1/(n 2). By Lemma 10.3.3, the chamber of  is well
determined modulo the action of Symn1 . The O(d 1)-diffeomorphism type of
BCdn (0 ) may be easily described. It is clear that BCd2 (2, 1) S d2 . Therefore, by
Lemma 10.3.4, BCdn (0 ) (S d1 )n2 S d2 . When d = 2, this is the only case
where BC2n () is not connected, as shown by Theorem 10.3.16. Note that Cdn (0 ) is
empty by Theorem 10.3.17, which is coherent with Proposition 10.3.10.

Example 10.3.23 Let  = (1 , . . . , n ) be a dominated generic length vector with

lops () = 2. We check that Shn () is poset isomorphic to Shn (0 ) where 0 =
(, . . . , , 1, 1, 1), with < 1/(n 3) ([85, Remark 2.4]). By Lemma 10.3.3, the
chamber of  is well determined modulo the action of Symn1 . As in the previous
example, we can describe the O(d 1)-diffeomorphism type of BCdn (0 ). Suppose
first that n = 3. The Morse function f : (S (d1 )2 [3, 3] of Lemma 10.3.1 has no
critical point between its minimum and the level set f 1 (1) = BCd3 (1, 1, 1). By the
Morse Lemma, BCd3 (1, 1, 1) is diffeomorphic to S 2(d1)1 . Using Lemma 10.3.4,
we deduce that BCdn (0 ) (S d1 )n3 S 2(d1)1 . In the same way, one proves that
Cdn (0 ) (S d1 )n3 S d2 . Using Formula (10.3.28), we see that lops () = 2 if
and only if C2n () is not connected.

The following lemma uses the nilpotency class nil introduced in Sect. 4.4.

Lemma 10.3.24 Let  = (1 , . . . , n ) be a generic length vector. Then

(a) If lops () > 1, then lops () = n nil H >0 (BCdn ()) + 1.
464 10 Miscellaneous Applications and Developments

(b) Suppose that  is dominated. If d > 2 or lops () > 2, then

lops () = n nil H >0 (Cdn ()) + 1.

Proof Let BC = BCdn (). Suppose that lops () = k 2. By (10.3.17) and (10.3.20),
the algebra H (BC) contains a copy of HG (BC)/(u) d1 (Sh n ()). By (10.3.33),

dim Shn () = n 2 k. Therefore, there exists x1 , . . . , xnk1 H d1 (BC)
whose cup product v does not vanish in H (nk1)(d1) (BC). By Poincar duality
Theorem 5.3.12, there is w H k(d1)1 (BC) such that v  w = 0 in H (n1)(d1)1
(BC). As k 2 the number k(d1)1 is strictly positive since d 2. Thus, v  w is
a non-vanishing cup product of length nk. Such a length is the maximal possible, as
seen using Sequence (10.3.17). Hence, nil H >0 (BCdn ()) = nk +1. This proves (a).
The proof of (b) is similar, using Theorem 10.3.17 and its proof instead of
Theorem 10.3.16. As dim C = (n 2)(d 1) 1, the class v is of degree
(k 1)(d 1) 1. The latter is strictly positive if d > 2 or k > 2. 

Corollary 10.3.25 Let  = (1 , . . . , n ) and  = (1 , . . . , n ) be two generic

length vectors. Suppose that, for some d 2, there exists a GrA-isomorphism
H (BCdn ()) H (BCdn ( )). Then

lops () = lops ( ) .

If  and  are both dominated, then the above equality holds true if there exists a
GrA-isomorphism H (Cdn ()) H (Cdn ( )).

Proof For the big chain space BCdn ( ), this is follows from Lemma 10.3.24, except
when d = 2 and lops () 1, cases which are covered by Examples 10.3.21
and 10.3.22. The argument for Cdn ( ) is quite similar. The case lops () = 2 is covered
by Example 10.3.23. The case lops () = 1 is not possible if  is dominated, so
lops () = 0 is equivalent to Cdn () = . 

10.3.4 The Inverse Problem

By Lemma 10.3.3, the diffeomorphism type of BCdn () or Cdn () is determined by
the chamber Ch() (up to the action of Symn1 ). The inverse problem consists of
recovering Ch() by algebraic topology invariants of BCdn () (or Cdn ()). We start by
the big chain space.

Proposition 10.3.26 Let  = (1 , . . . , n ) and  = (1 , . . . , n ) be two generic

length vectors. Then, the following conditions are equivalent.
(1) Ch( ) = Ch( ) for some Symn 1 .
(2) BCdn () and BCdn ( ) are O(d 1)-diffeomorphic.
(3) HG (BCdn ()) and HG (BCdn ( )) are GrA[u]-isomorphic.
Moreover, if d > 2 or n > 3, any condition (1)(3) above is equivalent to
10.3 Chain and Polygon Spaces 465

n n
(4) H (BC d ()) and H (BC d ( )) are GrA-isomorphic.
Finally, if d > 2 or if lops () = 2, then any condition (1)(3) above is equiva-
lent to
(5) H (BCdn ()) and H (BCdn ( )) are GrA-isomorphic.

That (5) implies (1) is not known in general if d = 2. That (4) implies (3) is wrong
3 3
if n = 3 and d = 2. Indeed, BC 2 (1, 3, 1) and BC 2 (1, 1, 1) are connected closed
1-dimensional manifolds, thus both diffeomorphic to S 1 but, by Corollary 10.3.7,
one has
 
HG (BC23 (1, 3, 1)) Z2 [u, A1 ] (A21 , u) while HG (BC23 (1, 1, 1)) Z2 [u] (u2 ) .

Implications like (4) (2) or (5) (2) are in the spirit of Proposition 4.2.5:
characterizing a closed manifold (within some class) by algebraic topology tools.
This was the historical goal of algebraic topology (see p. 201).
n
Proof As  and  are generic, one has H (BC d ()) HG (BCdn ()) and the same for
 . The following implications are then obvious, except (a) which was established in
Lemma 10.3.3.

(a)
(1) +3 (2) +3 (3) +3 (4)
CCCCC
CCCC
CCC
%
(5)

We shall now prove that (3) (1), (4) (3) and finally (5) (1).

(3) (1). A GrA[u]-isomorphism HG (BCdn ()) HG (BCdn ( )) descends to

a GrA-isomorphism : HG (BCdn ())/(u) HG (BCdn ( ))/(u). By (10.3.20), this
implies that d1 (Shn ()) and d1 (Sh

n ( )) are GrA-isomorphic. Using
Lemma 4.7.51 and Proposition 4.7.50, we deduce that the simplicial complexes
Sh  
n () and Shn ( ) are isomorphic. It follows that Shn () and Shn ( ) are poset
isomorphic. By Lemma 10.3.3, this implies (1).
n n
(4) (3). Let : H (BC d ()) H (BC d ( )) is a GrA-isomorphism and let
n
(u) = v. We must prove that v = u. This is obvious for d > 2 since H 1 (BC d ( )) =
Z2 . If d = 2, Corollary 10.3.7 implies that

xu = xv = x 2 (10.3.34)
n n
for all x H 1 (BC 2 ( )). By Corollary 10.3.7 again, H (BC 2 ( )) is generated in
n
degree 1, so (10.3.34) implies that (u + v)x = 0 for all x H (BC 2 ( )). By
n
Corollary 10.3.2, BC 2 ( ) is a closed manifold of dimension > 1 (since n 4). We
conclude that v = u by Poincar duality, using Theorem 5.3.12.
466 10 Miscellaneous Applications and Developments

(5) (1). Suppose first that d > 2. By (10.3.22) and (10.3.20), Condition (5) implies
that d1 (Sh 
n ()) and d1 (Shn ( )) are GrA-isomorphic. The argument is then
the same as that for (3) (1).
We now assume that d = 2. By Corollary 10.3.25, Condition (5) implies that
lops () = lops ( ). Let L = lops () = lops ( ). The cases L = 0, 1 were treated in
Examples 10.3.21 and 10.3.22. Let us assume that L > 2. By Theorem 10.3.16 and its
proof, the subalgebra of H (BC2n ()) (respectively: H (BC2n ( ))) generated by the
elements of degree one is isomorphic to 1 (Sh 
n ()) (respectively: 1 (Shn ( ))). By

Condition (5), this implies that 1 (Shn ()) and 1 (Shn ( )) are GrA-isomorphic
and the proof that Ch( ) = Ch( ) proceeds as that for (3) (1). 

Here is the analogue of Proposition 10.3.26 for the chain spaces.

Proposition 10.3.27 Let  = (1 , . . . , n ) and  = (1 , . . . , n ) be two generic

length vectors. Suppose that  and  are dominated. Then, the following conditions
are equivalent.
(1) Ch( ) = Ch( ) for some Symn 1 .
(2) Cdn () and Cdn ( ) are O(d 1)-diffeomorphic.
(3) HG (Cdn ()) and HG (Cdn ( )) are GrA[u]-isomorphic.
Moreover, if d > 2 or n > 4, then any condition (1)(3) above is equivalent to
(4) H (Cdn ()) and H (Cdn ( )) are GrA-isomorphic.
Finally, if d > 2 or if lops () = 3, then any condition (1)(3) above is equivalent
to
(5) H (Cdn ()) and H (Cdn ( )) are GrA-isomorphic.

Proof The proof is the same as that of Proposition 10.3.26, except for the following
small differences. For (3) (1), instead of (10.3.16), one uses Equation (10.3.29),
using that  and  are dominated. For (4) (3), the hypothesis that n > 4 guar-
antees that dim C2n ( ) > 1. For (5) (1), one uses Theorem 10.3.17 instead of
Theorem 10.3.16. 

Remark 10.3.28 In Proposition 10.3.27, implication (4) (1) is wrong for d = 2

and n = 4: BC24 (1, 1, 1, 2) and BC24 (1, 2, 2, 2) are connected closed 1-dimensional
manifolds, thus both diffeomorphic to S 1 . Implication (5) (1) is not known in
general if d = 2. It is however true if one uses the integral cohomology: this difficult
result, conjectured by Walker in 1985 [202] was proved by Schtz in 2010 [170],
after being established when lops () = 3 in [58, Theorem 4] (length vectors with
lopsidedness > 3 are called normal in [58, 170]).

The hypothesis that  is dominated in Proposition 10.3.27 is essential, as seen by

Proposition 10.3.29 and Lemma 10.3.30 below.

Proposition 10.3.29 Let  be a generic length vector and let Symn . If d = 3,

then H (Cdn ()) and H (Cdn ( )) are GrAisomorphic.
10.3 Chain and Polygon Spaces 467

Proposition 10.3.29 was first proved by V. Fromm in his thesis [65, Cor. 1.2.5].
It is wrong if d = 3: for small, C34 (, 1, 1, 1) is diffeomorphic to S 2 S 1 (see
Example 10.3.23) while C34 (1, 1, 1, ) is diffeomorphic to T 1 S 2 RP3 (see Example
10.3.14).
We give below a proof of Proposition 10.3.29 based on an idea of D. Schtz, using
the following lemma.

Lemma 10.3.30 If d = 2, 4, 8, then Cdn () is diffeomorphic to Cdn ( ) for any

Symn .

The hypothesis d = 2, 4, 8 is essential in the above lemma. Indeed, for

small, Cd4 (, 1, 1, 1) is diffeomorphic to S d1 S d2 (see Example 10.3.23) while
Cd4 (1, 1, 1, ) is diffeomorphic to T 1 S d1 (see Example 10.3.14). As d 2, these two
spaces have the same homotopy type only when d = 2, 4, 8 (see Example 5.4.18).

Proof Identifying Rd with C, H or O, we get a smooth multiplication on S d1 with

e1 as unit element. Consider the smooth map : N dn () Cdn () given by

(z1 . . . , zn ) = zn1 (z1 , . . . , zn1 ) .

The embedding j: Cdn () N dn () given by

j(z1 , . . . , zn1 ) = (z1 , . . . , zn1 , e1 )

is a section of . Consider the composed map

j
/ N n () h / N n ( ) / C n ( )
Cdn () d d d

where h (z) = z . Then, h j is a diffeomorphism: a direct computation shows

that its inverse is (h )1 j. 

Proof of Proposition 10.3.29 The case d = 2 is covered by Lemma 10.3.30, so we

assume that d 4. As observe in Remark 10.3.19, one has a GrA-isomorphism

H (d1) (Cdn () d () (10.3.35)

where d () is the quotient of Z2 [A1 , . . . , An1 ] (A

i of degree d 1) by the ideal
generated by A2i , AJ when J {n} is -long and jJ AJ{j} when J is -long.

By Lemma 10.3.30, there exists a ring isomorphism q4 : 4 () 4 ( ). But,
in the definition of d ( ), the integer d is only used to fix the degree of the vari-
ables Ai , and thus the grading of d ( ) (as non-graded rings, the rings d () are
isomorphic for all d). Therefore, the isomorphism q4 defines a GrA-isomorphism

qd : d () d ( ) for all d 2. Together with (10.3.35), this gives a GrA-

isomorphism qd : H (d1) (Cdn ())
H (d1) (Cdn ( )) when d 3.
468 10 Miscellaneous Applications and Developments

Without loss of generality, we may assume that  is dominated. Remark 10.3.20

thus provides an additive basis Z Y of H (Cdn ()). The set Z  = {ZJ = qd (ZJ ) |
J Sh()} is then a Z2 -basis of H (d1) (Cdn ( ). Let Y  = {YJ | J Sh()} be
the basis of H (d1)1 (Cdn ( ) which is Poincar dual to Z  , as in Remark 10.3.20.
Relations (10.3.30)(10.3.32) hold true both for Z Y in H (Cdn ()) and for Z  Y 

in H (Cdn ( )). Therefore, qd extends to a GrA-isomorphism q d : H (Cdn ())

H (Cdn ( )) when d > 3.

10.3.5 Spatial Polygon Spaces and Conjugation Spaces

The integral cohomology ring of the spatial polygon space N3n () has been computed
in [89, Theorem 6.4]. The result is as follows.
Theorem 10.3.31 Let  = (1 . . . , n ) be a generic length vector. Then, there is a
graded ring isomorphism
 n
Z[v, A1 . . . , An1 ] I 
H (N3 (); Z)

where the variables v and Ai are of degree 2 and I  is the ideal generated by the
families of relators

(1) A2i + Ai v i = 1, . . . , n 1
(2) A
 J J {1, . . . , n 1} and J {n} is long
(3) AK v |JK|1 J {1, . . . , n 1} and J is long
KJ

Theorem 10.3.31 says in particular that H odd (N3n (); Z) = 0. The Bockstein exact
sequence for 0 Z Z Z2 0 (see [179, Chap. 5, Sect. 2, Theorem 11])
thus implies that H (N3n ()) H (N3n (); Z) Z2 . Using Theorem 10.3.11, the
correspondence Ai  Ai and v  u thus provides a graded ring isomorphism

H (C2n ()) H (N 2n ())
H 2 (N3n ())

which divides the degrees by half. This suggests that N3n () is a conjugation space,
which we shall prove below. The involution on R3 given by the reflection through
the horizontal plane induces an involution, still called , on N3n (), with fixed point
set equal to N 2n ().
Proposition 10.3.32 Let  = (1 , . . . , n ) be a generic length vector. Then, the
space N3n () endowed with the involution is a conjugation manifold.
The proofs of this proposition use some Hamiltonian geometry, directly or indi-
rectly, so their understanding requires some knowledge in the subject, as presented
10.3 Chain and Polygon Spaces 469

in e.g. [12, Chapters 2 and 3]. Recall that, if  is generic, N3n () is a symplectic
manifold [88, 89, 111, 118]. The proof of Proposition 10.3.32 uses the following
lemma.

Lemma 10.3.33 Let  = (1 , . . . , n ) be a generic length vector. Then,

(a) N3n () is simply connected.
(b) induces the multiplication by (1) on H 2 (N3n (); Z).
(c) is anti-symplectic.

Proof The proof of (a) and (b) proceeds by induction on n. When n = 3, there are
two chambers up to permutation, represented by 0 = (1, 1, 2) and 1 = (1, 1, 1).
As N33 (0 ) = and N33 (1 ) = N 23 (1 ) = pt, (a) and (b) are true. For the induction
step, we consider the diagonal-length function : N3n () R given by (z) =
|n zn n1 zn1 |. By Lemma 10.3.3, we can slightly change n1 without modifying
the O(2)-diffeomorphism type of C3n () and thus the -equivariant diffeomorphism
type of N3n (). Therefore, we can assume that n1 = n , in which case is a smooth
map. Using [83, Theorem 3.2], we deduce that is Morse-Bott function. The critical
points are of even index and are isolated, except possibly for the two extrema. The
preimage of the maximum is either a point or N3n1 (1 , . . . , n2 , n1 + n ) and
the preimage of the minimum is either a point or N3n1 (1 , . . . , n2 , |n1 n |).
This proves (a) by induction.
The restriction  of to N 2n () N3n1 () is also Morse-Bott with Crit  =
Crit but the index of each critical point is divided in half. Thus, passing a critical
point of index 2 corresponds to add a conjugation 2-cell. This proves (b) by induction
on n.
nTo prove (c), we use that N3n1 () is the SO(3)-symplectic reduction at 0 of
2 2
i=0 Si where Si is the standard 2 sphere equipped with the SO(3)-homogeneous
symplectic form with symplectic volume equal to 2i (see [89]). The involution
is clearly anti-symplectic on S2i and this property descends to the symplectic
reduction. 

Proof of Proposition 10.3.32 We give below three proofs. The first one is the one
indicated in [87, Example 8.7].
1st proof. By induction on n, using the function : N3n () R of the proof of
Lemma 10.3.33. For n = 3, N33 () is either empty or a point, which starts the induc-
tion. The induction step uses that is the moment map of an S 1 -Hamiltonian action
on N3n () [118] and called the bending flow [111]. It may be visualized as a rota-
tion of zn1 and zn at constant speed around of axis n zn + n1 zn1 , leaving the
1
other zi s fixed. As a moment map for a circle action, it satisfies Crit = (N3n ())S .
We have seen that the critical points of are isolated or polygon spaces with fewer
1
edges. Hence, by induction hypothesis, (N3n ())S is a conjugation space. The invo-
lution is anti-symplectic by Lemma 10.3.33 and satisfies (z) = 1 (z) for all
z N33 () and S 1 . That N3n () is a conjugation manifold thus follows from [87,
Theorem 8.3].
470 10 Miscellaneous Applications and Developments

2nd proof. We use that N3n () is a symplectic reduction for the Hamiltonian action
of the maximal torus of U(n) on Gr(2; Cn ) (see [88, Theorem 4.4]). The complex
conjugation on Gr(2; Cn ) is anti-symplectic, with fixed point set equal to Gr(2; Rn ),
descends to the involution on N3n (). The manifold Gr(2; Cn ) with the complex
conjugation is a conjugation space (see p. 438 or Remark 9.7.9). That N3n () is a
conjugation space thus follows from [87, Theorem 8.12].
3rd proof. By Theorem 10.3.31, H odd (N3n ()) = 0 and H (N3n ()) is generated by
H 2 (N3n ()). By Lemma 10.3.33, N3n () is simply connected and H 2 is multipli-
cation by (1). Therefore, N3n () is a conjugation manifold by results of V. Puppe
(see [166, Theorem 5 and Remark 2]).
Remark 10.3.34 The quotient BN3n () = BC3n ()/SO(2) is called in [89] the
abelian polygon space (being an S 1 -symplectic reduction while N3n () is an SO(3)-
symplectic reduction). In [89, Theorem 6.4], the integral cohomology ring of BN3n ()
is computed: the statement is as that of Theorem 10.3.31 but with the relators of Corol-
lary 10.3.7. The involution is defined on BN3n () with fixed point set BC2n (). The
3rd proof of Proposition 10.3.32 may be easily adapted to show that is a conjugation
on BN3n ().

10.4 Equivariant Characteristic Classes

Let  be a topological group and let X be a -space. Let = (p : E X) be a

-equivariant vector bundle over X. Recall from Lemma 7.5.6 that is then induced
by i: X X from the vector bundle  over X . The Stiefel-Whitney classes of
 are called the equivariant Stiefel-Whitney classes of . Hence, w() = (w( ))
where : H (X) H (X) is the forgetful homomorphism.
An important example is given by the tautological bundle j (j = 1, . . . , r) over
the flag manifold Fl(n1 , . . . , nr ) (see Sect. 9.5.2), which is an O(n)-equivariant vector
bundle of rank nj .
Proposition 10.4.1 Let R be a closed subgroup of O(n). Then, as an H (BR)-
algebra, HR (Fl(n1 , . . . , nr )) is generated by the equivariant Stiefel-Whitney classes
wi ((j )R ) (j = 1 . . . , r, i = 1, . . . , nj ) of the tautological bundles. In particular,
Fl(n1 , . . . , nr ) is R-equivariantly formal.
Proof We have noticed above that wi (j ) = (wi ((j )R )). By Theorem 9.5.15,
H (Fl(n1 , . . . , nr )) is additively generated by the monomials in the wi (j )s. Such a
monomial is the image by of the corresponding monomial in the wi ((j )R )s and
Fl(n1 , . . . , nr ) is thus R-equivariantly formal. By the Leray-Hirsch theorem 4.7.17,
the monomials in the wi ((j )R)s generate HR (Fl(n1 , . . . , nr )) as an H (BR)-module,
whence the proposition. 
We specialize to the Grassmannian Gr(k; Rn ) = Fl(k, n k) with R = T2 ,
the maximal 2-torus of diagonal matrices in O(n). By (7.5.14), H (BT2 ) Z2
10.4 Equivariant Characteristic Classes 471

[u1 , . . . , un ], with deg(ui ) = 1. The tautological bundles are = 1 and = 2

(see Example 9.5.20). As Gr(k; Rn ) is T2 -equivariantly formal, the restriction to the
fixed points r: HT2 (Gr(k; Rn )) HT2 (Gr(k; Rn )T2 ) is injective by Theorem 7.6.6.
The fixed point set Gr(k; Rn )T2 is discrete and in bijection with [nk ], the set of binary

words = 1 n such that i = k. We identify Gr(k; Rn )T2 with [nk ] via this
bijection, which associates to the coordinate k-plane

 = {(t1 , . . . , tn ) Rn | ti = 0 if i = 0} .

Note that  is the center of the Schubert cell CF for the standard complete flag
F in Rn (see Sect. 9.5.3).

Proposition 10.4.2 For = 1 n [nk ], the restriction homomorphism

r : HT2 (Gr(k; Rn )) HT2 ({}) Z2 [u1 , . . . , un ]

satisfies
 
r (w(T2 )) = (1 + ui ) and r (w(( )T2 ) = (1 + ui ) .
i =1 i =0

We may note that the classes r (w(T2 )) and r (w(T2 )) satisfy the GKM-
conditions (see p. 483)

Proof Let E() be the fiber of over , seen as a T2 -equivariant vector bundle
over . One has a T2 -equivariant isomorphism

E() L(ui )
i =1

where L(ui ) is the T2 -equivariant line bundle over , on which T2 acts via the homo-
morphism dia(1 , . . . , n )  i . This homomorphism is associated to ui under the
bijection of (7.5.8). By Lemma 7.5.11, w(L(ui )T2 ) = 1 + ui and, using Lemma 7.5.7
and (9.4.3), we get

r (w(T2 )) = w((E() )T2 ) = (1 + ui ) .
i =1

The proof of the assertion for is similar, since

E( ) = {(t1 , . . . , tn ) Rn | ti = 0 if i = 1} . 
472 10 Miscellaneous Applications and Developments

Remark 10.4.3 Proposition 10.4.2 implies that the relation


n
w(T2 ) w(( )T2 ) = (1 + ui )
i=1

holds true in HT2 (Gr(k; Rn ). In fact, this provides a presentation of HT2 (Gr(k; Rn )
(see Corollary 10.5.7 and Remark 10.5.8).

Example 10.4.4 Consider Gr(1; R3 ) RP2 . The T 2 -fixed points are in bijection
with {100, 010, 001}. As a Z2 [u1 , u2 , u3 ]-algebra, HT2 (RP2 ) is generated by w1 =
w1 (T2 ) and w i = wi (T2 ) (i = 1, 2). The value of r for these classes are given in
Table 10.1.
One checks that the relations

w1 + w 1 = 1 , w 2 + w1 w 1 = 2 , w1 w 2 = 3 . (10.4.1)

are satisfied. For a generalization to Gr(1; Rn ), see Example 10.5.9.

Example 10.4.5 The Z2 [u1 , . . . , u4 ]-algebra HT2 (Gr(2; R4 )) is generated by wi =

wi (T2 ) and w i = wi (T2 ) (i = 1, 2). The value of r for these classes are given in
Table 10.2.
The following relations are thus satisfied (compare Example 10.5.10).

w1 + w 1 = 1
w2 + w1 w 1 + w 2 = 2
(10.4.2)
w2 w 1 + w1 w 2 = 3
w2 w 2 = 4 .

Table 10.1 The map r for w1 w

1 w
2
Gr(1; R3 ) (Example 10.4.4)
100 u1 u2 + u3 u2 u3
010 u2 u1 + u3 u1 u3
001 u3 u1 + u2 u1 u2

Table 10.2 The map r for w1 w

1 w2 w
2
Gr(2; R4 ) (Example 10.4.5)
1100 u1 + u2 u3 + u4 u1 u2 u3 u4
1010 u1 + u3 u2 + u4 u1 u3 u2 u4
1001 u1 + u4 u2 + u3 u1 u4 u2 u3
0110 u2 + u3 u1 + u4 u2 u3 u1 u4
0101 u2 + u4 u1 + u3 u2 u4 u1 u3
0011 u3 + u4 u1 + u2 u3 u4 u1 u2
10.4 Equivariant Characteristic Classes 473

The above results have their analogues in the complex case. The same proof as
for Proposition 10.4.1 gives the following proposition.

Proposition 10.4.6 Let R be a closed subgroup of U(n). Then, as an H (BR)-

algebra, HR (FlC (n1 , . . . , nr )) is generated by the equivariant Chern classes
ci ((j )R ) (j = 1 . . . , r, i = 1, . . . , nj ) of the tautological bundles. In particular,
FlC (n1 , . . . , nr ) is R-equivariantly formal.

As in the real case, we specialize to the Grassmannian Gr(k; Cn ) = FlC (k, n k)

and their tautological bundles = 1 and = 2 , with R = T , the maximal
torus of diagonal matrices in U(n). As seen in (7.5.14), H (BT ) Z2 [v1 , . . . , vn ],
with deg(vi ) = 2. The associated 2-torus is the maximal 2-torus T2 of diago-
nal matrices in O(n), and the action of T on Gr(k; Cn ) shows that Gr(k; Cn )T =
Gr(k; Cn )T2 . As Gr(k; Cn ) is T -equivariantly formal, the restriction to the fixed
points r: HT (Gr(k; Cn )) HT (Gr(k; Cn )T ) is injective by Theorem 7.6.11. The
same proof of Proposition 10.4.2 gives the analogous formulae
 
r (c(T )) = (1 + vi ) and r (c( )T ) = (1 + vi ) .
i =1 i =0

We now study the equivariant characteristic classes of a rigid -bundle (con-

versations with T. Holm were useful for this part). Recall that a -equivariant
p
vector bundle = (E X) is called rigid if the -action on X is trivial (see
p. 303). Since then X B X, the equivariant Stiefel-Whitney class w( )
belongs to H (X ) H (B) H (X). If is a -equivariant vector bundle over
a space Y which is -equivariantly formal, and if  is a 2-torus for instance, then
r(w( )) = w(|Y  ), r: H (Y ) H (Y  ) is injective (see Theorem 7.6.6) and |Y 
is rigid. Hence, in such cases, rigid equivariant vector bundles play an important role.
p
Let = (E X) be a rigid -equivariant vector bundle, where  is a 2-torus.
Let :  O(1) {1} be a homomorphism. We call a weight  -bundle with
respect to (or just a -weight  -bundle) if v = ()v for all  and v E.

Lemma 10.4.7 If  as 2-torus, then any rigid -equivariant vector

 bundle over a
locally contractible space decomposes into a Whitney sum = hom(,O(1))
of weight subbundles.
p
Proof Let = (E X) and let x X. Let : p1 (U) U Rr be a trivialization
of over an open set U X. Such trivialization is of the form (v) = (p(v), 2 (v)),
where 2 : p1 (U) Rr is a continuous map which is a linear isomorphism on
each fiber, and there is a bijection from the set of trivializations of over U and such

maps. A continuous map y  Ay from U to O(r) is thus defined by the equation

2 (v) = Ap(v) ()2 (v), required to be valid for all  and v p1 (U). If U
retracts by deformation onto b U, a classical folklore fact about representation
theory says that there is a continuous map y  gy from U to O(r), with gb = id,

such that Ay () = gy Ab ()gy1 (see e.g. [79, Lemma 1.2]). We can thus construct a
474 10 Miscellaneous Applications and Developments

1
new trivialization (v)
= (p(v), 2 (v)) of over U by setting 2 (v) = gp(v) 2 (v).
One checks that

2 (v) = Ab ()v , (10.4.3)

in other words, the map y  Ay is constant over U. As X is locally contractible,
there are such trivializations (Ux , x ) as above around each x X. Define

E = {v E | v = ()v, } .

As  as 2-torus, the vector space Ex decomposes into a direct sum of weight subspaces

Ex = Ex (10.4.4)
hom(,O(1))

(see the proof of Lemma 7.5.13). Clearly, E p1 (x) = Ex and, using (10.4.3),
E is the total space of a -weight subbundle of . By (10.4.4), this proves the
proposition. 

By Lemma 7.5.7,  is the Whitney sum of the bundles  . Hence, to compute
the Stiefel-Whitney classes of rigid -equivariant vector bundles, it suffices to know
those of -weight bundles. The following proposition provides the answer. We iden-
tify hom(, O(1)) with H 1 (B) via the bijection of (7.5.8).
Proposition 10.4.8 Let  be a 2-torus and let H 1 (B). Let be a -weight
-bundle of rank r over X. Then, in H (X) H (B) H (X), one has


r
 
w( ) = (1 + )k wrk () . (10.4.5)
k=0

Proof Let = (R pt) be the -weight line bundle over a point and let X = q ,
where q: X pt is the constant map. Hence, X is a -weight bundle over X whose
underlying bundle is a trivial line bundle. Note that any bundle over X may be
endowed with a structure of a -weight bundle using the isomorphism X .
In fact, using the definition of the tensor product bundle of (7.5.5), we see that any -
weight bundle is obtained this way:  X , where is the bundle endowed
with the trivial -action. By Lemma 7.5.7, one has

  ( X ) . (10.4.6)

Consider the two projections B : X B and X : X X (using that X

B X). As -acts trivially on E(), one checks easily that  = and thus
X
w( ) = 1w(). By (7.5.4), ( X ) = B
 and, by Lemma 7.5.11, w(  ) = 1+.
Hence, w(( X ) ) = (1 + ) 1. As ( X ) is a line bundle, the proposition follows
from (10.4.6) together with Lemma 9.6.7. 
10.4 Equivariant Characteristic Classes 475

Example 10.4.9 Let be a vector bundle of rank r over X. We let G = {1, } act on
by (v) = v. Hence, is an u-weight for u the generator of H 1 (BG). Using the
identification HG (X) H (X)[u], Formula (10.4.5) becomes


r
 
w(G ) = wrk ()(1 + u)k .
k=0

Thus,

wj (G ) = wj () + wj1 ()u + + w1 ()uj1 + uj .

In particular, the evaluation ev1 at u = 1 of the equivariant Euler class eG () = e(G )

satisfies

ev1 (eG ()) = ev1 (e(G )) = w() . (10.4.7)

Proposition 10.4.8 has its analogue for T -equivariant complex vector bundles,
where T is a torus. Let hom(T , U(1)), giving rise to () H 2 (BT ) (see
(7.5.11)). Then, if be a -weight complex vector T -bundle of rank r over X, its
equivariant Chern class c(T ) satisfies


r
 
c(T ) = (1 + ())k crk () . (10.4.8)
k=0

in HT (X) H (BT ) H (X). The proof is the same as for Proposition 10.4.8, using
at the end Lemma 9.7.17 instead of Lemma 9.6.7.

10.5 The Equivariant Cohomology of Certain

Homogeneous Spaces

The title of the section paraphrases that of Borels famous paper [15]. The main goal
is to prove and exemplify Theorem 10.5.1 below, due to A. Knutson (unpublished).
In addition, at the end of the section, we study the so-called GKM-conditions for the
flag manifolds.

Theorem 10.5.1 (A. Knutson)

(1) Let 1 , 2 be two closed subgroups of a compact Lie group . Suppose that
/1 or /2 is -equivariantly formal. Then, there is an GrA-isomorphism

1 ,,2 : H1 (/2 )
H (B1 ) H (B) H (B2 ) . (10.5.1)
476 10 Miscellaneous Applications and Developments

(2) Let (, 1 , 2 ) and (  , 1 , 2 ) be two data as in (1). Let
:    be a
continuous homomorphism such that
(i ) i . Denote by
i : i i the
restriction of
. Set = 1 ,,2 and  = 1 ,  ,2 . Then the diagram

H  (  / 2 )  / H (B  ) H (B  ) H (B  )

1 1 2

1
2 (10.5.2)
 
/ H (B1 ) H (B) H (B2 )
H1 (/2 )

is commutative (the vertical arrows are induced by

and
i , using the functo-
rialities of the Borel construction).

Remark 10.5.2 Point (3) says that 1 ,,2 is an isomorphism of H (B1 )-modules
with respect to the isomorphism 1 ,, . With the obvious vertical identification in
the commutative diagram

1 ,,
H1 (/ ) / H (B1 ) H (B) H (B)
O
,
1 
H (B1 ) / H (B1 )

the isomorphism 1 ,, is identified with a GrA-automorphism 1 of H (B1 ).

We do not know in general whether 1 coincides with the identity. This is however
the case in the following cases:
(1) 1 = {1}, since H (B{1}) Z2 [u].
(2) 1 a 2-torus. One uses (1), the naturality of 1 and that H 1 (B1 ) hom(1 ,
{1} (see (7.5.8)).
(3) 1 = O(n). One uses (2), the naturality of 1 and that H (BO(n)) H (BT2 )
is injective, where T2 is a maximal 2-torus of O(n) (see Theorem 9.6.1).
(4) 1 a torus or U(n). The argument is analogous to (1)(3) above.

Proof of Theorem 10.5.1 Let 1 2 acts on  by (1 , 2 ) = 1 21 . The

kernel of the projection 1 2 1 acts freely on . By Lemma 7.2.7, one has
a GrA-isomorphism

H1 (/2 )
H1 2 () . (10.5.3)

Let 1  2 acts on   by (1 , , 2 ) (  ,  ) = (1  1 ,  21 ).
Then  = 1  1 acts freely on   and the multiplication :   
coincides with the quotient map    )/ . Hence, as above, one has a
GrA-isomorphism
10.5 The Equivariant Cohomology of Certain Homogeneous Spaces 477

: H1 2 ()
H1 2 ( ) . (10.5.4)

The kernel of the projection 1  2  acts freely on  . By Lemma 7.2.7

again, one has a GrA-isomorphism

H (1 \ /2 )
H1 2 ( ) . (10.5.5)

Now, 1 \ and /2 being closed smooth manifolds, they are equivalent to finite
-CW-complexes (see [107]). If 1 \ or /2 is -equivariantly formal, then the
equivariant Knneth theorem 7.4.3 holds true, telling us that the strong equivariant
cross product gives a GrA-isomorphism

H (1 \) H (pt) H (/2 )
H (1 \ /2 ) . (10.5.6)

Using Example 7.2.4, we get a final GrA-isomorphism

H (1 \) H (pt) H (/2 ) H (B1 ) H (B) H (B2 ) . (10.5.7)

Combining the isomorphisms (10.5.3)(10.5.7) provides the GrA-isomorphism

1 ,,2 . Point (2) comes from the functoriality of the above constructions and
Point (3) is just an observation about Diagram (10.5.2).

Remark 10.5.3 As the right member of (10.5.1) is symmetric in 1 and 2 , one has
a GrA-isomorphism

H1 (/2 ) H2 (/1 ) .

This can be more easily deduced from (10.5.3) and thus, does not require /1 or
/2 being -equivariantly formal.

The first two applications of Theorem 10.5.1 concern the extreme cases 1 = 
and 1 = 1. The space /  = pt is -equivariantly formal. Therefore, for any
closed subgroup 2 of , one has the GrA-isomorphism

H (/2 ) H (B) H (B) H (B2 ) H (B2 ) . (10.5.8)

Note that one already has an identification H (/2 ) H (B2 ) established in

Example 7.2.4. We do not know whether these two identifications are the same
(it can be proved for special cases, as in Remark 10.5.2). In the case 1 = 1, we
478 10 Miscellaneous Applications and Developments

must assume that /2 is -equivariantly formal. We then get GrA-isomorphism

  
H (/2 ) Z2 H (B) H (B2 ) H (B2 ) image H (B) H (B2 ) .
(10.5.9)

We now concentrate on flag manifolds. The real flag manifold is O(n)-equivar-

iantly formal by Proposition 10.4.1. The complex flag manifold is U(n)-equivariantly
formal by Proposition 10.4.6. Below is a choice of examples.

Example 10.5.4 Let 1 : 1 O(n) denote the inclusion of the closed subgroup 1
in O(n). Theorem 10.5.1 together with Proposition 10.4.1 implies that

H1 (Fl(n1 , . . . , nr )) H (B1 ) H (BO(n)) H (BO(n1 ) BO(nr )) .

Recall from Theorem 9.5.8 that H (BO(n)) Z2 [w1 , . . . , wn ], where wi = wi (n ),

the Stiefel-Whitney classes of the tautological bundle n . Also, H
(BO(nj )) Z2 [w1 (nj ), . . . , wnj (nj )]. Using the Knneth formula, one has

H (BO(n1 ) BO(nr )) Z2 [wi (nj )] (j = 1, . . . , r , i = 1, . . . , nj ) .

By Theorem 9.6.2, the homomorphism H (BO(n)) H (BO(n1 ) BO(nr ))

sends wi to wi (n1 nr ). Hence

H1 (Fl(n1 , . . . , nr )) H (B1 )[wi (nj ) | j = 1, . . . , r , i = 1, . . . , nj ] I ,
(10.5.10)
where I is the ideal generated by

w (n1 nr ) + H 1 (w ) (w = 1 + w1 + w2 + ) .

In the particular case of the complete flag manifold Fl(1, . . . , 1), Isomorphism
(10.5.10) takes the form

H1 (Fl(1, . . . , 1)) H (B1 )[x1 , . . . , xn ] (i (x1 , . . . , xn ) = H 1 (wi )) ,
(10.5.11)

where i is the i-th elementary symmetric polynomial in the variables xj (see

Example 9.5.17).

Example 10.5.5 Let 1 = T2 O(1) O(1) be the maximal 2-torus

of the diagonal matrices in O(n). By Theorem 9.6.1 and its proof, H (BT2 )
Z2 [u1 , . . . , un ], with deg(ui ) = 1, and H 1 (wi ) = i (u1 , . . . , un ), where i denotes
the i-th elementary symmetric polynomial. Also, O(n)/T2 Fl(1, . . . , 1) is O(n)-
equivariantly formal by Proposition 10.4.1. Therefore, for any closed subgroup 2
10.5 The Equivariant Cohomology of Certain Homogeneous Spaces 479

in O(n), Theorem 10.5.1 provides the isomorphism

= T2 ,O(n),2 : HT2 (O(n)/2 )
Z2 [u1 , . . . , un ] H (BO(n)) H (B2 ) .
(10.5.12)
For instance, (10.5.10) implies that

HT2 (Fl(n1 , . . . , nr )) Z2 [u1 , . . . , un ][wi (nj ) | j = 1, . . . , r , i = 1, . . . , nj ] I ,
(10.5.13)

where I is the ideal generated by wi (n1 nr ) i (u1 , . . . , un ) (i = 1, . . . , n).

In the particular case of the full flag manifold Fl(1, . . . , 1), we thus get that

HT2 (Fl(1, . . . , 1)) Z2 [u1 , . . . , un , x1 , . . . , xn ] I , (10.5.14)

where I is the ideal generated by

i (u1 , . . . , un , ) = i (x1 , . . . , xn , ) (i = 1, . . . , n) .

Another example is given by the Stiefel manifold Stief(k, Rn ) = O(n)/O(nk) of

orthonormal k-frames in Rn . Here, H (BO(n)) H (BO(nk)) is just the obvious
epimorphism Z2 [w1 . . . , wn ] Z2 [w1 . . . , wnk ]. Hence, (10.5.12) implies that

HT2 (Stief(k, Rn )) Z2 [u1 , . . . , un ] (nk+1 , . . . , n ) . (10.5.15)

Note that Stief(k, Rn )) is not O(n)-equivariantly formal. At the contrary,

: HT2 (Stief(k, Rn )) H (Stief(k, Rn ))

is the zero homomorphism in positive degrees.

It is reasonable to conjecture that the isomorphism 1 ,O(n),O(n1 )O(nr ) of

Theorem 10.5.1 identifies the equivariant Stiefel-Whitney class wi ((j )1 ) (see
Sect. 10.4) with 1 wi (j ). The following proposition proves it for the maximal
2-torus T2 :

Proposition 10.5.6 Under the isomorphism = T2 ,O(n),O(n1 )O(nr ) , one has

(wi ((j )T2 )) = 1 wi (j ) .

Proof We start with the Grassmannian Gr(k; Rn ) = Fl(k, n k). For the inclusion

: (T2 , O(n), 1 O(n k)) (T2 , O(n), O(k) O(n k)), Diagram (10.5.2) has
the form
480 10 Miscellaneous Applications and Developments

/ H (BT2 ) H (BO(n)) H (BO(k) BO(n k))

HT2 (Gr(k; Rn ))

id
2
 
/ H (BT2 ) H (BO(n)) H (BO(n k))
HT2 (Stief(k, Rn ))

(we do not write the indices to the isomorphisms ). The O(n k)-principal bundle
Stief(k, Rn ) Gr(k; Rn ) is 1 , so Stief(k, Rn )T2 Gr(k; Rn )T2 is (1 )T2 . Hence,
w1 ((1 )T2 ) ker
. Using (10.5.15), one sees that ker(id
2 ) is the Z2 -vector
space generated by 1w1 (1 ). Therefore, (w1 ((1 )T2 )) = 1w1 (1 ). A symmetric
argument shows that (w1 ((2 )T2 )) = 1 w1 (2 ).
This starts an induction argument on k to prove the proposition for Gr(k; Rk+1 ) =
Fl(k, 1). The induction step uses Diagram (10.5.2) for the inclusion

: (T2 , O(k + 1), O(k) O(1)) (T2 , O(k + 2), O(k + 1) O(1))

which looks like

/ H (BT2 ) H (BO(k+2)) H (BO(k + 1) BO(1))

HT2 (Gr(k + 1; Rk+2 ))

id
2
 
/ H (BT2 ) H (BO(k+1)) H (BO(k) BO(1))
HT2 (Gr(k; Rk+1 ))

By induction hypothesis, (wi ((1 )T2 )) = 1wi (1 ) for i k and (w1 ((2 )T2 )) =
1 w1 (2 ). By Proposition 10.4.1, wk+1 ((1 )T2 ) is in ker
. On the other hand,
ker(id
2 ) is, in degree k + 1, the Z2 -vector space generated by 1 wk+1 (1 ).
This proves that (wk+1 ((1 )T2 )) = 1 wk+1 (1 ).
We now prove the proposition for Gr(k; Rn ) = Fl(k, n k), by induction on n.
The previous argument starts the induction for n = k+1. The induction step proceeds
as above, using Diagram (10.5.2) for the inclusion

: (T2 , O(n), O(k) O(n k)) (T2 , O(n + 1), O(k) O(n + 1 k))

and checking ker

and ker(id
2 ) in degree n k + 1.
Finally, consider the map j : Fl(n1 , . . . , nr ) Gr(nj ; Rn ) defined, in the mutu-
ally orthogonal subspaces presentation of Fl(n1 , . . . , nr ) given in (2) p. 376, by
j (W1 , . . . , Wr ) = Wj . Take the simplest permutation matrix O(n) so that
(0 Rnj 0) = Rnj 0. The conjugation with gives an homomorphism

: (T2 , O(n), O(n1 ) O(nr )) (T2 , O(n), O(n1 ), O(nj ) O(n nj ))
10.5 The Equivariant Cohomology of Certain Homogeneous Spaces 481

such that
2 (1 O(nj ) 1) = O(nj ) 1 and, for k = j,
2 (1 O(nk ) 1)
1 O(n nj ). Diagram (10.5.2) for
has the form

/ H (BT2 ) H (BO(nj )) H (BO(nj ) BO(n nj ))

HT2 (Gr(nj ; Rn ))

j id
2
 
/ H (BT2 ) H (BO(n)) H (BO(n1 ) BO(nr ))
HT2 (Fl(n1 , . . . , nr ))

Therefore,

(wi ((j )T2 )) = j (wi ((1 )T2 )) since j = j 1

=
(id
2 ) (wi ((1 )T2 ))
= (id
2 )(1 wi (1 )) (case Gr(nj ; Rn ) done above)
= 1 wi (j ) since j = j 1 

Using (10.5.13), Proposition 10.5.6 has the following corollary.

Corollary 10.5.7 One has an isomorphism of Z2 [u1 , . . . , un ]-algebra


HT2 (Fl(n1 , . . . , nr )) Z2 [u1 , . . . , un ][wi ((j )T2 ) | j = 1, . . . , r , i = 1, . . . , nj ] I ,

where I is the ideal generated by

wi ((1 )T2 (r )T2 ) i (u1 , . . . , un ) (i = 1, . . . , n) . (10.5.16)

Remark 10.5.8 The vanishing of the generators of I is equivalent to the relation


n
w((1 )T2 ) w((r )T2 ) = (1 + ui )
i=1

holding in HT2 (Fl(n1 , . . . , nr )). This relation may be obtained in the following way.
The trivializing map of (9.5.8) provides a morphism of T2 -equivariant bundles

E(1 r ) / Rn

 f

Fl(n1 , . . . , nr ) / pt

where T2 acts on Rn via the standard action of O(n). Using Example 7.5.12, one has


n

w((1 )T2 (r )T2 ) = f (w((R )T2 ) = n
(1 + ui ) .
i=1
482 10 Miscellaneous Applications and Developments

Example 10.5.9 Let  = O(n) and 2 = O(1)O(n1), so /2 RPn1 . Then

H (BO(1)) Z2 [w1 ] and H (BO(n 1)) Z2 [w 1 , . . . , w n1 ], where wi and w i
are the Stiefel-Whitney classes of the tautological bundles 1 and n1 . If 1 = T2 ,
we get, as in (10.5.14) that HT2 (RPn1 ) is the quotient of

Z2 [u1 , . . . , un ][w1 , w 1 , . . . , w n1 ]

by the relations

w1 + w 1 = 1 , w k +w1 w k1 = k (k = 2, . . . , n1) , w1 w n1 = n . (10.5.17)

These relations are the same as those of (10.4.1), no wonder given Corollary 10.5.7.
As RPn1 is T2 -equivariantly formal, H (RPn1 ) is the quotient of HT2 (RPn1 )
by the relations ui = 0. Relations (10.5.17) becomes w k = w1k and w1n = 0. Thus,
H (RPn1 ) Z2 [w1 ]/(w1n ) as expected.

Example 10.5.10 Let  = O(4) and 2 = O(2) O(2), so /2 Gr(2; R4 ).

As in Example 10.5.9, we see that HT2 (Gr(2; R4 )) is isomorphic to the quotient of
Z2 [u1 , . . . , u4 ][w1 , w2 , w 1 , w 2 ] by the relations

w1 + w 1 = 1
w2 + w1 w 1 + w 2 = 2
(10.5.18)
w2 w 1 + w1 w 2 = 3
w2 w 2 = 4 .

These relations are the same as those of (10.4.2) which is coherent with Corol-
lary 10.5.7. As in Example 10.5.9, we get a presentation of H (Gr(2; R4 )) by setting
ui = 0. This presentation is equivalent to that of Example 9.5.23.

Example 10.5.11 In the case 1 = {1} (the trivial subgroup), one has H{1} (Fl(n , . . . ,
1

nr )) H (Fl(n1 , . . . , nr )) and the above coincides with some theorems of Sect. 9.5.2
(e.g. Theorem 9.5.14 and Example 9.5.17).

Example 10.5.12 The analogues of the above examples works for the complex flag
manifolds FlC (n1 , . . . , nr ), where O(n) is replaced by U(n), T2 is replaced by the
maximal torus T (S 1 )n of the diagonal matrices in U(n) and the Stiefel-Whitney
classes are replaced by the Chern classes. The variables xi and yi have degree 2 instead
of 1. The analogue of Proposition 10.5.6 says that (ci ((j )T2 )) = 1 ci (j ). One
can show that these results are valid for the cohomology with coefficients in any
field.

Example 10.5.13 Consider the case where  = S 3 C2 and 1 = 2 = S 1 R2 .

Thus, /2 S 2 . By (7) p. 275, the inclusion i: S 2 (S 2 ) is homotopy equivalent
to a principal bundle with structure group S 3 . By Proposition 4.7.23, H 2 i is an
isomorphism and therefore S 2 is S 3 -equivariantly formal. Now, BS 1 CP and
BS 3 HP . Hence, H (B1 ) Z2 [x], H (B2 ) Z2 [y] and H (B) Z2 [p],
10.5 The Equivariant Cohomology of Certain Homogeneous Spaces 483

where x and y are of degree 2 and p of degree 4. The inclusion i : i  satisfies

H B1 (p) = x 2 and H B2 (p) = y2 (see Proposition 6.1.11). Therefore, using
Theorem 10.5.1, one gets

HS1 (S 2 ) Z2 [x, y] (x 2 + y2 ) .

We finish this section by studying the GKM-conditions for the flag manifolds.
Viewing Fl(n1 , . . . , nr ) SM(n) using (9.5.2), the fixed point set Fl(n1 , . . . , nr )T2
is clearly formed by the diagonal matrices. If  = dia(x1 . . . , xn ) is a diagonal
matrix and Symn , we set  = dia(x(1) . . . , x(n) ). We say that a class
a HT2 (Fl(n1 , . . . , nr )T2 ) satisfies the GKM-conditions if, for all transposition
= (i, j), the class a a is a multiple ui uj . This is an ad hoc formula-
tion for Fl(n1 , . . . , nr ) (in the spirit of [119]) of the conditions introduced in [71]
by M. Goresky, R. Kottwitz and R. MacPherson (whence the initials GKM). The
importance of the GKM-conditions is illustrated in the following result.
Proposition 10.5.14 The image of

r: HT2 (Fl(n1 , . . . , nr )) HT2 (Fl(n1 , . . . , nr )T2 )

is the set of classes satisfying the GKM-conditions.

Proof We first prove that the classes in the image of r satisfy the GKM-conditions.
Let = (i, j) with i < j. The GKM-condition for is trivial if  =  (i.e., when
xi = xj ). We may thus assume that xi = xj . In the proof of Proposition 9.7.1, we
have introduced an embedding r ij : SO(2) SO(n). The orbit of  under the action
of r ij (SO(2)) on Fl(n1 , . . . , nr ) by conjugation is a circle Cij joining  to  . This
circle is T2 -invariant and is actually an (ui uj )-circle in the sense of Example 7.6.10,
with fixed points {,  }. One has a commutative diagram

HT2 (Fl(n1 , . . . , nr )) r / H (Fl(n1 , . . . , nr )T2 ) / Z2 [u1 , . . . , un ]
T2
Fl(n1 ,...,nr )T2

 
HT2 (Cij ) r / H (C T2 ) / Z2 [u1 , . . . , un ]
T2 ij
,

where all the vertical arrows are induced by inclusions. The right vertical arrow is
just the projection. Therefore, the GKM-condition for comes nfrom Example 7.6.10.
For the converse, we use the weighted trace f (M) = j=1 j Mjj which is, by
Proposition 9.5.2, a Morse function with Critf = Fl(n1 , . . . , nr )T2 . Let Wx =
f 1 (, x] and let Tx be the set of those transpositions such that  and 
are in Wx . We claim that a class a HT2 (WxT2 ) is in the image of r: HT2 (Wx )
HT2 (WxT2 ) if and only if it satisfies the GKM-conditions for Tx . The only if part
484 10 Miscellaneous Applications and Developments

is proven as above since, if (i, j) Tx , then Cij Wx . The proof of the if part
proceeds by induction on the number nx of critical values of f in Wx , starting triv-
ially if nx = 0 or 1. For the induction step, choose z < y such that nz = ny 1. Let
Mz,y = Wy Wz . As in (7.6.14), one has the commutative diagram

0 / HT (Wy , Wz ) / HT (Wy ) / HT (Wz ) / 0
2 2 2

rz,y ry rz (10.5.19)
  

0 / HT (Mz,y ) / H (WyT2 ) / H (WzT2 ) / 0
2 T2 T2

where all arrows are induced by the inclusions and where the horizontal lines are
exact.
If a HT2 (WyT2 ) satisfies the GKM-conditions for Ty , so does a = (a) for
for some b H (Wz ). As is surjective,
Tz . By induction hypothesis, a = rz (b) T2
there exists b HT2 (Wy ) such that a ry (b) = (c) for some c HT2 (Wz,yT2
). By
the only if part the class ry (b) satisfies the GKM-conditions for Ty , and then so
T2
does a ry (b). let D Mz,y . Let TD be the set of transpositions in Ty such that
D = D. For each (i, j) TD , the class (a ry (b))D is a multiple of ui uj (since
(a ry (b)) = 0 when  = D). Since is injective, the class cD is a multiple of
ui uj for each (i, j) TD . By the proof of Proposition 9.5.2, the negative normal
bundle (D) for f at D is the Whitney sum

(D) = TD Cij .
(i,j)TD

As Z2 [u1 , . . . , un ] is a unique factorization domain, the class cD is a multiple of

 
(ui uj ) = e(TD Cij ) = e( (D)) .
(i,j)TD (i,j)TD

T2
This can be done for any D Mz,y . Using Diagram (7.6.13), we deduced that there
exists c HT2 (Wy , Wz ) such that rz,y (c) = c. Hence, a = ry ((
c) + b). 

Analogous GKM-relations hold true for the T -equivariant cohomology of the

complex flag manifold FlC (n1 , . . . , nr ). Here, T is the maximal torus of diagonal
matrix in U(n). It is naturally isomorphic to U(1)n and thus H (BT ) is isomorphic
to Z2 [v1 , . . . , vn ] where vi H 2 (BT ) is the class associated, under the map
of (7.5.11), to the projection of U(1)n onto its i-th factor. As in the real case, we
see FlC (n1 , . . . , nr ) HM(n) (using (9.7.2)), then FlC (n1 , . . . , nr )T is formed by
the diagonal matrices. We say that a class a HT (FlC (n1 , . . . , nr )T ) satisfies the
GKM-conditions if, for all transposition = (i, j), the class a a is a multiple
vi vj .
10.5 The Equivariant Cohomology of Certain Homogeneous Spaces 485

Proposition 10.5.15 The image of injective homomorphism of Z2 [v1 , . . . , vn ]

-algebras

r: HT (FlC (n1 , . . . , nr )) HT (FlC (n1 , . . . , nr )T )

is the set of classes satisfying the GKM-conditions.

Proof We use the injective homomorphism r ij : SU(2) U(n), introduced in
the proof of Proposition 9.7.1. The orbit of  under the action of r ij (SU(2)) on
FlC (n1 , . . . , nr ) by conjugation is a 2-sphere Sij (diffeomorphic to SU(2)/U(1)
S 2 ), whose intersection with FlC (n1 , . . . , nr )T is {,  }. This sphere is T -invariant
and is actually a -sphere in the sense of Example 7.6.12, with () = vi vj .
Note that FlC (n1 , . . . , nr )T = FlC (n1 , . . . , nr )T2 . The proof of Proposition 10.5.15
then goes as that of Proposition 10.5.14, replacing Cij by Sij and the mater-
ial of Proposition 9.5.2 and Example 7.6.10 by that of Proposition 9.7.1 and
Example 7.6.12. 

10.6 The Kervaire Invariant

In 1960, Michel Kervaire introduced an invariant for framed manifolds which enabled
him to construct the first topological manifold admitting no smooth structure (see
Theorem 10.6.12). The computation of the Kervaire invariant then led to one of the
most important problems in homotopy theory (see Theorem 10.6.11). In this section,
we give a survey of the geometric side of the Kervaire invariant, using the surgery
point of view of Wall (though well known by specialists, such a presentation is new in
the literature). The stable homotopy aspect of the invariant is only briefly mentioned,
being beyond the scope of this book. The notes of Weber [205] were helpful for
preparing this section. We start with the invariant introduced by Arf [10] in order to
classify quadratic forms in characteristic 2.
Let V be a finite dimensional Z2 -vector space. A quadratic form on V is a map
q: V Z2 such that the expression

B(x, y) = q(x) + q(y) + q(x + y) (10.6.1)

defines a bilinear form B on V , the bilinear form associated to q. It is obviously

symmetric and alternate, i.e. B(x, x) = 0.
For i Z2 , let i (q) =  q1 (i). The majority (or democratic) invariant of q is
the element maj(q) Z2 defined by

1 if 1 (q) > 0 (q)
maj(q) =
0 otherwise.

In fact, 0 (q) = 1 (q) when B is non-degenerate (see Proposition 10.6.1).

486 10 Miscellaneous Applications and Developments

Suppose that the associated bilinear form B is non-degenerate. Since it is alternate,

there exists a symplectic basis {a1 , . . . , ak , b1 . . . , bk } of V for B (see the proof of
Lemma 9.8.1). The Arf invariant Arf(q) Z2 of q is defined by


k
Arf(q) = q(ai )q(bi ) .
i=1

That Arf(q) is independent of the choice of the symplectic basis follows from
Point (2) of the following proposition. Two quadratic forms q and q on V are equiv-
alent if there is an automorphism h of V such that q = q h.

Proposition 10.6.1 Let V be a finite dimensional Z2 -vector space. Let q and q be

two quadratic forms on V with non-degenerate associated bilinear forms. Then
(1) 0 (q) = 1 (q).
(2) Arf(q) = maj(q).
(3) q and q are equivalent if and only if Arf(q) = Arf(q ).

Proof (Compare [24, Sect. III.1].) Suppose first that dim V = 2. Let A = {a, b} be
a symplectic basis for B on V with which we compute Arf(q) = q(a)q(b). Suppose
that Arf(q) = 0. There are two cases:
(i) q(a) + q(b) = 0, hence maj(q) = 0.
(ii) q(a) + q(b) = 1. By (10.6.1), one has q(a + b) = 0 and maj(q) = 0. By
symmetry, one may assume that q(a) = 0 and q(b) = 1. Then, the changing of
basis a = a and b = a + b makes this form equivalent to that of (i).
If Arf(q) = 1, then q(a) = q(b) = 1, thus q(a + b) = 1 and maj(q) = 1. Points (1),
(2) and (3) are thus proven when dim V = 2. We denote by qi the quadratic form for
which Arf(qi ) = i.
We now prove (1) and (2) by induction on dim V . Let B = {a1 , . . . , ak , b1 . . . , bk }
be a basis of V which is symplectic for B. One has V = V V where V is generated
by {aj , bj | j k 1} and V is generated by {ak , bk }. The restriction of q to V
(respectively: V ) is denoted by q (respectively: q ). We have

Arf(q) = Arf(q) + Arf(q) using the basis B

= maj(q) + maj(q) by induction hypothesis.

It thus suffices to prove that maj(q) + maj(q) = maj(q). Suppose that q = q1 . One
has 1 (q1 ) = 3 and 0 (q1 ) = 1. Therefore,

1 (q) = 30 (q) + 1 (q)
0 (q) = 31 (q) + 0 (q) .
10.6 The Kervaire Invariant 487

By induction hypothesis, 0 (q) = 1 (q). Therefore, maj(q) = maj(q), which

implies that

maj(q) = maj(q) + 1 = maj(q) + maj(q1 ) .

If q = q0 , a similar argument shows that maj(q) = maj(q) = maj(q) + maj(q0 ).

Point (1) is thus established.
It remains to prove Point (3). If q and q are equivalent, it is obvious that maj(q) =
maj(q ). Conversely, we easily deduce from above that q is an orthogonal sum of q0 s
and q1 s and that Arf(q) is the numbers of q1 s mod 2. That q and q are equivalent
when Arf(q) = Arf(q ) then comes from the equivalence

q1  q1  q0  q0

which is achieved by the automorphism

h(a1 ) = a1 + a2 , h(a2 ) = a2 , h(b1 ) = b1 , h(b2 ) = b1 + b2 . (10.6.2)


We now make some preparations for the Kervaire invariant. Let and  be two
vector bundles over the same space X and let r = (prX : X Rr X) denote the
product vector bundle of rank r over X. A stable isomorphism from to  is a family

of isomorphisms hr,r  : r  r for each r, r  sufficiently large, such that
if s r and s r  , the diagram

hr,r 
r /  r

 hs,s

s /  s

is commutative, where the vertical arrows are the inclusion morphisms. A stable
trivialization of a vector bundle over X is a stable isomorphism of with a product
bundle. A vector bundle admitting a stable trivialization is called stably trivial.
A framed manifold is a smooth manifold M together with a smooth stable trivial-
ization of its tangent bundle TM, called a stable framing of M. Two framed manifolds
M1 and M2 of the same dimension m and with Bd M1 = Bd M2 are framed cobordant
if there is exists a framed manifold W m+1 such that

BdW = M1 M2 and M1 M2 = Bd M1 = Bd M2

and whose stable framing extends those of M1 and M2 . The set of framed cobordism
fr
classes of framed closed manifolds of dimension m is denoted by m . It is an abelian
group for the disjoint union.
488 10 Miscellaneous Applications and Developments

Let M m be a closed manifold. Let Mk be the normal bundle of an embedding of

M into Rn+k . If k > m, such an embedding is unique up to isotopy, so the stable

isomorphism class of M
k is well defined. There is a canonical stable trivialization h
M

of TM M k since the latter is the restriction of T Rn to M. If h: TM r
m+r
is represents a stable isomorphism, the stable isomorphism represented by

hM
m+k+r o / TM r k / m+r k
h

TM M
k r
M M

represents a stable trivialization of M

k . For k large enough, such a stable trivialization

gives a vector bundle morphism

k )
E(M / Rk

 
M / pt

Applying the Pontryagin-Thom construction (see p. 430) to this morphism gives an

fr
element of m+k (S k ). This produces an isomorphism m m S from the framed

cobordism onto the m-stem

m
S
= lim

m+k (S k ) (10.6.3)
k

(see [152, Sect. 7]).

Example 10.6.2 Let (S n , F) be the standard sphere equipped with a stable framing
F. The comparison between F and the standard stable framing F0 (extending to
Dn+1 ) takes the form F = F F0 , where F : S n SO is a smooth map, whose
class [F ] n (SO) is unique (compare Lemma 9.1.2). The correspondence F 
[S n , F] nfr nS gives a map Jn : n (SO) nS which coincides with the
J-homomorphism of Whitehead [207].

We now describe framed surgery, following the point of view of Wall [204] (for
another approach, see p. 495). Let : S j Dmj M m (0 j m) be a smooth
embedding. We consider the (m + 1)-dimensional manifold

W = M [0, 1] Dj+1 Dmj

where is seen having image in M {1}. The corners of W may be smoothed in a

canonical way (see [17, Appendix, Theorem 6.2]) and thus W is a smooth cobordism
between M and M where

M = M int(im) |Sj Smj1 Dj+1 S mj1 .

10.6 The Kervaire Invariant 489

The manifold M is said being obtained from M by a surgery using . If M is endowed

with a stable framing F which extends to a stable framing of W , we say that the
surgery on is a stably framed surgery (for the framing F). Let M m be a manifold
and let : S j M be a continuous map, with j < m 1. We wish to perform a
surgery on M using a smooth embedding : S j Dmj M so that the restriction
of to S j {0} is homotopic to . Let Imm(S j Dmj , M) be the set of regular
homotopy classes of smooth immersions from S j Dmj into M. The restriction to
S j {0} provides a map

: Imm(S j Dmj , M) [S j , M] .

Proposition 10.6.3 Let M m be a manifold and let j < m 1. Then

(a) a stable framing F of M provides a map

F : [S j , M] Imm(S j Dmj , M)

which is a section of , i.e. F (a) = a for all a [S j , M].

(b) Suppose that F (a) contains an embedding . Then, the surgery on is a stably
framed surgery for the framing F.

Proof Let : S j M be a continuous map, which we extend to 1 : S j Dmj M

by 1 (x, z) = (x). The framing F of M gives rise to a stable trivialization of 1 TM.
On the other hand, T (S j Dmj ) has a canonical stable trivialization. Comparing
these two trivializations gives a stable isomorphism 1S from T (S j Dmj ) to 1 TM.
Since j < m 1, j (GL(m; R)) j (GL(m + N; R)) is an isomorphism and thus
1S is induced by a unique isomorphism from T (S j Dmj ) to 1 TM, giving rise to
an injective bundle map 1 : T (S j Dmj ) TM. Assertion (a) then follows from
the classification of immersions [93, Sect. 5], saying that Imm(S j Dmj , M) is (by
the tangent map) in bijection with the set of injective bundle maps from T (S j Dmj )
into TM (this also uses that j < m 1).
If is as in (b), the equation T = 1 is satisfied. This is exactly what is needed
to extend the stable framing F over W . For more details, see [204, Theorem 1.1]. 

Proposition 10.6.4 (Surgery below the middle dimension) A stably framed compact
m-dimensional manifold M m is stably framed cobordant to a manifold M  which is
([m/2] 1)-connected (i.e. i (M  ) = 0 for i m/2 1).

Proof Let F be the stable framing of M and let a [S k , M]. If k < m/2, general
position implies that F (a) contains an embedding : S k Dmk M using
which, by Proposition 10.6.3, a stably framed surgery may be performed on M. Up
to homotopy equivalence, W is obtained from M by attaching a (k +1)-cell and from
M by attaching a (m k)-cell. Hence, the inclusions i: M W and j: M W
satisfy

i: p (M) p (W ) is an isomorphism for p k 1. For p = k, it is surjective

and kills [].
490 10 Miscellaneous Applications and Developments

j: p (M ) p (W ) is an isomorphism for p m k 1.
Therefore, if k m/2 1, then p (M ) is isomorphic to p (M) for p k 1 and,
if [] = 0, k (M ) is isomorphic to a strict quotient of k (M). In particular, if M is
(k1)-connected, so is M and, in addition, the class [] has been killed. Therefore,
by a finite sequence of framed surgeries, one may obtain a stably framed manifold
M  which is [m/2] 1-connected. For more details, see e.g. [204, Theorem 1.2]. 
We now treat the middle dimensional surgery for an (k 1)-connected stably
framed closed manifold M m with m = 2k (k odd). Consider the Hurewicz homomor-
phisms h: k (M) Hk (M; Z) and h2 : k (M) Hk (M) (sending [: S k M]
to H ([S k ])); since k 3, we do not worry about base points). By the Hurewicz
theorem [82, Theorem 4.32], h is an isomorphism and, the universal coefficient
theorem [82, Theorem 3B.5], Hk (M) Hk (M; Z) Z2 . Hence, h2 descends to an
isomorphism

k (M) 2 k (M)
Hk (M) . (10.6.4)

Let : V v N n be an immersion of smooth manifolds. The self-intersection

SI() of is defined by

SI() = {x (V ) |  1 (x) > 1} .

When is in general position and 3v < 2n, then  1 (x) 2 and SI() is a
(2v n)-dimensional submanifold of N (see [77, Theorem 2.5])
Let a k (M). Choose an immersion : S k Dk M in general position
representing F (a) (where F is the stable framing of M). Let 0 be the restrictions
of to S k {0}. The self-intersection SI(0 ) is thus a finite number of points. Define

q () = SI(0 ) mod 2 .

Proposition 10.6.5 Let m = 2k 6 with k odd. Let M m be (k 1)-connected

manifold endowed with a stable framing F. Then
(a) the above correspondence a  q () induces, via (10.6.4), a well defined map

q = qM : Hk (M) Z2 .

(b) for all a, b Hk (M) one has

q(a + b) = q(a) + q(b) + a b

where a b denotes the (absolute) intersection form (see Sect. 5.3.3).

(c) q(a) = 0 if and only if F (a) contains an embedding.
Proof Let and be two immersions representing F (a) which are in general
position. There are thus joined by an immersion A: S k Dk I M I which
10.6 The Kervaire Invariant 491

we also assumed to be in general position. Then, SI(A0 ) is a compact 1-manifold

with boundary SI(0 ) SI( 0 ). As an arc has two ends, one has q () = q ().
Thus,
q () depends only on a k (M), so we can write q (a). In order to prove Point (a),
it remains to establish that q (2a) = 0, which will be done together with the proof
of (b).
Let a, b k (M). Let F (a) and F (b) be two immersions in general
position. Then, f (a + b) may be represented by an immersion obtained by con-
nected sum of and along a tube S k1 Dk I, disjoint from the images of
and except at its ends. Let B0 (, ) Z2 defined by

B0 (, ) =  [(0 (S k ) 0 (S k )] mod 2 .

Obviously, one has

q (a + b) = q (a) + q (b) + B0 (, ) . (10.6.5)

We claim that

B0 (, ) = a b . (10.6.6)

Indeed, if 0 and 0 are embeddings, this is just Corollary 5.4.13. We claim that a
and b may be represented by embeddings 0 and 0 such that 0 (S k ) 0 (S k ) =
0 (S k ) 0 (S k ). Suppose first that SI(0 ) is even. The points of SI(0 ) may be be
pairwise eliminated by Whitney procedure (since M is simply connected: see [210,
Theorem 4 and its proof]). This produces a regular homotopy 0t : S k M with
00 = 0 , so that 0 = 01 is an embedding. The control on the Whitney process
guarantees that the intersection of 0t (S k ) with (S k ) is constant in t. If SI(0 )
is odd, we use that there exists an immersion : S k Rm with SI() = 1 (see
[210, Sect. I.2]). We can compose with a chart Rm  M whose range is away
from 0 (S k ) 0 (S k ), obtaining an immersion  : S k M which, as a map, is
null-homotopic. Let : S k M be the immersion obtained by connected sum of

0 and . Then, 0 represents a and SI( 0 ) is even, so we can proceed as in the
previous case. The whole process may be independently applied to 0 . This proves
the claim and then Eq. (10.6.6).
Equations (10.6.5) and (10.6.6) imply that q (2a) = 0, because

a a = PD(a)  PD(a), [M]

= 0 .

Indeed, each x Hk (M) satisfies x  x = Sqk (x) = x  vk (M), where vk (M)

is the Wu class. But, as TM is stably trivial, its Stiefel-Whitney class satisfies
w(TM) = 1, which implies that v(M) = 1 by the Wu formula. This proves (a)
and, thus, Eqs. (10.6.5) and (10.6.6) imply (b).
For Point (c), suppose that q(a) = 0 and let F (a) be an immersion in general
position. The self-intersection SI(0 ) then consists of an even number of double
492 10 Miscellaneous Applications and Developments

points. These points can then be pairwise eliminated by the Whitney procedure as
explained above. 
By Point (b) of Proposition 10.6.5, qM : Hk (M) Z2 is a quadratic form asso-
ciated to the absolute intersection form of M. If Bd M is either empty or a Z2 -
homology sphere, the intersection form is non-degenerate (see Proposition 5.3.11).
Its Arf invariant Arf(qM ) is thus defined and is called the Kervaire invariant c(M)
of M. The case where Bd M is a Z2 -homology sphere is too general for our purpose
so we introduce the following definition: a compact manifold M is almost closed if
Bd M is either empty or a homotopy sphere.
Proposition 10.6.6 Let m = 2k 6 (k odd). Let M0 and M1 be two (k 1)-
connected stably framed almost closed manifolds of dimension m. If M0 and M1 are
stably framed cobordant, then c(M0 ) = c(M1 ).
Proof Let W0m+1 be a stably framed cobordism between M0 and M1 . If Bd M0 (and
thus Bd M1 ) is empty, we remove a tube Dm I out of W0 , getting a stably framed
manifold W m+1 whose boundary is the connected sum M = M0 M1 . If Bd M0
(and thus Bd M1 ) is a homotopy sphere, we set W = W0 and M = Bd W . The
manifold M is a (k 1)-connected stably framed closed manifold and, clearly,
c(M) = c(M0 ) + c(M1 ). It thus suffices to prove that c(M) = 0. By surgery
below the middle dimension (see Proposition 10.6.4), we may assume that W is
(k 1)-connected. To prove that c(M) = 0, it is enough to show that q(B) = 0
where B = ker(Hk (M) Hk (W )). Indeed, Proposition 5.3.9 and Kronecker dual-
ity imply that 2 dim B = dim Hk (M). The vanishing of q(B) thus implies, using
Proposition 10.6.1, that c(M) = Arf(q) = maj(q) = 0.
Let h and h denote the integral (co)homology. As M is (k 1)-connected and
almost closed, the Hurewicz theorem, the integral Poincar duality and the universal
coefficient theorem imply that

k (M) hk (M) hk (M, Bd M) hk (M) hom(hk (M); Z) . (10.6.7)

Therefore, all the groups in (10.6.7) are free abelian and the isomorphism
 hk (M) 
hk (M) sends BZ = ker(hk (M) hk (W )) onto BZ = Image hk (W ) hk (M) .
As hk (M) is free abelian, the same proof as for Proposition 5.3.9 shows that BZ is a
direct summand of hk (M) and rank hk (M) = 2 rank BZ (all groups in the analogue
of Diagram (5.3.7) are free abelian). Hence, BZ is a direct summand ofhk (M). The
homomorphism BZ /2 BZ B is then injective, where B = Image H k (W )

H k (M) . By Proposition 5.3.9, dim Hk (M) = 2 dim B, so B and BZ /2 BZ have the
same dimension. This shows that the homomorphism BZ B is surjective. Consider
the commutative diagram

k+1 (W ) / k+1 (W, M) / k (M) / k (W )

   
hk+1 (W ) / hk+1 (W, M) / hk (M) / hk (W )
10.6 The Kervaire Invariant 493

whose rows are exact. The bijectivities and surjectivity seen on vertical arrows
come from the Hurewicz-Whitehead theorem [179, Chap. 7, Sect. 5, Theorem 9],
since M and W are (k 1)-connected. By five-lemmas arguments, we deduce that
k+1 (W, M) BZ B is surjective.
Let b B. By the above, there is a map : S k M, representing b, which extends
to : Dk+1 W . Using the stable framing of W , the pair if map (, ) determines,
in its homotopy class, a pair of immersion (, (Dk+1 , S k ) (W, M) which
):
we may assume to be in general position. The self-intersection SI( 0 ) is a compact
1-dimensional manifold whose boundary is SI(0 ). As an arc has two ends, one has
q(b) = q (0 ) = 0. 

Example 10.6.7 The sphere S k has a standard stable framing. Let M = S k S k (k

odd), with the product stable framing. The manifold M is (k 1)-connected and,
by the Knneth formula, H k (M) H k (S k ) H 0 (S k ) H 0 (S k ) H k (S k ), with
generators a = [S k ] 1 and b = 1 [S k ]. As M is the boundary of S k Dk+1 and
of Dk+1 S k , one has c(M) = 0 and q(a) = q(b) = 0 by Proposition 10.6.6 and
its proof. As a b = 1, one has q(a + b) = 1. Note that a + b is represented by the
diagonal manifold of S k S k .

Proposition 10.6.5 permits us to define the Kervaire invariant for any stably framed
almost closed manifold M 2k (k odd), as c(M) = c(M  ) where M  is a (k 1)-
connected manifold stably framed cobordant to M. For closed manifolds, this gives
a map

fr
c: 2k Z2 (k odd) . (10.6.8)

fr
which is a homomorphism. Indeed, the sum in 2k may be represented by the con-
nected sum. If M1 and M2 are (k 1) connected stably framed closed manifolds,
then M = M1 M2 is (k 1)-connected and c(M) = c(M1 ) + c(M2 ).

Proposition 10.6.8 Let m = 2k 6 (k odd). Let M m be a stably framed almost

closed manifold. Then, c(M) = 0 if and only if M is stably framed cobordant to a
contractible manifold (if Bd M is not empty) or to a homotopy sphere (if Bd M is
empty).

Proof The if part is obvious since a contractible manifold of a homotopy sphere

is (k 1)-connected and its middle dimensional homology vanish.
The proof of the converse uses the integral (co)homology, denoted by h and h .
By surgery below the middle dimension, we may suppose that M is (k1)-connected.
As seen in (10.6.7), hk (M) is free abelian. Consider the integral intersection form
on hk (M) given by a b = PD(a)  PD(b), [M]Z
(for some choice of a gen-
erator [M]Z Hm (M, Bd M)). This form is unimodular since M is almost closed
(same proof as for Proposition 5.3.11, or see [97, p. 58]). As k is odd, the inte-
gral intersection form is alternate. Hence, hk (M) admits a skew-symplectic basis,
i.e. a basis a 1 , b 1 , . . . , a p , b b such that a i a j = b i b j = 0 and a i b j = ij
494 10 Miscellaneous Applications and Developments

(same proof as that of Lemma 9.8.1, or see [156, IV.1]). Under the isomorphism
hk (M)/2hk (M) Hk (M) the basis {ai , b i } gives a symplectic basis {ai , bi } for the
Z2 -intersection form.
By changing the basis {ai , b i }, we may assume that q(a1 ) = 0. Indeed, this can
be achieved if q(a1 )q(b1 ) = 0 (by exchanging a 1 with b 1 if necessary). Otherwise,
as Arf(q) = c(M) = 0, there exists j = 1 such that q(aj )q(bj ) = 1 (say j = 2). The
basis change of (10.6.2) then does the job.
We are thus in position to perform a stably framed surgery on an embedding
S k Dk M representing a 1 , giving a stably framed manifold M  . Let M0 =
:
M int((S k Dk )), contained in both M and M  .
By excision, hj (M, M0 ) hj (S k Dk , S k S k1 ) vanishes except for j = k where
it is infinite cyclic. As M is (k 1)-connected, the integral homology sequence of
the pair (M, M0 ) yields to the exact sequence

h j
0 hk (M0 ) hk (M) hk (M, M0 ) hk1 (M0 ) 0 (10.6.9)

where j: (M, ) (M.M0 ) denotes the pair inclusion. Since M is almost closed,
Poincar and Kronecker dualities provide the commutative diagram

h j
/ hk (M, M0 ) o
hk (M) hk (S k Dk , S k S k1 )
O O
[M] [S k Dk ]

/ hk (M) h / hk (S k Dk )
hk (M, Bd M)
k k
 

(h )
hom(hk (M); Z) / hom(h (S k Dk ); Z)
k

hk (S k Dk ) hk (M) is injective, so h is surjective.

As hk (M) is free abelian, h :
Hence, h j is surjective and, by (10.6.9), hk1 (M0 ) = 0. Note that hi (M0 ) hi (M) =
0 for i < k and, since k 2, van Kampens theorem applied to

M M0 (S k Dk ) and M0 (S k Dk ) S k S k1

implies that M0 is simply connected. Therefore, M0 is (k 1)-connected. Also,

from (10.6.9), we deduce that hk (M0 ) is free with rank hk (M0 ) = rank hk (M) 1.
From similar considerations with the pair (M  , M0 ), we deduce that M  is (k 1)-
connected and that hk (M  ) is free with rank hk (M  ) = rank hk (M) 2. The above
process may be repeated with M  showing that, thanks to a finite number of stably
framed surgeries, M is stably framed cobordant to M which is k-connected. By
= 0 if m1. If Bd M is not empty, then h (M)
Poincar duality, h (M) h (pt)
is simply connected, M
and, as M is contractible by the Hurewicz-Whitehead theorem
[82, Proposition 4.74]. If M is closed, then h (M) h (S m ) and by the Hurewicz
10.6 The Kervaire Invariant 495

hm (M)
theorem, m (M) Z. If : S m M represents a generator of m (M),

then is a homotopy equivalence by the Hurewicz-Whitehead theorem. Hence M is
a homotopy sphere. 
Corollary 10.6.9 Let m = 2k 6 with k odd. Let M m be a compact stably framed
manifold whose boundary is a homotopy sphere. Suppose that c(M) = 0. Then
 = Bd M is diffeomorphic to the standard sphere S m1 .
Proof By Proposition 10.6.8, the homotopy sphere  is also equal to Bd M where M

is contractible. Since  is simply connected, it is a consequence of the h-cobordism
that M is diffeomorphic to Dm (see [151, Proposition A, p. 108]). Hence,  is
diffeomorphic to S m1 . 
We have so far presented the stably framed surgery under the point of view of
Wall [204], using the theory of immersions. An earlier approach was introduced
by Milnor in [148] and developed in [24, 113] (for a presentation of the Kervaire
invariant in this framework, see [120, 133]). With this method, in order to per-
form a stably framed surgery on a class a j (M m ), we first represent a by an
embedding : S j M (this is possible when m > 2j and when m = 2j if M is
simply connected [77]). The stable framing of M gives a trivialization of N TM
where N is the trivial bundle of rank N (N large). We thus get a trivialization F of
N TM N TS j . The vector bundle 1 TS j is the restriction of the
tangent bundle to Rj+1 and therefore has a canonical field of orthonormal frames.
Thus, N TS j = 1 N1 TS j has a canonical field of orthonormal frames.
Together with the above trivialization F, this gives an element []
of the homotopy
group j (Stief(j + N, Rm+N )) of the Stiefel manifold Stief(j + N, Rm+N ) which is
shown to be the obstruction to perform a stably framed surgery on the class a. Homo-
topy groups of Stiefel manifold are known (see [181, Theorem 25.6]). For 2j < m,
j (Stief(j + N, Rm+N )) = 0, whence the surgery below the middle dimension. For
m = 2k with k odd, one has k (Stief(k + N, R2k+N )) = Z2 . This gives the quadratic
form q: Hk (M) Z2 , by q(a) = []. Consider the principal bundle
 
P = SO(k) SO(2k + N) Stief(k + N, R2k+N )

and its homotopy exact sequence

j i
k (SO(2k + N))
k (Stief(k + N, R2k+N ))
k1 (SO(k))
k1 (SO(2k + N)) .

Changing the stable framing of M adds to q(a) an element in the image of j . The
SO(k)-principal bundle P is the bundle of orthonormal frames in k . Hence,
(q(a)) k1 (SO(k)) k (BSO(k)) classifies . If k is odd and k = 1, 3, 7,
then ker i Z2 (see [24, Corollary IV.1.11]) and thus is injective. This proves
the following result.
Lemma 10.6.10 Let M be a (k 1)-connected 2k-dimensional stably framed
manifold, with k odd and k = 1, 3, 7. Then, a class a Hk (M) satisfies q(a) = 0 if
and only if it is representable by a k-sphere in M with trivial normal bundle.
496 10 Miscellaneous Applications and Developments

The Kervaire invariant has several other definitions (see, e.g. [23, 123] and also the
proof of Theorem 10.6.12). These more homotopic descriptions were much used for
computing the image of the Kervaire invariant c: m fr m Z2 , an outstanding
S

problems in stable homotopy theory. One of the main advance was due to Browder
[23]. An almost complete solution (except for m = 126) was provided by Hill,
Hopkins and Ravenel in 2009, who proved the following theorem (see [92]).

Theorem 10.6.11 Let m = 2k with k odd. Then, the Kervaire invariant c: m fr Z2

is surjective if m = 2, 6, 14, 30, 62 and possibly 126. In all other dimensions, the
Kervaire invariant vanishes.

The vanishing of the Kervaire invariant in dimension m implies the existence

of a closed PL-manifold K m which does not have the homotopy type of a closed
smooth manifolds. Below is a description of the construction of K m , originally due
to Kervaire [114].
Let pi : Di S k (i = 1, 2) be two copies of the unit disk bundle associated
to the tangent disk bundle to S k (k 3). Let A be a closed k-disk in S k . Choose

trivializations i : p1
i (A)
ADk Dk Dk over A (they are unique up to isotopy:
see Lemma 9.1.2). Let K0 be the quotient space of D1 D2 under the identification
1 (x, y) = 2 (y, x). After rounding the corners (see [17, Appendix, Theorem 6.2]),
K0 is a smooth compact manifold of dimension m = 2k, with boundary 0 = 0m1 .
This is an example of the so called plumbing technique (see [97, Sect. 8], [120]).
The boundary manifold 0 is a homotopy sphere. Indeed, K0 is (k 1)-connected
and hk (K0 ) is free abelian. The integral intersection form clearly induces an iso-

morphism hk (K0 ) hom(hk (K0 ), Z). By the analogue for the integral homology
of Proposition 5.3.11, we deduce that h (0 ) h (S m1 ) (compare [97, p. 58]).
Moreover, K0 is a thickening of S k S k . By general position (since k 3), one
has 1 (0 ) 1 (K0 ) = 1. Hence, 0 is a homotopy sphere (see the proof of
Proposition 10.6.8).
We claim that c(K0 ) = 1. Indeed, Hk (K0 ) has a symplectic basis a, b represented
by the two copies of S k . Recall that TS k is isomorphic to the normal bundle of the
diagonal sphere in S k S k (see Lemma 5.4.14). It then follows from Example 10.6.7
(or from Lemma 10.6.10 if k = 1, 3, 7) that qK0 (a) = qK0 (b) = 1. Therefore
c(K0 ) = 1.
The smooth structure on K0 determines a unique PL-structure (see (3) on p. 203).
The homotopy sphere 0 is PL-isomorphic to the standard sphere by Smales theorem
[177]. Let K be the PL-manifold obtained by gluing to K0 the cone over 0 . The
homotopy sphere 0 = 0m1 is called the Kervaire sphere while the PL-manifold
K = K m is called the Kervaire manifold.

Theorem 10.6.12 Let m = 2k 10 with k odd. Then the following assertions are
equivalent.
(a) The Kervaire invariant c: m
fr Z2 vanishes.
(b) The Kervaire sphere 0m1 is not diffeomorphic to the standard sphere.
10.6 The Kervaire Invariant 497

(c) The Kervaire manifold K m does not have the homotopy type of a smooth closed
manifold.
Thus, according to Theorem 10.6.11, (b) and (c) are true for 10 m = 14, 30, 62
and possibly 126. Theorem 10.6.12 goes back to Kervaire [114] who proved it in
1960 for m = 10, constructing the first example in history of a topological closed
manifold not admitting any smooth structure (even up to homotopy type). Together
with the discovery by J. Milnor in 1956 of several smooth structures on the 7-
sphere, Kervaires result, was quite influential in the history of differential topology
(see [53]).
Proof (c) (b). Suppose that (b) is not true, so there is a diffeomorphism
: S m1 0 . Then K0 Dm is a smooth closed manifold which is homeo-
morphic to K. This contradicts (c).
(b) (a). If (a) is not true, there is a closed smooth stably framed manifold N m with
c(N) = 1. Let N0 be N with an open m-disk removed. Thus, N0 is a smooth almost
closed stably framed manifold with c(N0 ) = 1. Let P be the boundary connected
sum of N0 and K0 . The boundary of P is 0  S m1 , diffeomorphic to 0 . But P is a
smooth almost closed stably framed manifold with c(P) = c(N0 ) + c(K0 ) = 0. By
Corollary 10.6.9, 0 is diffeomorphic to S m1 .
(a) (c). This is more complicated and we just sketch the idea of the proof.
Suppose that (c) is not true, so there is a smooth closed manifold M and a homotopy
equivalence f : K m M. Hence, M is of dimension m and is (k 1)-connected. Let
x be the cone point of K and y = f (x). Let D be an open m-disk in M around y and let
M0 = M D. Using boundary collars in K0 and M0 , one can construct continuous
maps of pairs fK : (K0 , 0 ) (K, x) and fM : (M0 , S m1 ) (M, y). These maps
induce isomorphism

H f
Hk (K) / Hk (M)

H fK H fM (10.6.10)
 

Hk (K0 ) o
Hk (M0 )

where  is defined to make the diagram commutative. For the sake of this proof,
an m-dimensional relative smooth manifold (X, Y ) is called acceptable if X is (k
1)-connected and H i (X, Y ; G) = 0 for k < i < n for all coefficient groups G.
The above pairs (K, x), (M, y), (K0 , 0 ) and (M0 , S m1 ) are acceptable. If (X, Y )
is an acceptable pair, Kervaire and Milnor in [113, pp. 531534] defined a map
X,Y : Hk (X) Z2 (this is the map 0 : Hk (X; Z) Z2 of [113, p. 534] which
descents to Hk (X)). The following two properties hold true.
(i) Let g: (X, Y ) (X  , Y  ) be a continuous map between acceptable pairs. If g
induces an isomorphism H (X  , Y  ; Z) H (X, Y ; Z), then

X,Y H g = X  ,Y  .
498 10 Miscellaneous Applications and Developments

(ii) Let (X, Y ) be an acceptable pair with X stably parallelizable. Then X,Y (a) = 0
if and only if a is representable by an embedded k-sphere with trivial normal
bundle [113, Lemma 8.3 and p. 534].
Another fact is that, as M is homotopy equivalent to K, it is stably parallelizable
(see [25, Theorem A2.1]). Therefore, the quadratic form qM and qM0 are defined.
Since (a) is true and m 10, one has m = 14 by Theorem 10.6.11 and thus
Lemma 10.6.10 applies. Therefore, for a Hk (M0 ) Hk (M), one has

qM (a) = qM0 (a)

= M,M0 (a) by (ii) and Lemma 10.6.10
= K,K0 (a) by (10.6.10) and (i)
= qK0 (a) by (ii) and Lemma 10.6.10.

Hence c(M) = maj(qM ) = maj(qK0 ) = 1, which contradicts Assertion (a). 

Besides giving the Kervaire invariant for a stably framed manifold, the Arf invari-
ant of a quadratic form determines the surgery obstruction group L2 () Z2 for
of order 2 [24, 203]. It has also applications in classical and high dimensional
knot theory (see [205] for a survey on these works).

10.7 Exercises for Chapter 10

10.1. Which closed surfaces admit a involution with scattered fixed point set?
10.2. Let X be a conjugation space. Prove that : H 0 (X) H 0 (X G ) coincides with
the homomorphism induced by the inclusion X G  X. Deduce that a G-space
Y is a conjugation space if and only if each path-connected component of Y
is a conjugation space.
10.3. Prove that the definition of a conjugation space may be expressed with the
reduced cohomology.
10.4. Let X and Y be two conjugation spaces which are G-equivariantly well
pointed. Prove that X Y is a conjugation space.
10.5. Let Xi (i = 1, 2) be two conjugation spaces which are G-equivariantly well
pointed. Suppose that X2 has finite cohomology type. Prove that X1 X2 is a
conjugation space.
10.6. Let X be a conjugation space which is equivariantly well pointed. Construct
a conjugation space Z such that Z G is homeomorphic to X G S 1 .
10.7. Prove that there is no conjugation space X such that X G is homotopy equivalent
to  k OP2 . [Hint: see Remark 10.2.16.]
10.8. Consider the generic length vector  = (1, . . . , 1, n 2) Rn>0 . Let C =
Cdn () and C = Cdn (). Compute Pt (C ) using Theorem 10.3.17. Compute
H (C ) using the transfer exact sequence or Theorem 10.3.11.
10.9. Let  = (1, 1, 2, 2, 3). Compute Pt (Cd5 ()). What are C25 () and C25 ()?
10.7 Exercises for Chapter 10 499

10.10. For which generic length vector  is the chain space C2n () an orientable
manifold? [Hint: this depends on n and on the lopsidedness lops ().]
10.11. For which generic length vector  does the spatial polygon space N3n () admit
a spin structure?
10.12. Let  = (1 , . . . , n ) be a generic length vector with lops () 2. Prove that
ulops ()2 = 0 in H (C2n ()) HG (C2n ()).
10.13. Do Example 10.4.5 for Gr(2; R5 ).
10.14. Use Theorem 10.5.1 to compute H1 (/2 ) for 1 = 2 = {1}, the center
of  = SU(2). Prove that H1 (/2 ) H (RP3 )[u]. Is that surprising?
Chapter 11
Hints and Answers for Some Exercises

11.1 Exercises for Chapter 2

2.1. Fn contains (n +

 n1)! n-simplexes (by induction on n).
2.2. Pt (Fnk ) = 1 + k+1 t k . By induction on k, starting with Pt (Fn0 ) = 1 + n and
with
n+1 n  n 
bk+1 (Fnk+1 ) = dim Ck+1 (Fn ) bk (Fnk ) = k+2 k+1 = k+2 .

as induction step. For the Euler characteristic, evaluate Pt (Fnk ) at t = 1.

2.3. (a) X is isometric to AB where A = {(0, 0, 0), (0, 1, 0), (1, 1, 1), (1, 0, 1)}
and B = {(1, 0, 0), (1, 1, 0), (0, 1, 1), (0, 0, 1)}. The complex A and B are
both quadrilaterals, and X = A B . Thus X is isomorphic to the double
suspension of B , so |X | S 3 .
2.4. 2, 3, 4}. (Sh()) = 3. Pt (Sh()) =
 = (1, 1, 1, 1, 3): Sh() = {5} F{1,
2+t . 2

 = (1, 1, 3, 3, 3): the maximal simplexes of Sh() are {1, 2, 3}, {1, 2, 4} and
{1, 2, 5}. (Sh()) = 1. Pt (Sh()) = 1.
 = (1, 1, 1, 1, 1): Sh() is the 1-skeleton of F{1, 2, 3, 4, 5}. (Sh()) = 5.
Pt (Sh()) = 1 + 6t.
2.5. a = {{2z} | z Z} and a = {{2z + 1} | z Z}.
2.10. By definition of a pseudomanifold, there exists a sequence = 0 , . . . , m =
 of n-simplexes such that, for i 1 < m, i and i+1 have an (n 1)-face
in common, called i . Then, a = {0 , . . . , m1 } satisfies (a) = {,  }.
2.11. If = (), then = ( + [M]). If also = ( ), then +  Z m (M),
therefore +  = 0 or [M] by Proposition 2.4.4.
2.12. This is the content of Proposition 2.5.7.
2.15. Hn (M) = 0. Indeed, if 0 = a Cn (M), there are a and  S(M) a
(since M is infinite). Using Point (c) of the definition of a pseudomanifold, one
can suppose that  = Sn1 (M). Then a. Thus, Z n (M) = 0.

Springer International Publishing Switzerland 2014 501

J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3_11
502 11 Hints and Answers for Some Exercises

2.17. One has S(K )S(K 2 ) = S(K 1 )S(K 0 ), whence H (K 1 , K 0 ) H (K , K 2 ).

The diagram for (b) is

01
Hr (K 0 ) / Hr (K 1 ) / Hr (K 1 , K 0 ) / Hr 1 (K 0 )

i .
   
j
Hr (K 2 ) / Hr (K ) / Hr (K , K 2 ) / Hr 1 (K 2 )

One checks that M V = 01 i 1 j works as a Mayer-Vietoris connecting

homomorphism.
2.19. By Mayer-Vietoris, Pt (M) = Pt (M1 ) + Pt (M2 ) t n 1.

11.2 Exercises for Chapter 3

3.1. In the triangulation of m I used in the proof of Proposition 3.1.30, the

number of (m + 1)-simplexes  is equal to m + 1. In that used in the proof of
m+1
Lemma 3.1.35, this number is k=1 k!.
3.2. Point (b) is proven by induction on the skeleta of (A, B), using (a) and that
D n = C S n1 . For (c), let X be a contractible space. Thus, there exists x0 X
and a continuous map F : X I X with F(x, 0) = x and F(x, 1) = x0 .
Note that the standard simplex n+1 is homeomorphic to the cone on n .
Therefore, a point of n+1 has coordinates [z, t], where (z, t) n I , with
[z, 1] = [z  , 1] for all z, z  n . A linear map D : Cm (X ) Cm+1 (X )
is then defined, for Sm (X ), by D()([z, t]) = F((z), t). It satisfies
D() = D() + for all Cm (X ) and all m 1. This implies that
Z m (X ) = Bm (X ) for m 1.
3.3. Using the homology sequence of (X, X A): Pt (S 2 A) = 1 + (n 1)t and
Pt (T 2 A) = 1 + (n + 1)t.
3.4. (RP 2 , RP 1 ) and (RP 2 , S) where S is the boundary of a 2-disk in RP 2 .
3.5. Define h(x, t) = x, u(A) = 0 and u(X A) = 1. Then (u, h) is a presentation
of (X, A) as a well cofibrant pair.
3.6. Let M be the Mbius band. Then H1 (BdM) H1 (M) is the zero homomor-
phism, contradicting Lemma 3.3.1 if there were a continuous retraction of M
onto BdM).
3.8. There is a unique smaller arc of great circle joining f (x) to f (x), whence
a homotopy to a map f satisfying f(x) = f(x). This means that f factors
through S n RP n which is of degree 0.
3.9. It is a braid diagram looking locally like
11.2 Exercises for Chapter 3 503

H k (X )L / H k (Z ) / H k+1 (Y, Z )
LLL pp8 PPP nn6 OOO
LL pp PPP nnn OOO
% ppp PP( n n n O'
k (Y ) k+1 (X, ) H7 k+1 (Y )
r9
H NNN H
nn 6 PPP
Z
oo
rr NNN nnn PPPP oo
rrr
N& nnn P( ooo
k
H (Y, Z ) / H k+1 (X, Y ) / H k+1 (X )

3.11. Like for Exercise 2.17.

3.13. If X is countable, so is its cellular chain complex and then H (X ) is countable.
But if, for example, X has infinitely many connected components, then H 0 (X )
is not countable by Corollary 3.1.12.
3.14. There is a retraction  rn : B Cn sending  B Cn to  0. This gives a
homomorphism r = Hrn : H1 (B) n=1 H 1 (C n ) =
n=1 Z2 . Divid-
ing [0, 1] using the intervals In = [1/(n + 1), 1/n], we can construct various
paths in B so that the  generator of H1 (In , Bd In ) is sent to that of H1 (Cn ) or
to 0. Any element of n=1 H1 (C n ) may thus be realized, proving that r is
surjective.
For (b), let Rk be the set of homotopy classes of maps from B to RP sending
Cn onto  a point for n > k. This set is finite. As RP is locally contractible,
the map
n=1 Rn [B, RP ] is surjective, which proves (b).
If B has the homotopy type of a CW-complex, then H 1 (B) [B, RP ]
by Proposition 3.8.3. As the latter is countable by (b), this contradicts (a) by
Kronecker duality.
3.15. Let Bq be a Z2 basis ofRq and let X q be a bouquet of q-spheres indexed by
Bq . The bouquet X = q1 X q satisfies Hq (X ) Rq .
3.16. Since 1 m + n = (X  ) = 1 b1 (X ) + b2 (X ). 
3.18. Note that H1 (X p ) 1 (X P )/[1 (X P ), 1 (X P )] Z2 . Thus, b1 (X P1 ) =
b1 (X P2 ) = 1 and b1 (X P3 ) = 0. As (X P1 ) = (X P2 ) = 0 and (X P3 ) = 1,
we get Pt (X P1 ) = Pt (X P2 ) = 1 + t and Pt (X P3 ) = 1.
3.20. Since X is not contractible, r j is not homotopic to a constant map, and nor
is j. One deduces that H n (X ) Z2 and then j r is homotopic to the identity
by Exercise 3.19.

11.3 Exercises for Chapter 4

4.5. As in Sect. 4.3.3, one defines a transfer homomorphism tr H (X ) H ( X )

by sending each singular simplex : m X to the set of its liftings into X .
As the number of these liftings is odd, one has H p tr = id. Hence H p is
surjective and H p is injective by Corollary 2.3.11.
4.6. Use Proposition 4.2.3.
4.7. From T to K , there is no map of degree one by Exercise 4.6. The same kind
of argument shows that there is no map of degree one from S 1 S 2 to RP 3
(the map a  a 3 is non trivial in H (RP 3 ) while it is trivial in H (S 1 S 2 )).
504 11 Hints and Answers for Some Exercises

For the other directions, let f : K T be a map. By Proposition 4.3.10,

H 1 f cannot be injective which, given the ring structure of H (T ), implies that
f is not of degree one. For a map g : S 1 S 2 RP 3 , we see above that
H 1 f = 0. Thus, the composition of g with the inclusion j : RP 3 RP is
homotopic to a constant map by Proposition 3.8.3. But H 3 j is an isomorphism
by Proposition 4.3.10, so g is of degree 0.
4.8. M is the a connected sum of M and RP n . By Propositions 4.2.1 and 4.3.10,
H (M)[a]/(a m + [M] ).
one has H ( M)
4.9. By Sect. 4.2.4 and Proposition 4.2.1, one has

H ((S 1 S 1 )  RP 2 ) Z2 [a, b, c]/(a 2 , b2 , c3 , ab + c2 , ac, bc) ,

with a, b and c of degree 1, and

H (RP 2  RP 2  RP 2 ) Z2 [x, y, z]/(x 3 , y 3 , z 3 , x 2 +y 2 , y 2 +z 2 , x y, x z, yz)

with x, y and z of degree 1. An isomorphism is given by a  x + y, b  x + z

and c  x + y + z.
4.11. Use Proposition 4.7.11 and (4.7.8).
4.12. (a) H (X ) Z2 [x1 , . . . , xn ], with xi of degree 1 (Knneth theorem). For (b)
and (c), set H (CP 2 ) Z2 [a]/(a 3 ) and H (CP 3 ) Z2 [b]/(b4 ), with a and
b of degree 2. Then
(b) As a Z2 -vector space, H (Y ) admits the basis ab, a 2 b, ab2 , ab3 , a 2 b2 , a 2 b3
(see Proposition 4.7.11; we write x y for the reduced cross product of
x and y). The only non trivial cup products are (ab)2 = a 2 b2 and
(ab)(ab2 )) = a 2 b3 .
(c) H (Z ) H (Y ) (see p. 172). All cup products vanish.
4.15. The space X = RP m RP n is a connected CW-complex of dimension m + 1,
thus cat (X ) m + n + 1 by Proposition 4.4.1. By Proposition 4.3.10 and
the Knneth theorem, H (X ) Z2 [a, b]/(a m+1 , bn+1 ). Thus, a m bn = 0,
so nil H >0 (X ) m + n + 1. Proposition 25.2 thus implies that cat (X ) =
m + n + 1.
4.16. cat (T n ) = n + 1 (same arguments as for Exercise 4.15).
4.17. For continuous maps f : X X  and g : Y Y  , the diagram

H (X Y ) / H (X ) H (Y )

H ( f g) H f H g


H (X  Y  ) / H (X  ) H (Y  )

is commutative. This is proven using Formula (4.6.8).

11.3 Exercises for Chapter 4 505

4.18. Using the Knneth formula,  the condition Pt (X ) = Pt (Y ) amounts to the
polynomial equality ri=1 (1 + ai t) = sj=1 (1 + b j t). Therefore r = s
and k (a1 , . . . , ar ) = k (b1 , . . . , br ) for k = 1 . . . , r , where k is the kth
elementary symmetric function. This implies the polynomial equality


r 
r
(x ai ) = (x bi ) ,
i=1 i=1

which implies that bi = a(i) .

4.19. For u H (X ) and v H (Y ), one has

u v, (a  ) (b  ) = u, a  v, b   by (4.6.7)

= u  a, v  b, 
= (u  a) (v  b),  by (4.6.7)
= (u  a) (v  b), 1 ( ) by (4.6.8)
= (u v)  (a b), 1 ( ) by Remark 4.6.4
= u v, (a b)  1 ( ) 
= u v, (a b)  1 ( )  by (4.6.8).

As the elements u v generate H (X ) H (Y ), this proves the formula.

4.20. Using Lemma 4.7.2 and Eq. (4.6.8).
4.21. Let i : E 0 E and j : (E, ) (E, E 0 ) denote the inclusions. The map
H i and H j are morphisms of H (B)-algebras. To see that : H (E 0 )
H +1 (E, E 0 ) is a morphism of H (B)-modules, let b H r (B) and v
H s (E, E 0 ). Then,

(H p0 (b)  v) = (H i H p(b)  v) = H p(b)  (v) ,

holds true in H r +s (E, E 0 ), thanks to the singular cohomology analogue of

Lemma 4.1.9.
4.22. If d is odd, then (Fd (K )) = 1 (K ). If d is even, (Fd (K )) = 1 + b(K ),
where (K ) is the total Betti number of K . This may be deduced from Corollary
4.7.52.
4.23. The Gysin sequence implies (a) and (b). Actually, e = 0 implies that the
restriction homomorphism H (E) H () is surjective (see Proposition
4.7.35), so the isomorphism of (b) is an isomorphism of H (S m )-modules
by the Leray-Hirsch theorem. If n > 2, is a GrA-isomorphism for degrees
reasons. The hypothesis n > 2 in (c) is necessary (see Remark 4.7.34).
4.24. Via a Riemannian metric, we identify D() with a closed tubular neigh-
bourhood of Q in M. For x Q, let (Dx , Sx ) (Dr , Sr 1 ) be the fiber
of (D(), S()) over x. The pair inclusions (Dx , Sx )  (D(), S()) 
(M, M Q) gives rise to the commutative diagram
506 11 Hints and Answers for Some Exercises

H r 1 (M Q) / H r 1 (S()) / H r 1 (Sx )

.
  
H r (M, M Q) / H r (D(), S()) / / H r (Dx , Sx )

The left vertical arrow is surjective since H r (M) = 0. Thus, H r 1 (S())

H r 1 (Sx ) is onto, which implies that e() = 0 by Proposition 4.7.35. Compare
Exercise 5.11.

11.4 Exercises for Chapter 5

5.1. Let X and Y be two homology manifolds and let (x, y) X Y . As (X (Y

y)) ((X x) Y ) = X Y (x, y), one can use a relative Knneth theorem
like Theorem 4.6.10 (see the comments after this theorem for the hypotheses
that we use, i.e. that (X (Y y), (X x) Y ) being excisive in X Y ,
which is true since they are both open).
5.4. It is enough to prove it for a manifold X connected with Y = Bd X = .
If n = dim X , Theorem 5.3.7 implies that Hn (X ) H 0 (X, Y ) = 0 and
Hn (X, Y ) H 0 (X ) = Z2 . Therefore, Hn1 (Y ) Hn1 (X ) is not injective
by the homology sequence of (X, Y ), which would be the cases if there were
a retraction from X onto Y (see Lemma 3.3.1).
5.5. Only the connected component of the point plays a role, so we can suppose
that M is connected. Instead of removing a point, we remove a closed n-disk
D around it. Thus M = M0 D, where M0 = M int D. As Hn (M) Z2
Hn (M, M0 ), the result follows from the homology sequence of (M, M0 ).
5.6. Since [N ] = [N1 + [N2 ] = [M] in Hn1 (M). If N2 = , then H f ([N1 ]) = 0.
5.8. There exists no degree one map from m to  n by Propositions 5.2.8 and 4.2.3.
The same arguments show that if there is a degree one map from  n to m ,
then m [(n 1)/]. The converse is true, using that RP  RP  RP 2 and
2 2

(S 1 S 1 )  RP 2 are homeomorphic (see [136, Lemma 7.1]).

5.9. No, because H f would be injective by Proposition 5.2.8, and, by the Knneth
theorem, there is no element a H 1 (M) with a m = 0.
5.10. Follows from Corollary 5.4.13.
5.11. Let be the normal bundle to Q. That H i([Q]) = 0 implies that the Poincar
dual PD(Q) of Q vanishes, whence e() = 0 by Lemma 5.4.4. Compare
Exercise 4.24.
5.13. Let A = PD(Q 1 ) PD(Q 2 ). For dimensional reasons, [M1 M2 ] =
1 ([M1 ] [M2 ]) and [Q 1 Q 2 ] = 1 ([Q 1 ] [Q 2 ]). Therefore,
11.4 Exercises for Chapter 5 507
   
A  [M1 M2 ] = PD(Q 1 ) PD(Q 2 )  1 ([M1 ] [M2 ])

= (PD(Q 1 )  [M1 ]) (PD(Q 2 )  [M2 ]) by (4.8.1)

= H i 1 ([Q 1 ]) H i 2 ([Q 2 ])

= (H i 1 H i 2 )([Q 1 ] [Q 2 ])
 
= H i([Q 1 Q 2 ]) by Exercise 4.17

where i : Q 1 Q 2 M1 M2 denotes the inclusion. As is an isomorphism,

this proves (5.5.1).
5.14. By Exercise 5.13, PD({x} M  ) = PD({x}) PD(M  ) = [Mx ] 1, where
Mx is the connected component of x in M.
5.18. Let a Q. The map associating to (t, x) I Q  the point ( 1 t 2 a, t 2 x)
parameterizes a smooth (q  + 1)-disk Da in , with boundary Q  , which
intersects Q transversally at a. By Proposition 5.4.22, l(Q, Q  ) = 1.

11.5 Exercises for Chapter 6

6.1. KP n may be covered by n + 1 chart domains which are contractible (see

(6.1.1) for K = C; the same formalism works for K = H). On the other hand,
nil H >0 (KP n ) = n + 1, thus cat (KP n ) = n + 1 by Proposition 4.4.2, like
for K = R: see Corollary 4.4.3).
6.3. Follows from Proposition 6.1.11.
6.4. The two parts of CC f give rise to the Mayer-Vietoris sequence

/ H k (CC f ) / H k (Y ) H k (Y ) / H k (X ) / H k+1 (CC f ) /

where = H f + H f . For f = j, this sequence together Proposition 6.1.11

gives the Poincar series of CC j :


1 + t3
Pt (CC j ) = t 4k + t 4k+3 = .
1 t4
k=0 k=1

Actually, one has a GrA-isomorphism H (CC j ) Z2 [a, b]/(b2 ) where a is

of degree 4 and b of degree 3 (see Exercise 7.20).

6.5. By the Knneth theorem and Proposition 4.2.1, H (X ) Z2 [a, b] (a 2 , b2 )

and H (X ) Z2 [c, d] (c3 , d 3 , cd, c2 + d 2 ) (generators in degree 4). Thus,

Pt (X ) = Pt (Y ) = 1 + 2t 4 + t 8 . But the cup-square map vanishes in H (X )
while not in H (Y ). (If the degree of the generators are set to 1, we get the
cohomology of the torus for X and of the Klein bottle for Y ).
508 11 Hints and Answers for Some Exercises

6.6. Use the Hopf vector bundle K over KP 1 . For n = 2k, one can take =
C C (k times) over (CP 1 )k . For n = 2k +1, take = C C R
over (CP 1 )k RP 1 .
6.7. There exits a map f : X S n such that a = H f () (see the proof of
Proposition 3.8.1). One can then take = f K where K is the vector Hopf
bundle over KP 1 for K = R, C, H and O.
6.10. Follows from Exercise 4.24.
h f
6.13. The maps h and f extend to maps on mapping cones Cg h Cg and Cg
C f g . The formula follows from the functionality of the cup product.
6.14. Let y  S m be another regular value for f . We may suppose that j is transversal
to Q  = f 1 ({y  }). Then,

Hopf ( f ) = l(Q, Q  ) by Proposition 6.3.7

=  j 1 (Q  ) mod 2 by Proposition 5.4.22
1 
=  ( f j) ({y }) mod 2
= deg( f j) as in Proposition 3.2.6.

11.6 Exercises for Chapter 7

7.2. Follows from Corollary 7.1.17. Note that this is wrong for p = 0 (see Example
7.1.16).
7.3. As X is equivariantly formal, there is a GrV[u]-isomorphism between H (X G )
and Z2 [u][B], where B is a GrV basis of H (X ). Therefore,

Pt (X ) t Pt (X )
Pt (X G ) = and Pt (u H (X G )) = .
1t 1t

7.6. By Proposition 7.1.12, : H Gd (X ) H d (X )G is an isomorphism and, as

d 1, H d (X )G is the diagonal subgroup of H d (X ) Z2 Z2 . By
Sequence (7.1.8), there exists 0 = a HGd (X ) and
Ann (u) = Z2 a.
Sequence (7.1.8) also implies that HG (X ) Z2 [u, a] (ua). If a 2 = 0, then
a 2 = u 2d which would contradict the equation ua = 0; indeed ua 2 = u 2d+1 =
0 by Corollary 7.1.5.
7.7. The pair (SU (2), ) is isomorphic to (S 3 , S 1 ), the groups of units quaternions
and complex numbers. For a topological group, let E n be the join of n
copies of , on which acts diagonally. Then, for n 1, E n (S 3 )/S 3
S 4(n1)+3 /S 3 HP n1 and E n (S 1 )/S 1 S 4(n1)+1 /S 1 CP n1 . Thus,
there are homotopy equivalences B S 3  E S 3 /S 3  HP and B S 1 
E S 1 /S 1  CP .
11.6 Exercises for Chapter 7 509

7.8. One check that (X Y ) X Y . The result then follows from the Knneth
theorem 4.6.7. One can also use the equivariant Knneth theorem 7.4.3 with
an additional hypothesis on .
7.9. The hypotheses of the Knneth theorems 4.6.7 and 7.4.3 are satisfied (see the
end of Remark 7.4.4). Therefore, there is a commutative diagram

H (X ) H ( pt) H (Y ) / H (X Y )


X Y X Y
 
/ H (X Y )
H (X ) H (Y )

As X and Y are equivariantly formal, X Y is surjective and thus, X Y is

surjective.
7.10. By Exercise 7.9 and (7.1.16), one has a GrA[u]-isomorphism

HG (X ) HG (S01 ) H (Y ) Z2 [u, a] (a 2 +ua) H (Y ) H (Y )[a] (a 2 +ua)

with a of degree [Link] Mayer-Vietoris sequence in question is equivalent to

0 / H (X ) r / H (X G ) J / H (Y ) / 0
G G

with the GrA[u]-isomorphism HG (X G ) H (Y )[u 1 ] H (Y )[u 2 ]. One has

J (b, c) = b + c for b, c H (Y ) (proving that J is onto) and J (u 1 , 0) =
J (0, u 2 ) = 0. One has r (u) = (u 1 , u 2 ), r (b) = (b, b) for b H (Y ) and
r (a) = (0, u 2 ).

7.11. By the equivariant Knneth theorem 7.4.3, HG (X ) Z2 [u, a, b] (a 2 +

ua, b2 + ub) with a and b of degree 1. The critical points of f , labeled by
there value, are x3 = (1, 1) (index 0), x1 = (1, 1) and x1 = (1, 1)
(index 1) and x3 = (1, 1) (index 2). Therefore

HG (X G ) Z2 [u 3 ] Z2 [u 1 ] Z2 [u 1 ] Z2 [u 3 ] .

One has r (u) = (u 3 , u 1 , u 1 , u 3 ), r (a) = (0, u 1 , 0, u 3 ) and r (b) =

(0, 0, u 1 , u 3 ), whence r (ab) = (0, 0, 0, u 23 ).
7.12. X = KP 2 (K = R, C, H or O) with the involution (x0 : x1 : x2 ) = (x0 :
x1 : x2 ). Related to this exercise is Exercise 8.9.
7.13. One has (Ppn )G = Pp Pn p1 . The total Betti number thus satisfy b(Ppn ) =
p n p1

b((Ppn )G ). This proves (a) by Proposition 7.3.7.

For (b), set H (Pp ) = Z2 [u, b1 ]/(b1 ), H (Pn p1 ) = Z2 [u, b2 ]/(b2 )
p p+1 n p1 n p

and H (Ppn ) = Z2 [a]/(a n+1 ). Thus, as a Z2 -module, HG (Ppn ) is generated

by the powers of an element A Hg1 (Ppn ) such that (A) = a. There are
two such elements, A and B, with the relation B = A + u. One has r (A) =
510 11 Hints and Answers for Some Exercises

(b1 + u, b2 ) or r (A) = (b1 , b2 + u). (use that r : H (Ppn ) H ((Ppn )G )

satisfy r(a) = (b1 , b2 ); but (b1 , b2 )n = 0 in HGn ((Ppn )G ) and (b1 +u, b2 +u)n
(u), contradicting (An ) = a n = 0). By exchanging A and B if necessary, one
may suppose that r (A) = (b1 + u, b2 ).
For (c), one uses that r (A) = (u, b2 ) (see (b)). Hence, r (Ak ) = (u k , 0) for k
n 1. Therefore, the relation r (An+1 ) = r (u An ) holds true in HG ((Ppn )G ). As
r is injective, the relation An+1 = u An holds true in HG (Ppn ). Therefore, there

is a surjective GrA[u]-morphism h : Z2 [u][A] (An+1 + u An ) HG (Ppn ).

n+1
But Z2 [u][A] (A + u An ) and HG (Ppn ) (which is a free Z2 -module over
{A, . . . , A }) have the same Poincar series. Hence, h is an isomorphism. The
n

second presentation follows by substituting B + u to A. Point (d) is proved in

the same way.
7.14. The isomorphism is given by A  B + u, with inverse given by B  A + u.
Note that R and S are the two presentations for HG (P02 ) given in Exercise
7.14 (c). Take X = HP 2 with (x0 : x1 : x2 ) = (x0 : x1 : x2 ).
7.16. If we consider S 1 acting on CP n by

g (x0 : : x p : p+1 : : xn ) = (gx0 : : gx p : x p+1 : : xn ) ,

the statements the are all the same, the degree of each cohomology classes
being doubled.
7.18. Write  X for  X . We see X as the union of two cones C X and C+ X
glued over their common base X . The homomorphism  comes from the
composition

/ H +1 (C X, X ) o / H +1 ( X )
H (X ) H +1 ( X, C+ X )

The homomorphism comes from the cohomology sequence of the pair

((C X ) , X ). As C X is -equivariantly contractible, the homomorphism
descends to an injection : H (X ) H +1 (C X, X ). The middle arrow is
induced by the pair inclusion (C X, X ) ( X, C+ X ) and is an isomorphism
by excision on the Borel constructions. By (7.2.10), the last arrow becomes an
isomorphism when composed by H +1 ( X ) H +1 ( X ). Whence the ho-
momorphism  which is injective. That  is a morphism of H ( pt)-module
comes from Exercise 4.21 applied to the bundle pair ((C X ) , X ) over B .
The morphism  is functorial in both X and in . In particular, the trivial
homomorphism {1} gives the commutative diagram


H (X ) / H +1 ( X )

 . (11.6.1)
 
 / H +1 ( X )
H (X )
11.6 Exercises for Chapter 7 511

When = G = X , one has H G (G) = 0 by Lemma 7.1.14 while H G1 (G G) =

Z2 , since G G is G-homeomorphic to the sphere S01 of Example 7.1.4.
For (c), let : H (X ) H (X ) be a section of . Using the commutativity
of Diagram (11.1), one defines a section  of  by  =  ( )1 .
Therefore,  X is -equivariantly formal and, as H ( pt)-module, H +1 ( X )
is generated by Image( ). As Image( ) Image( ), the morphism 
is an isomorphism.
7.19. The G-space  X is a finite dimensional G-complex satisfying b( X) =

b(( X ) ) < (since ( X ) = X ). Therefore,  X is equivariantly for-
G G G

mal by Proposition 7.3.7 and thus r : H G ( X ) H ((

G
X )G ) is injective by
Proposition 7.3.9. As r factor through H G i, we know that H G i is injective. By

Proposition 7.3.7 again, X is itself equivariantly formal and thus is surjective.

As H i = 0, the image of H G i is contained in ker . By (7.1.8), the later is
equal to the ideal (u) generated by u. By comparing the Poincar polynomials
of H G (
X ) with that of (u) (see Exercise 7.3) we deduce that the image of
HG i is actually equal to (u).
7.20. Let  be the subgroup of SU(2) formed by the diagonal matrices and let
+ = {dia(ei , ei ) | [0, ]}. Each conjugation class in has a unique
representative in + , so X/ is homeomorphic to [0, ]. Hence, each class
in E X has representatives of the form (z, b) with b + . We deduce
that X is homeomorphic to (E / [0, ]) (E / {0, }), that is
X CC j0 where j0 is the quotient map E / E / . The latter is
homotopy equivalent to the map j1 : B B induced by the inclusion
 (see Example 7.2.4). By Exercise 7.7, j1 is equivalent to the inclusion
j : CP HP . This proves (a). From the correction of Exercise 6.4 (see
p. 507), we have

(1 + t 3 )
Pt (H (X )) = = Pt (H ( pt)) Pt (X ) .
1 t4

This equality implies that X is -equivariantly formal (see the comment

following Corollary 4.7.20). The unique section H (X ) H (X ) of is
multiplicative for degrees reasons. Hence, the Leray-Hirsch theorem gives the
H ( pt)-algebra isomorphisms of (c).
7.21. (X , (X 1 ) , (X 2 ) , (X 0 ) ) is a Mayer-Vietoris data and X = (X 1 ) (X 2 ) .
By Lemma 7.2.11, ((X i ) , (X 0 ) ) is a well cofibrant for i = 1, 2. The Mayer-
Vietoris is then given by Lemma 3.1.38 and Proposition 3.1.51.
7.22. Let z X Y be the image of x and y. Using the Mayer-Vietoris sequence of
Exercise 7.21, one has a commutative diagram
512 11 Hints and Answers for Some Exercises

/ H (z)  / H (z) H (z) + / H (z) / 0

0
 

  
0 / H (X Y ) / H (X ) H (Y ) / H (z) / 0

where the vertical arrows are induced by the constant maps. Since the right
vertical map is an isomorphism, we check that the quotient of the second row
by the first one is an exact sequence (a particular case of the snake lemma: see
[28, Lemma 3.3]). Thus, the H -algebra homomorphism ( H i X , H i Y ) : H
(X Y ) H (X ) H (Y ) induced by the inclusions is an isomorphism.

11.7 Exercises for Chapter 8

8.1. No, since Sq1 vanishes in H (T ) while not in H ( K ).

8.2. Sq6 (a 5 b7 ) = a 10 b8 + a 9 b9 + a 6 b12 + a 5 b13
(use that Sq(a 5 b7 ) = Sq(a)5 Sq(b)7 = (a + a 2 )5 (b + b2 )7 ).
8.3. Comes form Lemma 8.5.1.
8.5. Follows from the Adem relations. Remark: by [1, Sect. 2.5], any finite set in
A generates a finite subalgebra; hence, all elements of A are nilpotent. For
estimates of nilpotency heights, see [212, Sect. 2.6].
8.6. If a H 1 (RP ) is the generator then, as A-module, H (RP ) admits {a r |
r = 2n1 } as a minimal set of generators. This follows from (8.2.2.). Note: if
b(X ) is infinite, then H (X ) is not finitely generated as an A-module [172].
8.7. Follows from Proposition 8.6.2 and Theorem 8.6.5.
8.8. We may assume that r > 1, otherwise there is nothing to prove. If U is an open
set of M over which p is trivial, then p 1 (U ) is homeomorphic to U Sr 1 ,
which implies that dim M = n r (using Theorem 3.3.4). For r > 1, the
Gysin exact sequence implies that H (M) is a GrA-quotient of Z2 [e], where
e H r (M) is the Euler class of p. Hence, n = kr and, using Exercise 8.7,
r = 1, 2, 4, 8.
8.9. One has n = dim M = 2r by Corollary 5.2.5. By Poincar duality, Pt (M) =
1 + t r + t 2r with non trivial cup-square. By Exercise 8.7, n = 2s with s =
1, 2, 3, 4.
8.10. The map f must have at least two critical points on each connected com-
ponent of M. Since M1 = pt, M is connected. We use the notation
of Sect. 7.6: Mx,y = f 1 ([x, y]), and Mx = f 1 ({x}). There are G-
equivariant diffeomorphisms : (D(1 ), S(1 )) (M1,0 , M0 ) and
+ : (D(1 ), S(1 )) (M0,1 , M0 ), where x is the normal bundle to Mx in
M. The pair (D(1 ), S(1 )) is G-equivariantly diffeomorphic to (D n , S n1 )
endowed with the antipodal involution. We thus have a smooth locally triv-
p
ial bundle Sr 1 S n1 M1 , where r is the codimension of M1 in M.
11.7 Exercises for Chapter 8 513

Since n is odd, Exercise 8.8 implies that r = 1, so p is a 2-fold covering. But p

is G-invariant, so it can be identified with the quotient map S n1 M1 .
Therefore, M1 is diffeomorphic to RP n1 and 1 is the Hopf bundle .
Thus, M is diffeomorphic to D n 1 D() which is homeomorphic to
+
RP n . Note that M may not be diffeomorphic to RP n if the diffeomorphism
1
+ : S
n1 S() = S n1 does not extend to a diffeomorphism of D n .

11.8 Exercises for Chapter 9

9.1. Use that w1 ( ) = w1 () + w1 () = 0 and, if w1 () = 0, that w2 ( ) =

w2 () + w2 () = 0.
9.2. As p is a local diffeomorphism, one has T X = p T X . Hence wi (T X ) =
H p(wi (T X )). But H p is injective by Exercise 4.5. The result follows from
Propositions 9.4.4 and 9.4.7.
9.3. Since p is a local diffeomorphism, one has T M p T M. Hence, H p(w1
(T M)) = w1 (T M) = 0. Thus, 0 = w1 (T M) ker(H p : H 1 (M)

H ( M)) which, by the transfer exact sequence, is generated by w.
1

9.4. Let m = dim M. Proposition 4.2.1 and its proof provide an epimorphism
: H >0 (M1 ) H >0 (M2 ) H >0 (M). For 0 < i < m, is induced by the
inclusions of Mi minus a disk into M. This implies that, for 0 < i < dim Mi ,
wi (M) is the image of (wi (M1 ), wi (M2 ). The same holds true for i = m, using
that (M) (M1 ) + (M2 ) mod 2 together with Corollary 5.4.16.
9.5. The Stiefel-Whitney classes w(T RP 4 ) (see Proposition 9.8.10) and w(T CP 2 )
(see Remark 9.8.12) are incompatible with such a decomposition, given
Formula (9.4.3).
9.6. Given Formula (9.4.1), this follows from Exercise 8.3.
9.9. Using the tautological bundles, as for Proposition 6.1.11. Using that BU (n) is
a conjugation space (see p. 438), this also follows from Corollary 10.2.11.
9.11. H (Gr(2; R5 )) has 10 basis elements: 1 = (1, 2), w1 = (1, 3), w12 = (2, 3) +
(1, 4), w2 = (2, 3), w13 = (1, 5), w1 w2 = (2, 4), w22 = (3, 4), w12 w2 =
(3, 4) + (2, 5) (w14 = w22 + w12 w2 ), w15 = w1 w22 = (3, 5) (w13 w2 = 0),
w22 w12 = (4, 5). The expression in the Schubert symbols were found using
Propositions 9.5.27 and 9.5.32.
9.12. Since M is orientable, Sq1 : H 5 (M) H 6 (M) vanishes (see Corollary
9.8.5). By Wus formula and the Adem relations, w6 (T M) = Sq3 (v3 (M)) =
Sq1 Sq2 (v3 (M)) = 0. The result follows from Corollary 5.4.16. The same
argument works when dim M = 10, using the Adem relation Sq5 = Sq1 Sq4 .
9.13. By Corollary 9.8.5, Sq1 : H n1 (Q) H n (Q) is surjective while Sq1 : H n1
(P) H n (P) vanishes. As Sq1 H n1 f = H n f Sq1 , this proves the
statement under Hypothesis (a). The argument for Hypothesis (b) is similar,
using Corollary 9.8.6.
514 11 Hints and Answers for Some Exercises

9.14. The normal bundle M of M = f 1 ({x}) in S n+k is trivial, since M = f Tx S k .

Thus, M admits an embedding into Rn+k+1 with trivial normal bundle. There-
fore, w(T
M) = 1 and thus w(T M) = 1. The results follows from Thoms
theorem 9.9.7.
9.15. By Thoms theorem 9.9.7, since they do not have the same Stiefel-Whitney
numbers.
9.16. The manifold M is of dimension n. Let a H 1 (M) be the non-zero class. By
Proposition 3.8.3, there exits a map f : M RP such that a = H f (). By
cellular approximation (or by the proof of Proposition 3.8.3), the image of f
is contained in RP n and thus H f : H (RP n ) H (M) is an isomorphism.
The result then follows from Corollary 9.9.10.

11.9 Exercises for Chapter 10

10.1. As usually drawn in R3 , the orientable surface g of genus g admits an axial

symmetry, with a scattered fixed point set, i.e. 2g + 2 = b(g ) fixed points
(this corresponds to a connected sum of g copies of S 1 S 1 with the involution
(z 1 , z 2 ) = (z 1 , z 2 )). A non-orientable surface does not admit any involution
with scattered fixed point set by Proposition 4.2.3 and Corollary 10.1.2 (or by
Corollary 10.1.3 when the involution is smooth).
10.2. The first assertion comes from Diagram (10.2.11) (for a more elementary
argument, see [87, Remark 3.1]). It implies that the H -frame respects the
path-connected component of X , proving the only if part of last assertion.
The if part is obvious.
10.3. Let (, ) be the H -frame of a conjugation space X . The isomorphism
satisfies (1) = 1 by Exercise 10.2 (or by Proposition 10.2.5). Therefore,
descends to a GrV-isomorphism : H 2 (X ) H (X G ). Also, (1) = 1,
so it descends to : H (X ) H G (X ) which is a section of : H G (X )
H (X ). The conjugation equation holds true for (, )
(call (, a H -
)
frame for X ). Conversely, such a H -frame determines an H -frame (it helps

to assume X path-connected, which we can do by Exercise 10.2).

10.4. Use Exercises 7.22 and 10.3.
10.5. By Exercise 10.3, one can use the reduced cohomology. Let ( i , i ) be the
H -frame for X i . Obviously, (X 1 X 2 )G = X 1G X 2G and X 2G is of finite
cohomology type (using 2 ). Using Proposition 4.7.11 and Lemma 7.4.9, one
has commutative diagrams

H 2 (X 1 ) H 2 (X 2 ) / H 2 (X X )
1 2

1 2
 


H (X 1 ) H (X 2G )
G / H (X G X G )
1 2
 11.9 Exercises for Chapter 10 515

and

1 2
H 2 (X 1 ) H 2 (X 2 ) / H 2 (X 1 ) H 2 (X 2 )
G G

G

 
/ H 2 (X 1 X 2 )
H 2 (X 1 X 2 ) G

which define (, ).
Checking the conjugation equation is straightforward.
10.6. Z = X S12 , where S12 is the 2-sphere with the linear involution of Example
7.1.14. This uses Exercise 10.5 .
10.7. In H (X ), Sq16 should not vanish by Proposition 10.2.6. This contradicts
Theorem 8.6.5.
10.8. The two sums in Proposition 10.3.7 have only one term and Pt (C ) = 1 +
t (n2)(d1)1 . Hence, H (C ) H (S (n2)(d1)1 ). By the transfer exact se-
quence, H (C ) is GrA-isomorphic to H (RP (n2)(d1)1 ). Theorem 10.3.11
gives the GrA[u]-presentation HG (C ) Z2 [u]/(u (n2)(d1) ). Actually, ar-
guments like in Example 10.3.9 enable us to prove that C S (n2)(d1)1
and C RP (n2)(d1)1 (see e.g. [85, Example 2.6]).
10.9. By Proposition 10.3.7, Pt (Cd5 ()) = 1+2t d1 +2t 2(d1)1 +t 3(d1)1 . When
d = 2, C25 () is an orientable surface by Corollary 10.3.2. As Pt (C25 () =
1 + 4t + t 2 , C25 () is an orientable surface of genus 2 and thus C25 () is a
non-orientable surface of genus 3.
10.10. Let C = C2n () and C = C2n (). Since  is generic, H (C ) HG (C )
and C is a closed manifold of dimension n 3 (see Corollary 10.3.2). By
Lemma 10.3.3, we may suppose that  is dominated. If lops () 1, C is
empty (see Examples 10.3.21 and 10.3.22). If lops () = 2, then Cdn () is
orientable, being diffeomorphic to (S 1 )n3 (see Example 10.3.23). Suppose
that lops () 3. By (10.3.28), C is connected and, since p : C C is of
degree 0, one has H n3 (C ) = u H n4 (C ) by the transfer exact sequence.
As HG (C ) is GrA-generated in degree 1 (see Theorem 10.3.11), there ex-
ists x1 , . . . , xn4 H 1 (C ) such that x1 xn4 u = 0 in HGn3 (C ) Z2 .
Then Sq1 (x1 xn4 ) = (n 4)x1 xn4 u by Theorem 10.3.7. Note that
the monomials of the form x1 , . . . , xn4 generate H n4 (C ). Using Corollary
9.8.5, we see that, for  generic and dominated, C2n () is non-orientable if and
only if lops () 3 and n is odd.
10.11. By Proposition 10.3.32, N3n () is a conjugation manifold with N3n ()G =
C2n (). Using Proposition 10.2.12, the answer is analogous to that of Exercise
10.10: for  generic and dominated, N3n () has no spin structure if and only
if lops () 3 and n is odd.
10.12. Let C = C2n (). By (10.3.33), the simplicial complex Sh n () (definition p. 459)
satisfies dim Sh n () = n 2 lops (). Therefore, in the face exterior algebra

1 (Sh n ()) (see Sect. 4.7.8), there exists at most n 2 lops () elements
 516 11 Hints and Answers for Some Exercises

whose product is not 0. By (10.3.20), HG (C )/(u)

. As H n3 (C ) = 0,
this implies that u lops ()2 = 0.
10.14. Like in Example 10.5.13, Theorem 10.5.1 gives the presentation H 1 ( / 2 )
Z2 [u, a]/(u 4 + a 4 ), where a and u are of degree one. Setting b = a + u, we
get H 1 ( / 2 ) Z2 [u, b]/(b4 ) which is isomorphic to H (RP 3 )[u]. This is
not surprising since SU(2)/{1} SO(3) RP 3 and that {1} acts trivially
on SU(2)/{1}.
References

1. Adams, J.F.: On the structure and applications of the Steenrod algebra. Comment. Math. Helv.
32, 180214 (1958)
2. Adams, J.F.: On the non-existence of elements of Hopf invariant one. Ann. of Math. 2(72),
20104 (1960)
3. Adem, A., Milgram, R.J.: Cohomology of finite groups. Grundlehren der Mathematischen
Wissenschaften [Fundamental Principles of Mathematical Sciences], vol. 309. Springer,
Berlin (1994)
4. Adem, J.: The iteration of the Steenrod squares in algebraic topology. Proc. Nat. Acad. Sci.
U. S. A. 38, 720726 (1952)
5. Adem, J.: The relations on Steenrod powers of cohomology classes. Algebraic geometry and
topology. A symposium in honor of S. Lefschetz, pp. 191238. Princeton University Press,
Princeton, NJ (1957)
6. Adem, J., Gitler, S., James, I.M.: On axial maps of a certain type. Bol. Soc. Mat. Mexicana
2(17), 5962 (1972)
7. Alexander, J.W.: A proof and extension of the Jordan-Brouwer separation theorem. Trans.
Amer. Math. Soc. 23(4), 333349 (1922)
8. Alexander, J.W.: Some problems in topology. Verh. Internat. Math. Kongr. Zrich, 1, 249257
(1932)
9. Allday, C., Puppe, V.: Cohomological methods in transformation groups. Cambridge Studies
in Advanced Mathematics. 32, vol. 11, pp. 470. Cambridge University Press, Cambridge
(1993)
10. Arf, C.: Untersuchungen ber quadratische Formen in Krpern der Charakteristik 2. I. J. Reine
Angew. Math. 183, 148167 (1941)
11. Atiyah, M.F.: K-theory and reality. Quart. J. Math. Oxford Ser. 2(17), 367386 (1966)
12. Audin, M.: The topology of torus actions on symplectic manifolds. Progress in Mathematics,
vol. 93. Birkhuser Verlag, Basel (1991)
13. Banyaga, A., Hurtubise, D.: Lectures on morse homology. Kluwer Texts in the Mathematical
Sciences, vol. 29. Kluwer Academic Publishers Group, Dordrecht (2004)
14. Beyer, W.A., Zardecki, A.: The early history of the ham sandwich theorem. Amer. Math.
Monthly 111(1), 5861 (2004)
15. Borel, A.: La cohomologie mod 2 de certains espaces homognes. Comment. Math. Helv. 27,
165197 (1953)
16. Borel, A., Haefliger, A.: La classe dhomologie fondamentale dun espace analytique. Bull.
Soc. Math. France 89, 461513 (1961)

Springer International Publishing Switzerland 2014 517

J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3
518 References

17. Borel, A., Serre, J.-P.: Corners and arithmetic groups. Comment. Math. Helv. 48, 436 491.
Avec un appendice: Arrondissement des varits coins, par A. Douady et L, Hrault (1973)
18. Bott, R.: Lectures on Morse theory, old and new. Bull. Amer. Math. Soc. (N.S.) 7(2), 331358
(1982)
19. Bott, R., Tu, L.W.: Differential Forms in Algebraic Topology. Springer, New York (1982)
(vol. 82 of Graduate Texts in Mathematics)
20. Bredon, G.E.: Introduction to Compact Transformation Groups. Pure and Applied Mathemat-
ics, vol. 46. Academic Press, New York (1972)
21. Brcker, T., Tom Dieck, T.: Representations of compact Lie groups. Graduate Texts in Math-
ematics, vol. 98. Springer, New York (1995)
22. Brouwer, L.E.J.: Collected works. In: Freudenthal, H. (ed.) Geometry, Analysis, Topology
and Mechanics, vol. 2. North-Holland Publishing Co., Amsterdam (1976)
23. Browder, W.: The Kervaire invariant of framed manifolds and its generalization. Ann. of Math.
2(90), 157186 (1969)
24. Browder, W.: Surgery on Simply-Connected Manifolds (Springer, NewYork, 1972). Ergeb-
nisse der Mathematik und ihrer Grenzgebiete, Band 65
25. Brown, E.H. Jr., Peterson, F.P.: The Kervaire invariant of (8k + 2)-manifolds. Amer. J. Math.
88, 815826 (1966)
26. Brown, K.S:. Cohomology of groups. Graduate Texts in Mathematics, vol. 87. Springer, New
York (1982)
27. Bullett, S.R., Macdonald, I.G.: On the Adem relations. Topology 21(3), 329332 (1982)
28. Cartan, H., Eilenberg, S.: Homological algebra. Princeton University Press, Princeton, N. J.
(1956)

29. Cech, E.: Multiplications on a complex. Ann. of Math. (2) 37(3), 681697 (1936)
30. Charney, R.: An introduction to right-angled Artin groups. Geom. Dedicata 125, 141158
(2007)
31. Chern, S.: Characteristic classes of Hermitian manifolds. Ann. of Math. 2(47), 85121 (1946)
32. Chern, S.: On the multiplication in the characteristic ring of a sphere bundle. Ann. of Math.
2(49), 362372 (1948)
33. Cohen, D.E.: On the Adem relations. Proc. Cambridge Philos. Soc. 57, 265267 (1961)
34. Constantin, A., Kolev, B.: The theorem of Kerkjrt on periodic homeomorphisms of the
disc and the sphere. Enseign. Math. (2), 40(34), 193200 (1994)
35. Cornea, O., Lupton, G., Oprea, J., Tanr, D.: Lusternik-Schnirelmann category, Mathematical
Surveys and Monographs, vol. 103. American Mathematical Society, Providence (2003)
36. Davis, D.M.: Table of immersions and embeddings of real projective spaces. Available on
Donald Daviss home page at Lehigh University. [Link]
37. tom Dieck, T.: Transformation Groups. Walter de Gruyter and Co., Berlin (1987)
38. tom Dieck, T.: Algebraic topology. In: EMS Textbooks in Mathematics. European Mathemat-
ical Society (EMS), Zrich (2008)
39. tom Dieck, T., Kamps, K.H., Puppe, D.: Homotopietheorie. In: Lecture Notes in Mathematics,
vol. 157. Springer-Verlag, Berlin (1970)
40. Dieudonn, J.: A History of Algebraic and Differential Topology. 19001960. Birkhuser-
Boston Inc., Boston (1989)
41. Dold, A.: Erzeugende der Thomschen Algebra N. Math. Z. 65, 2535 (1956)
42. Dold, A.: Partitions of unity in the theory of fibrations. Ann. of Math. 2(78), 223255 (1963)
43. Dold, A.: Lectures on algebraic topology. In: Classics in Mathematics. Springer, Berlin (1995)
(Reprint of the 1972 edition)
44. Dugundji, J.: Topology. Allyn and Bacon Inc., Boston (1966)
45. Duistermaat, J.J.: Convexity and tightness for restrictions of Hamiltonian functions to fixed
point sets of an antisymplectic involution. Trans. Amer. Math. Soc. 275(1), 417429 (1983)
46. Ebeling, W.: Lattices and codes. Advanced Lectures in Mathematics. Friedr. Vieweg & Sohn,
Braunschweig. A course partially based on lectures by F. Hirzebruch (1994)
47. Ehresmann, C.: Sur la topologie de certains espaces homognes. Ann. of Math. (2) 35(2),
396443 (1934)
References 519

48. Ehresmann, C.: Sur la topologie de certaines varits algbriques relles. J. Math. Pures Appl.
IX, 69100 (1937)
49. Eilenberg, S.: Singular homology theory. Ann. of Math. 2(45), 407447 (1944)
50. Eilenberg, S., MacLane, S.: Acyclic models. Amer. J. Math. 75, 189199 (1953)
51. Eilenberg, S., Steenrod, N.: Foundations of Algebraic Topology. Princeton University Press,
Princeton (1952)
52. Eilenberg, S., Steenrod, N.E.: Axiomatic approach to homology theory. Proc. Nat. Acad. Sci.
U. S. A. 31, 117120 (1945)
53. Eliahou, S., de la Harpe, P., Hausmann, J.-C., Weber, C.: Michel Kervaire 19272007. Notices
Amer. Math. Soc. 55(8), 960961 (2008)
54. Farber, M.: Topological complexity of motion planning. Discrete Comput. Geom. 29(2), 211
221 (2003)
55. Farber, M.: Topology of robot motion planning. In: Morse Theoretic Methods in Nonlinear
Analysis and in Symplectic Topology, NATO Science Series II: Mathematics, Physics and
Chemistry, vol. 217, pp. 185230. Springer, Dordrecht (2006)
56. Farber, M.: Invitation to topological robotics. European Mathematical Society (EMS), Zrich,
Zurich Lectures in Advanced Mathematics (2008)
57. Farber, M., Fromm, V.: The topology of spaces of polygons. Trans. Amer. Math. Soc. 365(6),
30973114 (2013)
58. Farber, M., Hausmann, J.-C., Schtz, D.: On the conjecture of Kevin Walker. J. Topol. Anal.
1(1), 6586 (2009)
59. Farber, M., Hausmann, J.-C., Schtz, D.: The Walker conjecture for chains in Rd . Math. Proc.
Camb. Philos. Soc. 151(2), 283292 (2011)
60. Farber, M., Schtz, D.: Homology of planar polygon spaces. Geom. Dedicata 125, 7592
(2007)
61. Farber, M., Tabachnikov, S., Yuzvinsky, S.: Topological robotics: motion planning in projec-
tive spaces. Int. Math. Res. Not. 34, 18531870 (2003)
62. Ferry, S.C.: Constructing U V k -maps between spheres. Proc. Amer. Math. Soc. 120(1), 329
332 (1994)
63. Franz, M., Puppe, V.: Steenrod squares on conjugation spaces. C. R. Math. Acad. Sci. Paris
342(3), 187190 (2006)
64. Fritsch, R., Piccinini, R.A.: Cellular Structures in Topology, Volume 19 of Cambridge Stud-
iesin Advanced Mathematics. Cambridge University Press, Cambridge (1990)
65. Fromm, V.: The topology of spaces of polygons. Ph.D. thesis, Durham University. http://
[Link]/3208/ (2011)
66. Fulton, W.: Young Tableaux. With Applications to Representation Theory and Geometry.
London Mathematical Society Student Texts, vol. 35. Cambridge University Press, Cambridge
(1997)
67. Giever, J.B.: On the equivalence of two singular homology theories. Ann. of Math. 2(51),
178191 (1950)
68. Gonzalez, J., Landweber, P.: Symmetric topological complexity of projective and lens spaces.
Algebr. Geom. Topol. 9, 473494 (2009) (electronic)
69. Gordon, McA.: On the higher-dimensional Smith conjecture. Proc. London Math. Soc. 3(29),
98110 (1974)
70. Gordon, McA.: Knots in the 4-sphere. Comment. Math. Helv. 51(4), 585596 (1976)
71. Goresky, M., Kottwitz, R., MacPherson, R.: Equivariant cohomology, Koszul duality, and the
localization theorem. Invent. Math. 131(1), 2583 (1998)
72. Gottlieb, D.H.: Fiber bundles with cross-sections and noncollapsing spectral sequences.
Illinois J. Math. 21(1), 176177 (1977)
73. Gray, B.: Homotopy Theory. Academic Press (Harcourt Brace Jovanovich Publishers), New
York (1975)
74. Greenberg, M.J., Harper, J.R.: Algebraic topology, volume 58 of Mathematics Lecture Note
Series. Advanced Book Program (Benjamin/Cummings Publishing Co. Inc., Reading, 1981)
(A first course)
520 References

75. Griffiths, P., Harris, J.: Principles of Algebraic Geometry. Pure and Applied
[Link]-Interscience, New York (1978)
76. Gubeladze, J.: The isomorphism problem for commutative monoid rings. J. Pure Appl. Alge-
bra 129(1), 3565 (1998)
77. Haefliger, A.: Plongements diffrentiables de varits dans varits. Comment. Math. Helv.
36, 4782 (1961)
78. Haefliger, A.: Points multiples dune application et produit cyclique rduit. Amer. J. Math.
83, 5770 (1961)
79. Hambleton, I., Hausmann, J.-C.: Equivariant bundles and isotropy representations. Groups
Geom. Dyn. 4(1), 127162 (2010)
80. Hambleton, I., Hausmann, J.-C.: Conjugation spaces and 4-manifolds. Math. Z. 269(12),
521541 (2011)
81. Hatcher, A.: Vector Bundles and K-Theory. Available on Allen Hatchers home page at Cornell
University
82. Hatcher, A.: Algebraic topology. Cambridge University Press, Cambridge (2002)
83. Hausmann, J.-C.: Sur la topologie des bras articuls. In Algebraic topology Poznan 1989,
volume 1474 of Lecture Notes in Math (Springer, Berlin, 1991), pp. 146159
84. Hausmann, J.-C.: On the Vietoris-Rips complexes and a cohomology theory for metric spaces.
In: Prospects in Topology (Princeton, NJ, 1994), volume 138 of Ann. of Math. Stud., pp. 175
188. Princeton University Press, Princeton (1995)
85. Hausmann, J.-C.: Geometric descriptions of polygon and chain spaces, in Topology and
robotics, volume 438 of Contemp. Math. (American Mathametical Society, Providence, 2007),
pp. 4757
86. Hausmann, J.-C., Holm, T.: Conjugation spaces and edges of compatible torus actions, in Geo-
metric aspects of analysis and mechanics, volume 292 of Progr (Math. Birkhuser/Springer,
New York, 2011), pp. 179198
87. Hausmann, J.-C., Holm, T., Puppe, V.: Conjugation spaces. Algebr. Geom. Topol. 5:923964
(2005) (electronic)
88. Hausmann, J.-C., Knutson, A.: Polygon spaces and Grassmannians. Enseign. Math. (2),
43(12), 173198 (1997)
89. Hausmann, J.-C., Knutson, A.: The cohomology ring of polygon spaces. Ann. Inst. Fourier
(Grenoble) 48(1), 281321 (1998)
90. Hausmann, J.-C., Rodriguez, E.: The space of clouds in Euclidean space. Exp. Math. 13(1):31
47, 2004. [Link]
91. Hausmann, J.-C., Tolman, S.: Maximal Hamiltonian tori for polygon spaces. Ann. Inst. Fourier
(Grenoble) 53(6), 19251939 (2003)
92. Hill, M.A., Hopkins, M.J., Ravenel, D.C.: The Arf-Kervaire invariant problem in algebraic
topology: introduction, in Current Developments in Mathematics, 2009 (International Press,
Somerville, 2010), pp. 2357
93. Hirsch, M.W.: Immersions of manifolds. Trans. Amer. Math. Soc. 93, 242276 (1959)
94. Hirsch, M.W.: Smooth regular neighborhoods. Ann. of Math. 2(76), 524530 (1962)
95. Hirsch, M.W.: Differential topology. Graduate texts in mathematics, No. 33 (Springer-verlag,
New York, 1976)
96. Hirzebruch, F.: Topological Methods in Algebraic Geometry. Third enlarged edition. New
appendix and translation from the second German edition by R.L.E. Schwarzenberger, with
an additional section by A. Borel. Die Grundlehren der Mathematischen Wissenschaften,
Band 131. Springer, New York (1966)
97. Hirzebruch, F., Mayer, K.H.: O(n)-Mannigfaltigkeiten, exotische Sphren und Singularitten.
Lecture Notes in Mathematics, No. 57 (Springer, Berlin, 1968)
98. Hopf, H.: Zur Algebra der Abbildungen von Mannigfaltigkeiten. J. Reine Angew. Math. 163,
7188 (1930)
99. Hopf, H.: ber die Abbildungen der dreidimensionalen Sphren auf die Kugelflche. Math.
Ann. 104(1), 637665 (1931)
References 521

100. Hopf, H.: ber die Abbildungen von Sphren auf Sphren niedrigerer Dimension. Fundam.
Math. 25, 427440 (1935)
101. Hopf, H.: Ein topologischer Beitrag zur reellen Algebra. Comment. Math. Helv. 13, 219239
(1941)
102. Hopf, H.: Einige persnliche Erinnerungen aus der Vorgeschichte der heutigen Topologie. In:
Colloque de Topologie (Brussels, 1964), pp. 816. Librairie Universitaire, Louvain (1966)
103. Hsiang, Wu-chung: Hsiang, Wu-yi: Differentiable actions of compact connected classical
groups. I. Amer. J. Math. 89, 705786 (1967)
104. Hudson, J.F.P.: Piecewise Linear Topology. University of Chicago Lecture Notes prepared
with the assistance of J. L. Shaneson and J. Lees. W. A. Benjamin Inc, New York (1969)
105. Husemoller, D.: Fibre bundles. In: Graduate Texts in Mathematics, vol. 20, 3rd edn. Springer,
New York (1994)
106. Illman, S.: Smooth equivariant triangulations of G-manifolds for G a finite group. Math. Ann.
233(3), 199220 (1978)
107. Illman, S.: The equivariant triangulation theorem for actions of compact Lie groups. Math.
Ann. 262(4), 487501 (1983)
108. James, I.M.: On category, in the sense of Lusternik-Schnirelmann. Topology 17(4), 331348
(1978)
109. James, I.M., Whitehead, J.H.C.: The homotopy theory of sphere bundles over spheres. II.
Proc. London Math. Soc. 3(5), 148166 (1955)
110. Kamiyama, Y., Tezuka, M., Toma, T.: Homology of the configuration spaces of quasi-
equilateral polygon linkages. Trans. Amer. Math. Soc. 350(12), 48694896 (1998)
111. Kapovich, M., Millson, J.J.: The symplectic geometry of polygons in Euclidean space. J.
Differential Geom. 44(3), 479513 (1996)
112. Kendall, D.G., Barden, D., Carne, T.K., Le, H.: Shape and shape theory. Wiley Series in
Probability and Statistics. John Wiley and Sons Ltd., Chichester (1999)
113. Kervaire, M., Milnor, J.W.: Groups of homotopy spheres. I. Ann. of Math. 2(77), 504537
(1963)
114. Kervaire, M.A.: A manifold which does not admit any differentiable structure. Comment.
Math. Helv. 34, 257270 (1960)
115. Kervaire, M.A.: Smooth homology spheres and their fundamental groups. Trans. Amer. Math.
Soc. 144, 6772 (1969)
116. Kirby, R.C., Siebenmann, L.C.: Foundational Essays on Topological Manifolds, Smooth-
ings, and Triangulations. Princeton University Press, Princeton, NewJersey (1977) (Annals
of Mathematics Studies, No. 88)
117. Kirwan, F.C.: Cohomology of quotients in symplectic and algebraic geometry. Mathematical
Notes. vol. 31. Princeton University Press, Princeton, NJ (1984)
118. Klyachko, A.A.: Spatial polygons and stable configurations of points in the projective line,
in Algebraic geometry and its applications (Yaroslavl 1992), Aspects Math., E25 (Vieweg,
Braunschweig, 1994), pp. 6784
119. Knutson, A., Tao, T.: Puzzles and (equivariant) cohomology of Grassmannians. Duke Math.
J. 119, 221260 (2003)
120. Kosinski, A.A.: Differential Manifolds. Academic, Boston, MA (1993) (vol. 138 of Pure and
Applied Mathematics)
121. Kreck, M., Puppe, V.: Involutions on 3-manifolds and self-dual, binary codes. Homology,
Homotopy Appl. 10(2), 139148 (2008)
122. Lang, S.: Algebra, 2nd edn. Addison-Wesley Publishing Company Advanced Book Program,
Reading, MA (1984)
123. Lannes, J.: Sur linvariant de Kervaire des varits fermes stablement paralllises. Ann. Sci.
cole Norm. Sup. (4), 14(2), 183197 (1981)
124. Lannes, J.: Sur les espaces fonctionnels dont la source est le classifiant dun p-groupe ablien
lmentaire. Inst. Hautes tudes Sci. Publ. Math. (75), 135244, (1992). With an appendix
by Michel Zisman
522 References

125. Lannes, J., Zarati, S.: Sur les foncteurs drivs de la dstabilisation. Math. Z. 194(1), 2559
(1987)
126. Lashof, R.K.: Equivariant bundles. Illinois J. Math. 26(2), 257271 (1982)
127. Lashof, R.K., May, J.P.: Generalized equivariant bundles. Bull. Soc. Math. Belg. Sr. A
38(265271), 1986 (1987)
128. Latour, F.: Double suspension dune sphre dhomologie [daprs R. Edwards]. In: Sminaire
Bourbaki, 30e anne (1977/78), vol. 710 of Lecture Notes in Math., pages Exp. No. 515, pp.
169186. Springer, Berlin (1979)
129. Latschev, J.: Vietoris-Rips complexes of metric spaces near a closed Riemannian manifold.
Arch. Math. (Basel) 77(6), 522528 (2001)
130. Lawson, H.B. Jr., Michelsohn, M.-L.: Spin Geometry. Princeton Mathematical Series, vol.
38. Princeton University Press, Princeton (1989)
131. Lebesgue, H.: Oeuvres scientifiques (en cinq volumes), vol. IV. LEnseignement mathma-
tique, Geneva (1973) (Sous la rdaction de Franois Chtelet et Gustave Choquet)
132. Lefschetz, S.: Introduction to Topology (Princeton Mathematical Series, vol. 11). Princeton
University Press, Princeton, New Jersey (1949)
133. Levine, J.P.: Lectures on groups of homotopy spheres. In: Algebraic and Geometric Topology,
New Brunswick, New Jersey, 1983, vol. 1126 of Lecture Notes in Math., pp. 6295. Springer,
Berlin (1985)
134. MacLane, S.: Categories for the Working Mathematician, Volume 5 of Graduate Texts in
Mathematics, 2nd edn. Springer, New York (1998)
135. Massey, W.S.: Imbeddings of projective planes and related manifolds in spheres. Indiana Univ.
Math. J. 23, 791812 (1973/74)
136. Massey, W.S.: A Basic Course in Algebraic Topology, Volume 127 of Graduate Texts in
Mathematics. Springer-Verlag, New York (1991)
137. Massey, W.S.: A history of cohomology theory. In: History of Topology, pp. 579603. North-
Holland, Amsterdam (1999)
138. Matouek, J.: Lectures on Discrete Geometry. Graduate Texts in Mathematics, vol. 212.
Springer, New York (2002)
139. May, J.P.: Equivariant homotopy and cohomology theory. CBMS Regional Conference Se-
ries in Mathematics., vol. 91. Published for the Conference Board of the Mathematical Sci-
ences, Washington, DC (1996). With contributions by Cole, M., Comezaa, G., Costenoble,
S., Elmendorf, A.D., Greenlees, J.P.C., Lewis, L.G. Jr., Piacenza, J.R.J., Triantafillou, G.,
Waner, S.
140. May, J.P.: A concise course in algebraic topology. Chicago Lectures in Mathematics. Univer-
sity of Chicago Press, Chicago, IL (1999)
141. McCleary, J.: A users guide to spectral sequences. In: Cambridge Studies in Advanced Math-
ematics, vol. 58, 2nd edn. Cambridge University Press, Cambridge (2001)
142. Milgram, R.J., Trinkle, J.C.: The geometry of configuration spaces for closed chains in two
and three dimensions. Homology Homotopy Appl. 6(1), 237267 (2004)
143. Miller, H.: The Sullivan conjecture on maps from classifying spaces. Ann. Math. (2) 120(1),
3987 (1984)
144. Milnor, J.W.: Construction of universal bundles. II. Ann. of Math. 2(63), 430436 (1956)
145. Milnor, J.W.: On manifolds homeomorphic to the 7-sphere. Ann. of Math. 2(64), 399405
(1956)
146. Milnor, J.W.: The geometric realization of a semi-simplicial complex. Ann. of Math. 2(65),
357362 (1957)
147. Milnor, J.W.: On spaces having the homotopy type of CW-complex. Trans. Amer. Math. Soc.
90, 272280 (1959)
148. Milnor, J.W.: A procedure for killing homotopy groups of differentiable manifolds, in Pro-
ceedings of the Symposium Pure Mathematics, vol. III (American Mathematical Society,
Providence, 1961), pp. 3955
149. Milnor, J.W.: Morse theory. Based on lecture notes by M. Spivak and R. Wells. Annals of
Mathematics Studies, No. 51. Princeton University Press, Princeton, N.J. (1963)
References 523

150. Milnor, J.W.: Spin structures on manifolds. Enseignement Math. 2(9), 198203 (1963)
151. Milnor, J.W.: Lectures on the h-cobordism theorem. Notes by L. Siebenmann and J. Sondow.
Princeton University Press, Princeton, N.J. (1965)
152. Milnor, J.W.: Topology from the differentiable viewpoint. In: David, W.W. (ed.) Based on
Notes. The University Press of Virginia, Charlottesville, VA (1965)
153. Milnor, J.W., Stasheff, J.D.: Characteristic Classes. Princeton University Press, Princeton
(1974) Annals of Mathematics Studies, No. 76
154. Mosher, R.E., Tangora, M.C.: Cohomology operations and applications in homotopy theory.
Harper and Row Publishers, New York (1968)
155. Munkres, J.R.: Elements of algebraic topology. Addison-Wesley Publishing Company, Menlo
Park, CA (1984)
156. Newman, M.: Integral Matrices, in Pure and Applied Mathematics, vol. 45 (Academic Press,
New York, 1972)
157. Olbermann, M.: Conjugations on 6-manifolds, Ph.D. thesis, University of Heidelberg. http://
[Link]/archiv/7450 (2007)
158. Olbermann, M.: Conjugations on 6-manifolds. Math. Ann. 342(2), 255271 (2008)
159. Olbermann, M.: Involutions on S 6 with 3-dimensional fixed point set. Algebr. Geom. Topol.
10(4), 19051932 (2010)
160. Pitsch, W., Scherer, J.: Conjugation spaces and equivariant Chern classes. Bull. Belg. Math.
Soc. Simon Stevin 20, 7790 (2013)
161. Poincar, H.: Analysis situs. J. Ec. Polytech. 21, 1123 (1895)
162. Poincar, H.: Complment lAnalysis situs. Palermo Rend. 13, 285343 (1899)
163. Pont, J.-C.: La topologie algbrique des origines Poincar. Presses Universitaires de France,
Paris (1974)
164. Postnikov, M.M.: The structure of the ring of intersections of three-dimensional manifolds.
Doklady Akad. Nauk. SSSR (N.S.) 61, 795797 (1948)
165. Puppe, V.: Group actions and codes. Canad. J. Math. 53(1), 212224 (2001)
166. Puppe, V.: Do manifolds have little symmetry? J. Fixed Point Theory Appl. 2(1), 8596 (2007)
167. Quillen, D.: The mod 2 cohomology rings of extra-special 2-groups and the spinor groups.
Math. Ann. 194, 197212 (1971)
168. Ranicki, A.A.: On the Hauptvermutung. In: The Hauptvermutung Book, pp. 331. Kluwer
Academic, Dordrecht (1996) (vol. 1 of K-Monogr. Math.)
169. Ronga, F.: Schubert calculus according to Schubert. arXiv:math/0608784v1[[Link]]
170. Schtz, D.: The isomorphism problem for planar polygon spaces. J. Topol. 3(3), 713742
(2010)
171. Schwartz, L.: Unstable modules over the Steenrod algebra and Sullivans fixed point set con-
jecture. Chicago Lectures in Mathematics. University of Chicago Press, Chicago, IL (1994)
172. Schwartz, L.: propos de la conjecture de non-ralisation due N. Kuhn. Invent. Math.
134(1), 211227 (1998)
173. Scull, L.: Equivariant formality for actions of torus groups. Canad. J. Math. 56(6), 12901307
(2004)
174. Seifert, H., Threlfall, W.: Lehrbuch der Topologie. B. G. Teubner, Leipzig und Berlin (1934)
175. Serre, J.-P.: Cohomologie modulo 2 des complexes dEilenberg-MacLane. Comment. Math.
Helv. 27, 198232 (1953)
176. Seymour, R.M.: On G-cohomology theories and Knneth formulae, in Current trends in alge-
braic topology, Part 1 (London, Ont., 1981), vol. 2. CMS Conferences Proceedings (American
Mathematical Society, Providence, RI, 1982), pp. 257271
177. Smale, S.: Generalized Poincars conjecture in dimensions greater than four. Ann. of Math.
2(74), 391406 (1961)
178. Smith, P.A.: Transformations of finite period. Ann. of Math. 2(39), 127164 (1938)
179. Spanier, E.H.: Algebraic topology. Springer-Verlag, New York (1981)
180. Spivak, M.: A Comprehensive Introduction to Differential Geometry, vol. V, 2nd edn. Publish
or Perish Inc., Wilmington, USA (1979)
181. Steenrod, N.: The Topology of Fibre Bundles. Princeton University Press, New Jersey (1951)
524 References

182. Steenrod, N.E.: Cohomology invariants of mappings. Ann. of Math. 2(50), 954988 (1949)
183. Steenrod, N.E.: The work and influence of Professor S. Lefschetz in algebraic topology,
in Algebraic Geometry and Topology. A Symposium in Honor of S. Lefschetz (Princeton
University Press, New Jersey, 1957), pp. 2443
184. Steenrod, N.E.: Cohomology operations. Lectures by N. E. Steenrod written and revised by
D. B. A. Epstein. Annals of Mathematics Studies, vol. 50. Princeton University Press, Prince-
ton (1962)
185. Steenrod, N.E.: A convenient category of topological spaces. Michigan Math. J. 14, 133152
(1967)
186. Stong, R.E.: Notes on cobordism theory. Mathematical notes. Princeton University Press,
Princeton, N.J. (1968)
187. Sullivan, D.P.: On the intersection ring of compact three manifolds. Topology 14(3), 275277
(1975)
188. Sullivan, D.P.: Geometric Topology: Localization, Periodicity and Galois Symmetry, Volume
8 of K-Monographs in Mathematics. Springer, Dordrecht (2005) The 1970 MIT notes, Edited
and with a preface by Andrew Ranicki
189. Sumners, D.W.: Smooth Z p -actions on spheres which leave knots pointwise fixed. Trans.
Amer. Math. Soc. 205, 193203 (1975)
190. Thom, R.: Espaces fibrs en sphres et carrs de Steenrod. Ann. Sci. Ecole Norm. Sup. 3(69),
109182 (1952)
191. Thom, R.: Quelques proprits globales des varits diffrentiables. Comment. Math. Helv.
28, 1786 (1954)
192. Thom, R.: Un lemme sur les applications diffrentiables. Bol. Soc. Mat. Mexicana 2(1), 5971
(1956)
193. Thomas, E.: On the cohomology groups of the classifying space for the stable spinor groups.
Bol. Soc. Mat. Mexicana 2(7), 5769 (1962)
194. Thomas, E.: On the mod 2 cohomology of certain H-spaces. Comment. Math. Helv. 37,
132140 (1962)
195. Thurston, W., Meeks, J.: The mathematics of three-dimensional manifolds. Scientific Amer-
ican 251(1), 94106 (July 1984)
196. Tietze, H.: ber die topologischen Invarianten mehrdimensionaler Mannigfaltigkeiten.
Monatsh. Math. Phys. 19(1), 1118 (1908)
197. Toda, H.: Composition methods in homotopy groups of spheres. Annals of Mathematics
Studies, vol. 49. Princeton University Press, Princeton (1962)
198. Tolman, S., Weitsman, J.: On the cohomology rings of hamiltonian T-spaces, in Northern
California Symplectic Geometry Seminar, vol. 196. Amer. Math. Soc. Transl. Ser. 2 (American
Mathematical Society, Providence, RI, 1999), pp. 251258
199. Tolman, S., Weitsman, J.: The cohomology rings of symplectic quotients. Comm. Anal. Geom.
11(4), 751773 (2003)
200. Veblen, O., Alexander, J.W. II.: Manifolds of N dimensions. Ann. Math. (2), 14(14), 163178
(1912/1913)
201. Vietoris, L.: ber den hheren Zusammenhang kompakter Rume und eine Klasse von zusam-
menhangstreuen Abbildungen. Math. Ann. 97(1), 454472 (1927)
202. Walker, K.: Configuration spaces of linkages. Bachelors thesis, Princeton. [Link]
net/math/[Link] (1985)
203. Wall, C.T.C.: Surgery of non-simply-connected manifolds. Ann. of Math. 2(84), 217276
(1966)
204. Wall, C.T.C.: Surgery on Compact Manifolds (Academic Press, London, 1970) (London Math-
ematical Society Monographs, No. 1)
205. Weber, C.: The Robertello-Arf invariant of knots Preprint (University of Geneva, Geneva)
206. Weinberger, S.: Homology manifolds. In: Sher, R.B., Daverman, R.J. (eds.) Handbook of
Geometric Topology, pp. 10851102. North-Holland, Amsterdam (2002)
207. Whitehead, G.W.: Elements of Homotopy Theory, Volume 61 of Graduate Texts in Mathematics
(Springer, New York, 1978)
References 525

208. Whitehead, J.H.C.: On C 1 -complexes. Ann. of Math. 2(41), 809824 (1940)

209. Whitney, H.: On products in a complex. Ann. of Math. (2) 39(2), 7432 (1938)
210. Whitney, H.: The singularities of a smooth n-manifold in (2n1)-space. Ann. of Math. 2(45),
247293 (1944)
211. Whitney, H.: Moscow 1935: Topology moving toward America, in A Century of Mathematics
in America, Part I, History of Mathematics, vol. 1 (American Mathematical Society, Provi-
dence, 1988), pp. 97117
212. Wood, R.M.W.: Problems in the Steenrod algebra. Bull. London Math. Soc. 30(5), 449517
(1998)
213. Wen-tsn, Wu: Classes caractristiques et i-carrs dune varit. C. R. Acad. Sci. Paris 230,
508511 (1950)
214. Wen-tsn, Wu: Les i-carrs dans une varit grassmannienne. C. R. Acad. Sci. Paris 230,
918920 (1950)
Index

A of finite type, 175

Action pair, 177
code associated to an, 435 principal, 274, 298, 358
semi-free, 436 trivialization, 358
with scattered fixed point set, 433 universal, 274, 299
Acyclic carrier, 49 vector, 186189, 355358
Adem relations, 344 Bundle characteristic map, 189
Admissible sequence, 347 Bundle gluing map, 189
Affine order, 9
Alexander duality, 220
Almost closed manifold, 492 C
Annihilator, 224, 265 C 1 -triangulation, 203
Arf invariant, 485 Cap product
Axial map, 246 functoriality, 152, 153
Axioms of a cohomology theory, 123 in relative simplicial cohomology, 152
simplicial, 150
singular, 155
B Category
Barycenter, 11 CW, 112
Barycentric subdivision, 11 GrA, 129
Betti number, 23 GrA[u], 263
total, 285 GrV, 31
Binomial coefficients mod 2, 248 Lusternik-Schnirelmann, 147
Bockstein homomorphism, 350 RCW, 115
Borel construction, 260, 275 Simp, 7
Borsuk-Ulam theorem, 245 Top, 8
Boundary of a simplex, 6 Top2 , 69
Boundary operator TopG , 260
cellular, 102 Top , 275
in a chain complex, 14 Cell
ordered, 51 attaching a, 97
simplicial, 13 characteristic map of a, 97
singular, 61 convex, 9
Bouquet, 84 homology, 108
cohomology ring of, 134 open, 97
Bundle, 172 Cellular
induced, 183, 356 chain, 102
Springer International Publishing Switzerland 2014 527
J.-C. Hausmann, Mod Two Homology and Cohomology, Universitext,
DOI 10.1007/978-3-319-09354-3
528 Index

cochain, 102 simplicial, 42

cohomology, 102 singular, 68
homology, 102 extension of the fiber, 172
map, 107 of a cochain complex, 16
Chain operation, 326
cellular, 102 ordered, 52
complex, 14 positive parts, 134
morphism of, 17 reduced
homotopy, 34, 73 singular, 63
map, 17 relative
ordered, 51 simplicial, 41
simplicial, 11 singular, 67
singular, 60 sequence
Chain space, 444 simplicial, 42
big, 444 simplicial, 13
Chamber of a length vector, 447 singular, 61
Characteristic class theory, 123
of a 2-fold covering, 142 natural transformation, 123
of a vector bundle, 355, 372 Colouring definitions
Characteristic map of cellular (co)chains, 103
global, 99 of ordered (co)chains, 51
of a 2-fold covering, 140, 141 of simplicial (co)chains, 11
Characteristic map of a cell, 97 of singular (co)chains, 60
Chern classes Combinatorial manifold, 202
action of the Steenrod algebra on, 410 Complex
integral, 408 chain, 14
mod2, 408 cochain, 14
uniqueness, 410 convex-cell, 9
Circuit, 22 full, 24
Cobordant manifolds, 429 simplicial, 5
Cobordism, 431 Component
Coboundary operator of a simplicial complex, 8
cellular, 102 Cone
in a chain complex, 14 on a simplicial complex, 23
ordered, 52 on a topological space, 74
simplicial, 13 Conjugation
Cochain cell, 438
cellular, 102 complex, 438
complex, 14 equation, 437
morphism of, 17 universal, 441
map, 17 manifold, 441
ordered, 51 space, 437
simplicial, 11 Conjugation space, 409
singular, 60 Connected sum, 135
unit, 20, 62 cohomology ring of, 136
Cocycles Connecting
simplicial, 13 1-cochain, 62, 142
singular, 61 homomorphism
Code, 435 cohomology, 36
Cofibrant pair, 79 homology, 39
Cohomology Connecting homomorphism
cellular, 102 simplicial, 44
connecting homomorphism, 36 Contiguous simplicial maps, 34
Index 529

Contractible space, 74 local, 91

Convex cell, 9 Degree of a map
Convex-cell complex, 9 between manifolds, 204
2-fold covering between pseudomanifold, 32
characteristic class of, 142 between spheres, 90
deck involution, 141 Democratic invariant, 485
transfer exact sequence, 145 Dimension
transfer homomorphism, 145 of a simplex, 5
Cross product of a simplicial complex, 6
associativity, 166 topological invariance, 95
equivariant, 291 Disjoint union axiom, 123
strong, 291 Duality
functoriality, 158 Alexander, 220
in cohomology, 157 Kronecker, 1419
in homology, 159 ordered, 52
in relative cohomology, 158 singular, 61
of maps, 156 Lefschetz, 211
reduced, 169 Poincar, 203
equivariant, 296
Cross-square map, 333
equivariant, 339 E
Cup product Eilenberg-MacLane space, 119
commutativity, 129 cohomology of, 349
commutativity of, 132 Equivariant
functoriality, 129, 132 cross product, 291
in relative simplicial cohomology, 130 CW-complex, 260, 287
in relative singular cohomology, 133, 155 Stiefel-Whitney classes, 470
simplicial, 128 vector bundle, 300
singular, 132 Equivariant cohomology
Cup-square map, 138, 249, 352 for a pair with involution, 263
CW in general, 276
complex, 97 Equivariantly formal, 265, 278
Cartesian product, 99 Euclidean bundle, 187
equivariant, 260, 287 Euler
finite, 106 characteristic
homology-cell complex, 108 of a finite CW-complex, 106
regular, 104 of a finite simplicial complex, 25
weak topology, 97 of a manifold, 204, 217, 230
pair, 97 class, 184, 187
space, 96 equivariant, 303
structure, 97 Exact sequence
on S n , 99 cohomology, 36
on projective spaces, 99 homology, 40
perfect, 106 of cochain complexes, 35
Cycles simplicial (co)homology, 42, 45
simplicial, 13 of a triple, 46
singular, 61 singular (co)homology, 68
of a triple, 71
Exactness axiom, 123
D Excision
Deck involution axiom, 123
of a 2-fold covering, 141 property, 77
Degree simplicial, 57
530 Index

Excisive couple, 132 complex case, 484

Extension of a cochain Global characteristic map, 99
simplicial, 42 G-map, 260
singular, 67 Good pair, 78
GrA, 129
Graded algebras
F category of, 129
Face Graph
exterior algebra, 195 simplicial, 6
front, back, 131 Grassmannian, 376
in a regular CW-complex, 104 complex, 405
inclusion, 60 infinite, 382, 407
of a simplex, 6 tautological bundle over, 381, 407
space, 194 GrA[u], 263
Fiber inclusion, 172 GrV, 31
Finite (co)homology type, 23, 72, 88 Gubeladzes theorem, 196
Finite type (graded vector space), 23 Gysin
Flag complex, 194 exact sequence, 184, 186
Flag manifold, 376 homomorphism, 225
complete, 377, 405
complex, 404
tautological bundle over..., 407
H
tautological bundle over..., 381
Ham sandwich theorem, 246
Forgetful homomorphism (in equiv. coho-
Hawaiian earring, 125
mology), 264, 278
Homogeneous space, 375
Framed
cobordism, 487 Homology
manifold, 487 cell, 108
Framed bundle (of a K-vector bundle), 300 complex, 108
Framed bundle (of a vector bundle), 359 cellular, 102
Framing (stable), 487 connecting homomorphism, 39
Full complex, 24 of a chain complex, 16
Fundamental class ordered, 52
of a polyhedral homology manifold, 203 reduced
of a pseudomanifold, 22 singular, 63
of a relative homology manifold, 213 relative
simplicial, 43
singular, 67
G sequence
G-contractible, 260 simplicial, 45
Generic length vector, 446 simplicial, 13
Genus singular, 61
of a nonorientable surface, 30 sphere, 220
of an orientable surface, 29 Homology manifold, 202
Geometric relative, 211
open simplex, 7 Homotopic maps, 73
simplex, 7 Homotopy, 73
Geometric realization axiom, 123
of a simplicial complex, 6 chain, 34, 73
metric topology, 6 equivalence, 74
weak topology, 6 equivalent pairs or spaces, 74
G-homotopy, 260 property (in singular (co)homology), 73
Giambelli formula, 399 quotient, 260, 275
GKM-conditions, 483 relative, 356
Index 531

type, 74 generic, 446

Hopf lopsidedness of a, 463
bundle, 241244 normal, 466
vector, 244 Leray-Hirsch Theorem, 173, 177
invariant, 249, 353 2nd version, 175
map, 242 Link, 7
Linking number, 232
Localization
I principle, 223
Intersection form, 218 theorem, 283, 287
Invariance of dimension, 95 Long subset, 446
Inverse problem, 464 Lopsidedness, 463
Lusternik-Schnirelmann category, 147

J
Join M
simplicial, 7 Majority invariant, 485
topological, 171 Manifold
Jordan Theorem, 96 combinatorial, 202
PL, 202
polyhedral homology, 202
K relative polyhedral homology, 211
Kervaire topological, 95, 135
invariant, 492, 493 Map
manifold, 496 cellular, 107
sphere, 496 of simplicial triple, 46
Kirwan surjectivity theorems, 317321 of topological pairs, 69
Klein bottle piecewise affine, 6
cohomology algebra of, 138 Mapping cylinder
triangulation of, 28 bundle pair, 182
Kronecker double, 257
duality, 1419 neighbourhood, 79
ordered, 52 Mapping torus, 191
singular, 61 exact sequence, 191
pair, 15 Maximal simplex, 6
morphism of, 17 Mayer-Vietoris
pairing connecting homomorphisms
extended, 61 simplicial, 48
on (co)chains, 12 singular, 87
on (co)homology, 16 data, 87
on cellular (co)chains, 103 sequence
on ordered (co)chains, 51 simplicial, 48
on singular (co)chains, 60 singular, 87
Knneth theorem, 160 Meridian sphere, 233
equivariant, 292 M-involution, 434
reduced, 169 Morse-Bott
relative, 164, 179 function, 309
critical manifold, 309
perfect, 311
L self-indexed, 377
Lefschetz duality, 211 inequalities, 310
Length vector, 444 lacunary principle, 311
chamber of a, 447 polynomial, 310
dominated, 461 Morse function, 309
532 Index

N free, 445
Natural transformation, 112 spatial, 468
NDR-pair, 80 Polyhedral homology
Negative normal bundle, 309 manifold, 202
Nilpotency class, 147 relative, 211
Non-singular map, 246, 353 Pontryagin-Thom construction, 430
Null-homotopic, 121 Product
cap, see cap product
in relative simplicial cohomology,
O 152
Octonionic projective plane cross, 156
cohomology algebra of O P n , 243 cup, see cup product
Odd map, 245 in relative singular cohomology, 155
Open simplex, 7 of CW-complexes, 99
Order slant, 412
affine, 9 tensor, 157
simplicial, 8 Projective space
Ordered R P as Eilenberg-space, 121
chain, 51 cohomology algebra of C P n , 241
cochain, 51 cohomology algebra of H P n , 242
cohomology, 52 cohomology algebra of R P n , 146, 239
homology, 52 complex, 240
simplex, 50 octonionic (projective plane), 243
Orientation quaternionic, 242
of a vector bundle, 360 real, 99
of a vector space, 360 simplicial cohomology algebra of R P 2 ,
transport along a path, 366 137
standard CW-structure on R P n , 99
tautological line bundle, 366
triangulation of R P 2 , 26
P
Proper map, 310
Pair Pseudomanifold, 21
CW, 97 fundamental class of, 22
good, 78 fundamental cycle of, 22
Kronecker, 15
simplicial, 41
topological, 66 Q
map of, 69 Quadratic form, 485
Perfect
CW-structure, 106
Morse-Bott function, 311 R
Piecewise affine map, 6 Reduced
Pieri formula, 398 equivariant cohomology, 266, 279
PL-manifold, 202 singular (co)homology, 63
Poincar dual suspension, 170
functoriality, 224 Reduced symmetric square, 420
of a homology class, 217 Regular
of a submanifold, 221 CW-complex, 104
Poincar duality, 203 value (topological), 91
Poincar series/polynomial, 23, 72, 88 Relative
Pointed space, 84 cohomology
well, 85, 168 ordered, 55
Polygon space, 444 simplicial, 41
abelian, 470 singular, 67
Index 533

homology order, 8
simplicial, 43 for the barycentric subdivision, 11
singular, 67 pair, 41
singular cycle, 68 pairs
Representable functor, 119 map of, 45
Retraction, 78, 94 relative
by deformation, 78 cohomology, 41
Riemannian metric, 187 homology, 43
Robot arm map, 445 suspension, 56
triad, 47
triple, 46
S map of, 46
Scattered fixed point set, 433 Simplicial complex
Schubert complex, 5
cell, 390 finite, 6
variety, 390 component of, 8
Schubert calculus, 393400 connected, 8
complex, 411 Euclidean realization, 6
Giambelli formula, 399 geometric realization, 6
Pieri formula, 398 locally finite, 6
Segre class, 411 Simplicial excision, 57
Self-intersection, 490 Singular
Semi-free action, 436 chain, 60
Sequence, see exact sequence cochain, 60
Short subset, 56, 446 cocycles, 61
Shn (), 446 cohomology, 61
Sh n (), 459
cycles, 61
Simp, 7 homology, 61
Simplex, 5 reduced (co)homology, 63
boundary of a, 6 relative
geometric, 7 cohomology, 67
geometric open, 7 homology, 67
maximal, 6 simplex, 59
ordered, 50 small, 75
singular, 59 Skeleton
small, 75 of a CW-complex, 97
standard, 59 of a simplicial complex, 6
Simplicial Slant product, 412
category, 7 Slice inclusion, 165
chain, 11 Small
cochain, 11 map, 75
cocycles, 13 singular simplex, 75
cohomology, 13 Smash product, 168
exact sequence, 42, 45, 46 Smith inequality, 285
cohomology connecting homomorphism, Spin structure, 370
42 Splitting principle, 402, 409
cycles, 13 for complex bundles, 410
graph, 6 generalized, 401, 410
homology, 13 Star, 7
homology connecting homomorphism, open, 7
44 Steenrod
map, 7 algebra, 346
of simplicial pairs, 45 squares, 334
534 Index

characterization, 351 complexity, 149

Stiefel manifold, 381, 407, 479 pair, 66
Stiefel-Whitney classes, 372 regular value, 91
action of the Steenrod algebra on, 403 2-torus (2-elementary abelian group), 304
dual, 389 Torus (Lie group), 307
equivariant, 470 associated 2-torus, 307
first, 363 Torus (manifold)
of projective spaces, 416417 cohomology algebra of T 2 , 137
second, 367 triangulation of T 2 , 27
uniqueness, 403 Transfer, 145
Stiefel-Whitney numbers, 429 exact sequence, 145, 186
Subcomplex Transport of an orientation, 366
of a CW-complex, 97 Triangulable
of a simplicial complex, 6 pair, 211
Subdivision, 10 space, 9
barycentric, 11 Triangulation, 9, 211
operator, 209 Triple
Subset definitions topological, 70
of (co)chains, 11 map of, 71
of cellular (co)chains, 102
of ordered (co)chains, 51
of singular (co)chains, 60 U
Sullivan conjecture, 240, 353 Umkehr homomorphism, 225
Surface, 26 Unit cochain, 20, 52, 62
cohomology algebra of, 137, 138
genus
V
of a nonorientable surface, 29
Vector bundle, 186189, 355358
of an orientable surface, 29
associated framed bundle, 300, 359
nonorientable, 29
complex, 362
orientable, 29
equivariant, 300
Surgery, 489
rigid, 303, 473
Suspension, 83
Euclidean, 187, 360
isomorphism, 84
induced, 188, 356
reduced, 170
isomorphism of, 355
simplicial, 56
orientable, 360
Symplectic basis, 411
product, 356
stable isomorphism of, 487
structures on, 359
T tensor product, 302
Tautological vector bundle, 244, 366, 381 for line bundles, 364, 409
complex, 407 Thom space, 426
Tensor product, 157 trivial, 356
Thom trivialization, 356
class, 182 stable, 487
isomorphism theorem, 182, 187 Whitney sum, 188
space, 426 equivariant, 301
Thom-Pontryagin construction Vertex,vertices, 5
see Pontryagin-Thom construction, 430
Top, 8
Top2 , 69 W
TopG , 260 Wang exact sequence, 190
Top , 275 Weak homotopy equivalence, 122
Topological Weak topology (for CW-complexes), 97
Index 535

Wedge, see bouquet presentation of, 80

Weight Well pointed space, 85, 168
bundle, 473 equivariantly, 295
Weighted trace, 377, 405 Wu
Well cofibrant pair, 80 class, 411
equivariantly, 280 formula, 412