Sasakian Geometry Charles Boyer Krzysztof

Galicki download

https://ebookbell.com/product/sasakian-geometry-charles-boyer-
krzysztof-galicki-6986112

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

A Mod P Jacquetlanglands Relation And Serre Filtration Via The

Geometry Of Hilbert Modular Varieties Splicing And Dicing Fred Diamond

https://ebookbell.com/product/a-mod-p-jacquetlanglands-relation-and-
serre-filtration-via-the-geometry-of-hilbert-modular-varieties-
splicing-and-dicing-fred-diamond-52181396

Sasanian Studies Late Antique Iranian World Sasanidische Studien

Sptantike Iranische Welt Vol 1 Edited By Shervin Farridnejad Touraj
Daryaee

https://ebookbell.com/product/sasanian-studies-late-antique-iranian-
world-sasanidische-studien-sptantike-iranische-welt-vol-1-edited-by-
shervin-farridnejad-touraj-daryaee-49199662

Sasanian Clay Sealings In The Bandar Abbas Museum Kamal Aldin Niknami

https://ebookbell.com/product/sasanian-clay-sealings-in-the-bandar-
abbas-museum-kamal-aldin-niknami-49994596

Sasanian Archaeology Settlements Environment And Material Culture St

John Simpson Editor

https://ebookbell.com/product/sasanian-archaeology-settlements-
environment-and-material-culture-st-john-simpson-editor-50202504
Sasanian Persia 1st Edition Touraj Daryaee

https://ebookbell.com/product/sasanian-persia-1st-edition-touraj-
daryaee-50455120

Sasanian Persia Touraj Daryaee

https://ebookbell.com/product/sasanian-persia-touraj-daryaee-50677030

Sasanian Persia Between Rome And The Steppes Of Eurasia Eberhard Sauer

https://ebookbell.com/product/sasanian-persia-between-rome-and-the-
steppes-of-eurasia-eberhard-sauer-51971812

Sasanian Iran In The Context Of Late Antiquity The Bahari Lecture

Series At The University Of Oxford Touraj Daryaee

https://ebookbell.com/product/sasanian-iran-in-the-context-of-late-
antiquity-the-bahari-lecture-series-at-the-university-of-oxford-
touraj-daryaee-56401674

Sasanian Persia Between Rome And The Steppes Of Eurasia 1st Edition
Eberhard Sauer

https://ebookbell.com/product/sasanian-persia-between-rome-and-the-
steppes-of-eurasia-1st-edition-eberhard-sauer-6874054
 Sasakian Geometry

Charles P. Boyer and Krzysztof Galicki

Department of Mathematics and Statistics, University of New Mex-

ico, Albuquerque, N.M. 87131
 Preface

This book is a result of a fifteen year long collaboration which has led us to
understand and appreciate the importance of Sasakian manifolds as an integral
part of Riemannian Geometry. In the early nineties neither of us was aware or
particularly interested in Sasakian structures. This rapidly changed in 1992 when,
together with Ben Mann, we realized that smooth 3-Sasakian manifolds, which are
automatically Einstein, are orbibundles over positive quaternionic Kähler orbifolds.
In the smooth case this bundle was first described by Konishi. As quaternionic
Kähler orbifolds are easily manufactured via quaternionic reduction, likewise we
were able to construct large families of smooth compact 3-Sasakian spaces with
relative ease. Searching through the literature afterwards we were surprised that a
folklore conjecture attributed to Tanno stated that any 3-Sasakian manifold should
be a spherical space form.
Soon after we began to understand that the 3-Sasakian geometry, though very
interesting, was too specialized a case. To a large extent our main motivation
had to do with Einstein metrics. We realized that the 3-Sasakian manifolds were
merely a special case of the more general theory of Sasaki-Einstein manifolds and
shifted our focus to that more general case. With time it became clearer and clearer
that Sasakian geometry is one of the richest sources of complete Einstein metrics of
positive scalar curvature. We began to appreciate that Sasakian geometry, naturally
“sandwiched” between two different Kähler geometries, was at least as interesting
and important as the latter.
It was Nigel Hitchin who first suggested to us that perhaps it is time to write
a modern book on the subject. We started thinking about the project in 2002. It
was a challenging endeavor: we aimed at writing a monograph which should give a
fairly complete account of our own contributions to the subject combined with an
advanced graduate level textbook describing the foundational material in a modern
language. Unlike the Kähler case, there are very few books of this sort discussing
Sasakian manifolds.
Another difficulty which we were to discover later was the rapid development of
the subject. Some new results, such as the construction of Sasaki-Einstein metrics
on exotic spheres with János Kollár, had to do with the dynamics of our own
research program. But many other important results were obtained by others.
To complicate things even more, Sasaki-Einstein manifolds appear to play a very
special rôle in the so-called AdS/CFT duality conjecture and they have received an
enormous amount of attention among physicists working in this area. Some of the
intriguing new results which we decided to include in the book, albeit very briefly,
indeed came quite unexpectedly from considerations in Superstring Theory.
In the end we know our book will not be as complete as we would have wanted
or hoped it to be. We finally had to stop writing while knowing all too well that the

iii
iv PREFACE

book is neither complete nor perfect. Few books of this sort are. We are pleased to
see so much renewed interest in the subject and hope that our work will be of help
to a number of mathematicians and physicists, researchers and graduate students
alike.
Over the years we have benefited immensely from many discussions with many
colleagues and collaborators about the mathematics contained in this book. Here
we are happy to take this opportunity to thank: B. Acharya, I. Agricola, D. Alek-
seevsky, V. Apostolov, F. Battaglia, C. Bär, H. Baum, F. Belgun, L. Bérard Bergery,
R. Bielawski, C. Böhm, A. Buium, D. Blair, O. Biquard, J.-P. Bourguignon, R.
Bryant, D. Calderbank, A. Čap, J. Cheeger, T. Colding, A. Dancer, O. Dearricott,
T. Draghici, Y. Eliashberg, J. Figueroa O’Farrill, M. Fernandez, T. Friedrich, E.
Gasparim, P. Gauduchon, H. Geiges, W. Goldman, G. Grantcharov, K. Grove, M.
Harada, T. Hausel, G. Hernandez, L. Hernandez, R. Herrera, O. Hijazi, N. Hitchin,
H. Hofer, T. Holm, J. Hurtubise, J. Isenberg, G. Jensen, J. Johnson, A. de Jong,
D. Joyce, L. Katzarkov, J. Konderak, M. Kontsevich, J. Kollár, H. B. Lawson,
C. LeBrun, B. Mann, P. Matzeu, S. Marchiafava, R. Mazzeo, J. Milgram, A. Mo-
roianu, P.-A. Nagy, M. Nakamaye, T. Nitta, P. Nurowski, L. Ornea, H. Pedersen,
P. Piccinni, M. Pilca, Y.-S. Poon, E. Rees, P. Rukimbira, S. Salamon, R. Schoen,
L. Schwachhöfer, U. Semmelmann, S. Simanca, M. Singer, J. Sparks, J. Starr, S.
Stolz, A. Swann, Ch. Thomas, G. Tian, M. Verbitsky, M. Wang, J. Wiśniewski, R.
Wolak, D. Wraith, S.-T. Yau, D. Zagier, and W. Ziller. It is probably inevitable
that we have missed the names of some friends and colleagues to whom we deeply
apologize.
We owe our deep gratitude to János Kollár. He is not just a collaborator on
one of the more important papers we wrote. His continuous help and involvement
throughout writing of this book was invaluable. Much of the material presented
in chapters 10 and 11 was written with his expert help and advice. We also wish
to thank Ilka Agricola, Elizabeth Gasparim, Eugene Lerman, Michael Nakamaye,
and Santiago Simanca and Thomas Friedrich for carefully reading certain parts
of our book and providing invaluable comments and corrections. We thank Evan
Thomas for helping with computations used to compile the various tables appearing
in the appendices. We also thank our graduate students: J. Cuadros, R. Gomez,
D. Grandini, J. Kania, and R. Sanchez-Silva for weeding out various mistakes while
sitting in several courses we taught at UNM using early versions of the book.
The second author would like to express special thanks to the Max-Planck-
Institut für Mathematik in Bonn for the generous support and hospitality. Several
chapters of this book were written during K.G.’s sabbatical visit at the MPIM
during the calendar year 2004 and also later during two shorter visits in 2005 and
2006. Both of us thank the National Science Foundation for continuous support of
our many projects, including this one.
Finally, we thank Jessica Churchman and Alison Jones from the Oxford Uni-
versity Press for their patience, continuing interest in our work on the project and
much help in the latter stages of preparing the manuscript for publication.
Last but not least we would like to thank our families, especially Margaret and
Rowan, for support and patience and for putting up with us while we were “working
on The Book” day and night.

Albuquerque, February 2007.

 Contents

Preface iii

Introduction 1

Chapter 1. Structures on Manifolds 9

1.1. Sheaves and Sheaf Cohomology 9
1.2. Principal and Associated Bundles 14
1.3. Connections in Principal and Associated Vector Bundles 18
1.4. G-Structures 23
1.5. Pseudogroup Structures 36
1.6. Group Actions on Manifolds 38

Chapter 2. Foliations 51
2.1. Examples of Foliations 51
2.2. Haefliger Structures 52
2.3. Leaf Holonomy and the Holonomy Groupoid 54
2.4. Basic Cohomology 59
2.5. Transverse Geometry 60
2.6. Riemannian Flows 69

Chapter 3. Kähler Manifolds 75

3.1. Complex Manifolds and Kähler Metrics 76
3.2. Curvature of Kähler Manifolds 82
3.3. Hodge Theory on Kähler Manifolds 88
3.4. Complex Vector Bundles and Chern Classes 92
3.5. Line Bundles and Divisors 94
3.6. The Calabi Conjecture and the Calabi-Yau Theorem 102

Chapter 4. Fundamentals of Orbifolds 105

4.1. Basic Definitions 105
4.2. Orbisheaves and orbibundles 108
4.3. Groupoids, Orbifold Invariants and Classifying Spaces 115
4.4. Complex Orbifolds 123
4.5. Weighted Projective Spaces 133
4.6. Hypersurfaces in Weighted Projective Spaces 138
4.7. Seifert Bundles 144

Chapter 5. Kähler-Einstein Metrics 151

5.1. Some Elementary Considerations 152
5.2. The Monge-Ampère Problem and the Continuity Method 153
5.3. Obstructions in the Positive Case 160
v
vi CONTENTS

5.4. Kähler-Einstein Metrics on Hypersurfaces in CP(w) 162

5.5. Automorphisms and the Moduli Problem 173

Chapter 6. Almost Contact and Contact Geometry 179

6.1. Contact Structures 180
6.2. Almost Contact Structures 190
6.3. Almost Contact Metric Structures 195
6.4. Contact Metric Structures 198
6.5. Structures on Cones 201

Chapter 7. K-Contact and Sasakian Structures 207

7.1. Quasi-regularity and the Structure Theorems 207
7.2. The Transverse Geometry of the Characteristic Foliation 214
7.3. Curvature Properties of K-Contact and Sasakian Structures 219
7.4. Topology of K-Contact and Sasakian Manifolds 229
7.5. Sasakian Geometry and Algebraic Geometry 236
7.6. New Sasakian Structures from Old 250

Chapter 8. Symmetries and Sasakian Structures 257

8.1. Automorphisms of Sasakian Structures and Isometries 257
8.2. Deformation Classes of Sasakian Structures 265
8.3. Homogeneous Sasakian Manifolds 272
8.4. Symmetry Reduction and Moment Maps 276
8.5. Contact and Sasakian Reduction 290

Chapter 9. Links as Sasakian Manifolds 299

9.1. Preliminaries 299
9.2. Sasakian Structures and Weighted Homogeneous Polynomials 300
9.3. The Milnor Fibration and the Topology of Links 302
9.4. The Differential Topology of Links 312
9.5. Positive Sasakian Structures on Links 319
9.6. Links of Complete Intersections 326

Chapter 10. Sasakian Geometry in Dimensions Three and Five 329

10.1. Sasakian Geometry in Dimension Three 329
10.2. Sasakian Structures and the Topology of 5-Manifolds 335
10.3. Sasakian Links in Dimension Five 352
10.4. Regular Sasakian Structures on 5-Manifolds 360

Chapter 11. Sasaki-Einstein Geometry 369

11.1. Foundations of Sasaki-Einstein Geometry 370
11.2. Extremal Sasakian Metrics 378
11.3. Further Obstructions to Sasaki-Einstein Structures 382
11.4. Sasaki-Einstein Metrics in Dimensions Five 388
11.5. Sasaki-Einstein Metrics on Homotopy Spheres 403
11.6. The Sasaki-Einstein semi-group 407
11.7. Sasaki-Einstein Metrics in Dimensions Seven and Higher 409
11.8. Sasakian η-Einstein Metrics 417

Chapter 12. Quaternionic Kähler and Hyperkähler Manifolds 421

12.1. Quaternionic Geometry of Hn and HPn 422
 CONTENTS vii

12.2. Quaternionic Kähler Metrics 428

12.3. Positive Quaternionic Kähler Manifolds and Symmetries 434
12.4. Quaternionic Kähler Reduction 438
12.5. Compact Quaternionic Kähler Orbifolds 443
12.6. Hypercomplex and Hyperhermitian Structures 453
12.7. Hyperkähler Manifolds 455
12.8. Hyperkähler Quotients 458
12.9. Toric Hyperkähler Metrics 461
12.10. ALE Spaces and Other Hyperkähler Quotients 465
Chapter 13. 3-Sasakian Manifolds 473
13.1. Almost Hypercontact Manifolds and 3-Sasakian Structures 474
13.2. Basic Properties 478
13.3. The Fundamental Foliations Fτ and FQ 480
13.4. Homogeneous 3-Sasakian Manifolds 491
13.5. 3-Sasakian Cohomology 494
13.6. Symmetry Reduction 499
13.7. Toric 3-Sasakian Manifolds 506
13.8. Cohomogeneity One 3-Sasakian 7-Manifolds 522
13.9. Non-Toric 3-Sasakian Manifolds in Dimension 11 and 15 525

Chapter 14. Sasakian Structures, Killing Spinors, and Supersymmetry 529

14.1. The Dirac Operator and Killing Spinors 529
14.2. Real Killing Spinors, Holonomy and Bär’s Correspondence 533
14.3. Geometries Associated with 3-Sasakian 7-manifolds 535
14.4. Geometries Associated with Sasaki-Einstein 5-manifolds 542
14.5. Geometric Structures on Manifolds and Supersymmetry 545
Appendix A 551
A.1. Preliminaries on Groupoids 551
A.2. The Classifying Space of a Topological Groupoid 555

Appendix B 559
B.1. Reid’s List of K3 Surfaces as hypersurfaces in CP4 (w) 559
B.2. Differential topology of 2k(S 3 × S 4 ) and 2k(S 5 × S 6 ) 560
B.3. Tables of Kähler-Einstein metrics on hypersurfaces CP(w) 561
B.4. Positive Breiskorn-Pham Links in Dimension 5 564
B.5. The Yau-Yu Links in Dimensions 5 567
Bibliography 569
Index 607
 Introduction

In 1960 Shigeo Sasaki [Sas60] began the study of almost contact structures
in terms of certain tensor fields, but it wasn’t until [SH62] that what are now
called Sasakian manifolds first appeared under the name of “normal contact metric
structure”. By 1965 the terms “Sasakian structure” and “Sasakian manifold” be-
gan to be used more frequently replacing the original expressions. For a number of
years these manifolds were intensively studied by a group of Japanese geometers.
The subject did get some attention in the United States mainly due to the papers
of Goldberg and Blair. Nevertheless, the main interest in the field remained in-
side Japan, finding it hard to spread out and attract broader attention beyond its
birthplace either in the United States or in Europe.
Over a period of four years between 1965 and 1968 Sasaki wrote a three part
set of lecture notes which appeared as an internal publication of the Mathematical
Institute of the Tôhoku University under the title Almost contact manifolds, Part
I-III [Sas65, Sas67, Sas68]. Put together, the work amounts to almost 500 pages.
Even today, after 40 years, the breath, depth and the relative completeness of the
Sasaki lectures is truly quite remarkable. It is hard to understand why they did not
make it as a monograph in some prestigious Western book series; it is a pity. As it
is, the notes are not easily available and, consequently, not well-known1. Outside
Japan the first and important attempt to give a broader account of the subject was
given eight years later by Blair [Bla76a].
After 1968 Sasaki himself was less active although he continued to publish
until 1980. Yet he had already created a new subfield of Riemannian geometry
which slowly started to attract attention worldwide, not just in Japan. In 1966
Brieskorn wrote his famous paper describing a beautiful geometric model for all
homotopy spheres which bound parallelizable manifolds [Bri66]. In 1976 Sasaki
[ST76, SH76] realized that Brieskorn manifolds admit almost contact and con-
tact structures. (This very important fact was independently observed by several
other mathematicians: Abe-Erbacher [AE75], Lutz-Meckert [LM76], and Thomas
[Tho76].) Thirty years later the Brieskorn-Pham links as well as more general links
of weighted homogeneous polynomials, are the key players in several chapters of
our book. Yet again, Sasaki seemed to have had both the necessary intuition and
a broad vision in understanding what is and what is not of true importance2.
1We became aware of the Sasaki notes mainly because of our work on this book. We obtained
a copy of the lectures in 2003 form David Blair and we would like to thank him for sharing them
with us.
2We know little about Sasaki’s non-mathematical life. He was born in 1912 and what we do
know is from the volume of his selected papers edited in 1985 by Tachibana [Sas85]. There one
finds a short introduction by S. S. Chern and an essay by Sasaki in which he mostly discusses
his life as a working mathematician. He apparently died almost 20 years ago on August 12th,
1987. Sadly, his death passed without any notice, strangely forgotten. We could not find an
1
2 INTRODUCTION

Over the years Sasakian geometry has taken a back seat to other areas of
Riemannian geometry, most prominently to the study of Riemannian geometry
with reduced holonomy groups. Nevertheless, as we shall see, Sasakian geome-
try is closely related to all these other geometries. Although generally a Sasakian
manifold has the generic holonomy SO(n), the Riemannian cone over a Sasakian
manifold does have reduced holonomy making the study of Sasakian geometry quite
tractable. A quick perusal of Berger’s list of possible irreducible Riemannian holo-
nomy groups (see Table 1.4.1 below) shows that there are five infinite series of these
holonomy groups. All five are related to Sasakian geometry: generally a Sasakian
metric itself has generic holonomy SO(n), and its Riemannian cone has holonomy
U (n); the Riemannian cone of a general Sasaki-Einstein structure has holonomy
SU (n); whereas, the Riemannian cone of a general 3-Sasakian structure has holo-
nomy Sp(n). The remaining infinite series in Table 1.4.1 is the group Sp(n)Sp(1)
of quaternionic Kähler geometry which is closely related to 3-Sasakian geometry as
we discuss shortly. The two remaining irreducible Riemannian holonomy groups
G2 and Spin(7) are related to Sasakian geometry in a less direct way as we discuss
in Chapter 14. In this same spirit our book attempts to show Sasakian geometry
not as a separate subfield of Riemannian geometry but rather through its inter-
relation to other geometries. This is perhaps the most important feature of the
subject. The study of Sasakian manifolds brings together several different fields
of mathematics from differential and algebraic topology through complex algebraic
geometry to Riemannian manifolds with special holonomy.
The closest relative of Sasakian geometry is Kählerian geometry, the importance
of which is difficult to overestimate mainly because of its role in algebraic geometry.
But Sasakian geometry also has a very algebro-geometric flavor. In fact, there is an
‘algebraic structure’ on every Sasakian manifold; whereas, Voisin has shown recently
that there are Kähler manifolds that admit no algebraic structure whatsoever. To
better understand the relation between Sasakian and Kählerian geometries we begin
with the more familiar relation between contact and symplectic geometries. Let
(M, η, ξ) be a contact manifold where η is a contact form on M and ξ is its Reeb
vector field. It is easy to see that the cone (C(M ) = R+ × M, ω = d(tη)) is
symplectic. Likewise, the Reeb field defines a foliation of M and the transverse
space Z is also symplectic. When the foliation is regular the transverse space is a
smooth symplectic manifold giving a projection π called Boothby-Wang fibration,
and π ∗ Ω = dη relates the contact and the symplectic structures as indicated by

(C(M ), ω) ←- (M,η, ξ)


π
y
(Z, Ω).

English language mathematical obituary honoring his life and work and commemorating his death;
apparently, his passing was noted just in a short obituary of a local newspaper. After his death
Tokyo Science University where he briefly worked after retiring from Tôhoku took care of his
mathematical heritage with some of his manuscripts placed at the university library. We are
grateful to Professor Yoshinobu Kamishima for this information.
 INTRODUCTION 3

We do not have any Riemannian structure yet. It is quite reasonable to ask

if there is a Riemannian metric g on M which “best fits” into the above diagram.
As the preferred metrics adapted to symplectic forms are Kähler metrics one could
ask for the Riemannian structure which would make the cone with the warped
product metric ḡ = dt2 + t2 g together with the symplectic form ω into a Kähler
manifold? Then ḡ and ω define a complex structure Φ̄. Alternatively, one could ask
for a Riemannian metric g on M which would define a Kähler metric h on Z via a
Riemannian submersion. Surprisingly, in both cases the answer to these questions
leads naturally and uniquely to Sasakian Geometry. Our diagram becomes

(C(M ), ω, ḡ, Φ̄) ←- (M, ξ,η, g, Φ)



π
y
(Z, Ω, h, J).

From this point of view it is quite clear that Kählerian and Sasakian geome-
tries are inseparable, Sasakian Geometry being naturally sandwiched between two
different types of Kählerian Geometry. Yet the two fared differently over the years.
Since Erich Kähler’s seminal article “several dominant figures of the mathematical
scene of the XXth century have, step after step along a 50 year period, transformed
the subject into a major area of Mathematics that has influenced the evolution of
the discipline much further than could have conceivably been anticipated by any-
one” writes Jean-Pierre Bourguignon in his tributary article The unabated vitality
of Kählerian geometry published in [Käh03]. Sasakian Geometry has not been as
lucky. There always has been interesting work in the area, but for unclear reasons
it has never attracted people with the same broad vision, people who would set
out to formulate and then work on fundamental problems. Yet, arguably Sasakian
manifolds are at least as interesting as Kählerian manifolds.
Our own research of the last decade has been an attempt to bring Sasakian
geometry back into the main stream. We believe that a modern book on Sasakian
manifolds is long overdue. Most of the early results are scattered, often buried
in old and hard to get journals. They are typically written in an old fashioned
language. Worse than that, articles with interesting results are drowned in a vast
sea of papers of little importance. There are very few graduate level texts on the
subject. There is the book Structures on manifolds written by Yano and Kon in 1984
[YK84], but this book is over 20 years old and treats both Sasakian and Kählerian
geometries as a subfield of Riemannian geometry. Recently, Blair substantially
updated his well-known Contact manifolds in Riemannian geometry in the Springer
Lecture Notes series [Bla76a] with Riemannian geometry of contact and symplectic
manifolds [Bla02]. Again the major emphasis as well as the techniques used are
Riemannian in nature. Our monograph naturally complements Blair’s as it employs
an entirely different philosophy and follows a different approach. First we develop
the important relations between Sasakian geometry and the algebraic geometry of
Kähler (actually projective algebraic) orbifolds. Secondly, our major motivation
to begin with was in proving the existence of Einstein metrics. So we have the
understanding of Sasaki-Einstein metrics as a main goal toward which to work, but
4 INTRODUCTION

we have also come to appreciate the beauty and richness of Sasakian geometry in
its own right.
Our book breaks more or less naturally into two parts. Chapters 1 through
9 provide an introduction to the modern study of Sasakian geometry. Starting
with Chapter 10 the book becomes more of a research monograph describing many
of our own results in the subject. However, the extensive introduction should
make it accessible to graduate students as well as non-expert researchers in related
fields. We have used parts or our monograph as a textbook for advanced graduate
courses. For example, assuming that the students have some basic knowledge of
Riemannian geometry, algebraic geometry, and some algebraic topology, a course
treating Sasakian geometry starting with Chapter 4 through Chapter 11 is possible,
drawing on the first three chapters as review. Many of the results in Chapters 4-11
are given with full proofs bringing the student to the forefront of research in the area.
As a guide we use the principle that proofs, or at least an outline of proofs, are given
for important results that are not found in another book. Our book also contains
many examples, for we believe that the learning process is substantially enhanced
by working through examples. We also have exercises scattered throughout the
text for the reader to sharpen her/his skills. Open problems of varying or unknown
difficulty are listed, some of which could be the basis of a dissertation. The text is
aimed at mathematicians, but we hope it will find many readers among physicists,
particularly those working in Superstring Theory.
We begin in Chapter 1 by introducing various geometries that play more or
less important roles in the way they relate to Sasakian structures. We espouse the
point of view that a geometric structure is best described as a G-structure which,
in addition, may or may not be (partially) integrable. As Sasakian manifolds are all
examples of Riemannian foliations with one-dimensional leaves, Chapter 2 takes the
reader into the world of foliations with a particular focus on the Riemannian case.
The literature is full of excellent books on the subject so we just select the topics
most relevant to us. Chapter 3 reviews some basic facts about Kähler manifolds.
Again, we are very selective choosing only what is needed later in describing the
two Kähler geometries of the “Kähler-Sasaki sandwich”. Of particular interest is
Yau’s famous proof of the Calabi conjecture.
A key tool that allows for connecting Sasakian structures to other geometric
structures is the theory of Riemannian orbifolds and orbifold bundles or ‘orbibun-
dles’. For that reason Chapter 4 is crucial in setting the stage for an in depth study
of Sasakian manifolds which begins later in Chapter 7. Orbifolds just as manifolds
have become a household name to the well trained geometer. Nevertheless, a lot
of important results are scattered throughout the literature, and orbifolds typically
appear within a specific context. There is a forthcoming book on orbifolds, Orb-
ifolds and Stringy Topology [ALR06] by Adem, Leida, and Ruan, but this has a
particularly topological bent. Hence, we take some time and effort to prepare the
reader introducing all the basic concepts from the point of view needed in subse-
quent chapters. By Chapter 5 we are ready for a second trip into the realm of
Kähler geometry. However, now the focus is on Kähler-Einstein metrics, in partic-
ular positive scalar curvature Kähler-Einstein metrics on compact Fano orbifolds.
We introduce some basic techniques that allow for proving various existence results.
We also briefly discuss obstructions. Chapter 6 presents the necessary foundational
 INTRODUCTION 5

material on almost contact and contact geometry. This leads directly to the defi-
nition of a Sasakian structure introduced at the very end.
The study of Sasakian geometry finally begins with Chapter 7. We first present
the important structure theorems, and then gather all the results concerning the
geometry, topology, and curvature properties of both K-contact and Sasakian man-
ifolds. Most of the curvature results are standard and can be found in Blair’s book
[Bla02], but our main focus is different: we stress the relation between Sasakian
and algebraic geometry, as well as the basic cohomology associated with a Sasakian
structure. A main tool used in the text is the transverse Yau Theorem due to El
Kacimi-Alaoui. In the companion Chapter 8 we present known results concerning
symmetries of Sasakian structures. We introduce the Sasakian analogue of the bet-
ter known symplectic/contact reduction. Then we study toric contact and toric
Sasakian manifolds and prove several Delzant-type results.
Chapter 9 is devoted to the geometry of links of isolated hypersurface singulari-
ties as well as a review of the differential topology of homotopy spheres a la Kervaire
and Milnor [KM63]. A main reference for the study of such links is Milnor’s classic
text Singular points of complex hypersurfaces [Mil68], but also Dimca’s Singular-
ities and topology of hypersurfacs [Dim92] is used. The differential topology of
links is a beautiful piece of mathematics, and this chapter offers a hands-on “user’s
guide” approach with much emphasis on the famous work of Brieskorn [Bri66]. Of
importance for us is that when the singularities arise from weighted homogeneous
polynomials the links have a natural Sasakian structure with either definite (posi-
tive or negative) or null basic first Chern class. Emphasis is given to the positive
case which corresponds to having positive Ricci curvature. In Chapter 10 we dis-
cuss the Sasakian geometry in low dimensions. In dimension 3 there is a complete
classification. Dimension 5 is large enough to be interesting, yet small enough to
hope for some partial classification. We concentrate on the simply connected case
as there we can rely on the Smale-Barden classification. In terms of Sasakian struc-
tures our main focus is on the case of positive Sasakian structures. In considerable
detail we describe several remarkable theorems of Kollár which show how positivity
severely restricts the topology of a manifold which is to admit a positive Sasakian
structure.
Chapter 11 is central to the whole book and perhaps the main reason and jus-
tification for it. Much of this chapter is based on a new method for proving the ex-
istence of Einstein metrics on odd dimensional manifolds introduced by the authors
in 2001 [BG01b]. We realized there that links of isolated hypersurface singulari-
ties obtained from weighted homogeneous polynomials admit Sasakian structures.
Moreover, by using an orbifold version [BG00b] of an old result of Kobayashi, we
successfully tied the problem to making use of the continuity method for proving
the existence of Kähler-Einstein metrics on compact Kähler orbifolds. For the au-
thors this was the original “raison d’etre” for Chapters 4, 5, and 9 of the book.
In a series of papers the authors and their collaborators have successfully applied
this method to prove the existence of Sasaki-Einstein metrics on many 5-manifolds,
on odd dimensional homotopy spheres that bound parallelizible manifolds, as well
as on odd dimensional rational homology spheres. Furthermore, our method has
been substantially generalized by Kollár who has pushed our understanding much
further especially in dimension five. Although a complete classification is perhaps
6 INTRODUCTION

not within reach we now begin to have a really good grasp of Sasaki-Einstein ge-
ometry in dimension 5. In addition, we discuss toric Sasaki-Einstein geometry in
dimension 5 which began with the work of Gauntlett, Martelli, Sparks, and Wal-
dram [GMSW04b], and culminates with the very recent work of Cho, Futaki,
Ono, and Wang [FOW06, CFO07] which shows in arbitrary odd dimension that
any toric Sasakian manifold with positive anticanonical Sasakian structure admits
a compatible Sasaki-Einstein metric. We also discuss extremal Sasaki metrics de-
fined in analogy with the extremal Kähler metrics, and introduce the Sasaki-Futaki
invariant [BGS06]. In addition to lifting the well-known obstructions of positive
Kähler-Einstein metrics we also present some new results due to Gauntlett, Martelli,
Sparks, and Yau [GMSY06] involving two well known estimates, one due to Lich-
nerowicz, and the other to Bishop. We also present Sasaki-Einstein metrics obtained
via the join construction described earlier in Chapter 7. We end this long chapter
with a brief discussion of Sasakian-η-Einstein metrics.
Chapter 12 gives an extensive overview of various quaternionic geometries. The
main focus is on the positive quaternionic Kähler (QK) manifolds (orbifolds) and
on the hyperkëhler manifolds (orbifolds). The reason for such an extensive treat-
ment has to do with Chapter 13. The 3-Sasakian manifolds studied there cannot
be introduced without a deeper understanding of these two quaternionic geome-
tries, just as Sasakian and Sasaki-Einstein manifolds cannot be studied without
Kählerian and Kähler-Einstein geometry. The Sasaki-Einstein manifolds of Chap-
ter 13 have a completely different flavor than the ones that appeared in Chapter
11. It is not only that these occur only in dimensions 4m + 3, but also that they
have a somewhat richer geometric structure. In addition the method in which
the metrics are obtained is completely different. In Chapter 11, with some excep-
tions, we mostly get our existence results via the continuity method applied to the
Monge-Ampère equation. Very few metrics are known explicitly, though there are
exceptions. Most of the 3-Sasakian metrics we consider are obtained via symmetry
reduction similar to the hyperkähler and quaternionic Kähler reduction. Indeed the
three quotients are all related. So the manifolds and the metrics we get are quite
often explicit and can be studied as quotients. Again, there are some exceptions.
Finally, Chapter 14 gives a very brief overview of the rich theory of Killing spinors.
There we describe some other geometries and show how they relate to 3-Sasakian
7-manifolds and Sasaki-Einstein 5-manifolds. At the end we very briefly comment
on how Sasakian geometry naturally appears in various supersymmetric physical
theories. Both Sasaki-Einstein geometry and geometries with exceptional holonomy
have appeared in various models of supersymmetric String Theory fuelling vigorous
interest in them by mathematicians and physicists alike.
We also have added two appendices. The first appendix gives a very brief in-
troduction to groupoids and their classifying spaces which are employed in Chapter
4, while the second gives many tables listing links of hypersurface singularities that
are used throughout the book.
We have compiled a very extensive bibliography. There are various reasons for
its size. We should remark that in this day and age of easy internet access, with
MathSciNet and Google, it would make no sense to simply compile a bibliography
of every paper with the words ‘Sasaki’ or ‘Sasakian’ in the title. Anyone with an
access to MathSciNet can easily compile such a list of 809 papers3 so it would serve
3As we checked on February 19, 2007.
 INTRODUCTION 7

no purpose to do it for this book. In a way, in spite of its size, we were very careful
and selective in choosing all bibliography items. Our book brings together so many
different areas of mathematics, that having good references becomes essential. In
some cases the proofs we give are only sketches and in such instances we wanted
to refer the reader to the best place he/she could find more details. That is quite
often the original source but not always. We refer to various books, monographs,
and lecture notes. We have tried to be both selective and accurate in attributing
various results with care. This can be at times a hard task. We suspect that we
did not always get the references and proper credits exactly right. This is almost
inevitable considering we are not experts in many of the areas of mathematics that
substantially enter as part of the book. In any case, we apologize for any omissions
and errors; these are certainly not intentional4.
Finishing this book was a challenging task. We had to deal with an increasing
number of new and interesting results appearing every few months. It is always a bit
dangerous to include material based on articles that have not yet been published.
To make the book as up-to-date as possible we took the risk to include some of
these new results without giving proofs. On the other hand we are happy to see
that the field is active and we very much hope that our book will become somewhat
outdated in a few years.

4There are many examples in the literature where incorrect attributions are made. A recent
case that we just uncovered is that of nearly Kähler manifolds which are usually attributed to
Gray [Gra69b, Gra70], yet they were discovered 10 years earlier by Tachibana [Tac59]. The
fact that any nearly Kähler 6-manifold is Einstein is also attributed to Gray in [Gra76], yet it
was proven earlier by Matsumoto [Mat72]. See Section 14.3.2.
 CHAPTER 1

Structures on Manifolds

A unifying viewpoint for doing differential geometry involving different struc-

tures is that of a G-structure, where G can stand for ‘geometric’ or more appro-
priately a ‘Lie group’. A (first order) G-structure is just a reduction of the bundle
of frames of a manifold from the general linear group to a subgroup. For example,
from the point of view of G-structures a Riemannian metric on a manifold corre-
sponds to a reduction of the frame bundle with group GL(n, R) to the orthonormal
frame bundle with subgroup, the orthogonal group O(n, R). Many other examples
will be given below. It will be important to have at our disposal the general theory
of connections in principal and associated fibre bundles of which the G-structures
mentioned above are special cases. However, before embarking on the study of such
structures we give a short review of sheaves and their cohomology groups. Sheaves
are more general than bundles, but they are a bit too general for describing geomet-
ric structures on manifolds. They are mainly used to pass from local information
to global information.

1.1. Sheaves and Sheaf Cohomology

Sheaves, which were invented by Jean Leray as a prisoner of war during World
War II, have become an important tool in geometry. Here we give a very brief tour
of sheaf theory referring to the literature [Bre97, GR65, Hir66, Wel80] for more
detail and proofs. Our presentation follows [Wel80] fairly closely.
Definition 1.1.1: Let X be a topological space. A presheaf on X is an assignment
to each nonempty open set U ⊂ X a set F(U ) together with maps, called restriction
maps,

ρU
V : F(U )−
−−−→F(V )

for each pair of open sets U and V with V ⊂ U that satisfy the conditions ρU
U = idU
and ρUW = ρ U
V ◦ ρ V
W whenever, W ⊂ V ⊂ U.
Very often the sets F(U ) have some additional algebraic structure, such as a
group, a ring, or a module structure. In this case we assume that the restriction
maps preserve the algebraic structure. So a presheaf is a contravariant functor
from the category of open sets of a fixed topological space with inclusion maps as
morphisms to the category of groups, rings, or modules whose morphisms are the
homomorphisms of that category.
9
10 1. STRUCTURES ON MANIFOLDS

A morphism of presheaves f : F−−→G on X is a family of homomorphisms

fU : F(U )−−→G(U ) such that the diagram
hU
F(U −−−−−→
 ) G(U
)
 
 
(1.1.1) ρUV ρUV
 
y y
hV
F(V ) −−−−−→ G(V )
commutes, where V ⊂ U ⊂ X. If the homomorphisms fU : F(U )−−→G(U ) are
monomorphisms then F is a subpresheaf of G.
Given a presheaf F over X we can consider the direct limit
Fx = lim
−→
F(U )
x∈U

with respect to the restriction maps {ρU V }. Clearly, Fx inherits whatever algebraic
structure the sets F(U ) have. Fx is called the stalk of F at x, and for x ∈ U there
is a natural projection ρUx : F(U )− −→Fx by sending an element s ∈ F(U ) to its
equivalence class sx in the direct limit. sx is called the germ of s at x.
Definition 1.1.2: A presheaf F is called a sheaf if for every collection Ui of open
subsets of X with U = ∪Ui , the presheaf F satisfies the following two conditions
(i) If s, t ∈ F(U ) and ρU U
Ui (s) = ρUi (t) for all i, then s = t.
(ii) If si ∈ F(Ui ) and if for any j with Ui ∩ Uj 6= ∅ the equality ρU
Ui ∩Uj (si ) =
i

U
ρUji ∩Uj (sj ) holds for all i, then there is an s ∈ F(U ) such that ρUUi (s) = si
for all i.
A morphism of sheaves is a morphism of the underlying presheaves. In par-
ticular, an isomorphism of sheaves is a morphism of sheaves such that the all the
maps fU : F(U )−−→G(U ) are isomorphisms. Not every presheaf is a sheaf. For
example consider C and the presheaf that assigns to every open set U ⊂ C the
algebra of bounded holomorphic functions B(U ) on U. Since there are no bounded
holomorphic functions on all of C this presheaf violates condition (ii) of definition
1.1.2 above. Nevertheless, as we shall see shortly we can associate a sheaf to any
presheaf in a fairly natural way. First we give an important example of a presheaf
that is a sheaf.
Example 1.1.3: Let X and Y be topological spaces and consider the presheaf CX,Y
on X which associates to every open subset U of X the set CX,Y (U ) of all continuous
maps from U to Y. The restriction map ρU V is just the natural restriction, i.e. if
f ∈ CX,Y (U ) and V ⊂ U, then ρU V (f ) = f |V . It is easy to see that CX,Y satisfies
the two conditions of Definition 1.1.2 and thus defines a sheaf on X. An important
special case is obtained by taking Y to be either of the continuous fields F = R or
C. In this case we note that CX,F has the structure of an F-algebra. Two particular
cases of interest are when X is either a smooth real manifold or a complex manifold.
In these cases we are more interested in the subsheaf EX ⊂ CX,R of smooth functions
and the subsheaf OX ⊂ CX,C of holomorphic functions, respectively. Such sheaves
are called the structure sheaf of X.
Definition 1.1.4: An étale space over a topological space X is a topological space
Y together with a continuous map π : Y −−→X that is a local homeomorphism.
 1.1. SHEAVES AND SHEAF COHOMOLOGY 11

For each open set U ⊂ X we can consider the set Γ(U, Y ) of continuous sections
of π, that is, the subset of f ∈ CX,Y (U ) that satisfy π◦f = idU . Then the assignment
that assigns to each open subset U of X the set of continuous sections Γ(U, Y ) forms
a subsheaf of CX,Y which we denote by SX,Y . We shall now associate to any presheaf
F an étale space F̃. Then by taking sections of F̃ we shall get a sheaf. Define F̃ by

(1.1.2) F̃ = tx∈X Fx .

Let π : F̃−−→X denote the natural projection that sends any sx ∈ Fx to x. We

define a topology on F̃ as follows: for each s ∈ F(U ) define a map s̃ : U −−→F̃ by
s̃(x) = sx . This satisfies π ◦ s̃ = idU . Then we let the sets

{s̃ | U is open in X, s ∈ F(U )}

be a basis for the topology on F̃. Then π and all functions s̃ are continuous, and it
is easy to check that π is a local homeomorphism, so that F̃ is an étale space over
X. But we have already seen that the set of sections of an étale space form a sheaf.
We denote the sheaf associated to the étale space F̃ by S̃X,Y . Thus, beginning
with a presheaf F we have associated a sheaf S̃X,Y . This sheaf is called the sheaf
associated to or generated by the presheaf F.
Exercise 1.1: Show that if one starts with a sheaf F then the sheaf associated to
F is isomorphic to F.
Definition 1.1.5: A ringed space is a pair (X, A) consisting of a topological
space X together with a sheaf of rings A on X, called the structure sheaf. (X, A)
is a locally ringed space if for each point x ∈ X, the stalk Ax is a local ring.
All ringed spaces considered in this book will be locally ringed spaces, so we
often omit the word ‘locally’ and just refer to a ringed space. We now consider the
structure sheaves EM and OM on a real or complex manifold M, respectively. We
denote by A the sheaf E on a real manifold or the sheaf O on a complex manifold.
Definition 1.1.6: Let (X, A) be a ringed space with structure sheaf A given by E
(or O), respectively. A sheaf F of A-modules is said to be locally free of rank
r if X can be covered by open sets U such that there is an isomorphism of sheaves
r times
F|U ≈ A|U ⊕ ··· ⊕ A|U . A locally free rank 1 sheaf is called an invertible sheaf.
Let π : E−−→M be an F-vector bundle over a smooth manifold M . As in the
case of an étale space the subset of f ∈ CM,E satisfying π ◦ f = idM defines a
subsheaf of CM,E , called the sheaf of germs of continuous sections of E. Actually,
we are more interested in the sheaf of germs of smooth sections of E which we
denote by E(E). On a complex manifold M a C-vector bundle can have a special
type of structure.
Definition 1.1.7: Let M be a complex manifold. A complex vector bundle π :
E−−→M on M is said to be holomorphic if E is a complex manifold with π
holomorphic, and the transition functions for E can be taken to be holomorphic
functions.
Of course, the transition functions being holomorphic is equivalent to the local
trivializations being holomorphic. Then we have
Proposition 1.1.8: Let M be a real or complex manifold. There is a one-to-one
correspondence between smooth vector bundles on M and locally free EM -sheaves on
12 1. STRUCTURES ON MANIFOLDS

M. Similarly, when M is a complex manifold, there is a one-to-one correspondence

between holomorphic vector bundles on M and locally free OM -sheaves on M.
The correspondence is given by associating to a holomorphic vector bundle
E the sheaf of germs of holomorphic sections of E. Because of Proposition 1.1.8
we shall often use interchangeably the concepts of vector bundles and locally free
sheaves. Of particular interest is the rank one case in the holomorphic category
which gives a one-to-one correspondence between holomorphic line bundles and
invertible OM -sheaves.
As mentioned previously the power of sheaf theory is as a tool in passing from
the local to the global. This is accomplished by the way of sheaf cohomology theory
which we now briefly describe.

1.1.1. Sheaf Cohomology. There are several approaches to sheaf cohomol-

ogy theory. There is the axiomatic approach used in [GR65], the derived functor
approach using the resolution by discontinuous sections due to Godement [God58]
and espoused in [Bre97, Wel80], and finally the Čech theoretic approach described
in [GH78b] and [Kod86]. We prefer to begin with the latter and then present the
so-called Abstract de Rham Theorem using resolutions as in [Wel80].
For simplicity we assume that X is paracompact and Hausdorff. Let F be a
sheaf X, and let U = {Ui }i be a locally finite cover by open sets. Define the set
C p (U, F) of p-chains of F by
Y
(1.1.3) C p (U, F) = F(Ui0 ∩ · · · ∩ Uip ),
i0 ,...,ip

where we assume that the indices ij are distinct. We denote elements of C p (U, F)
by σi0 ,...,ip . The coboundary operator δp : C p (U, F)−−→C p+1 (U, F) is defined by
p+1
X
(1.1.4) (δp σ)i0 ,...,ip+1 = (−1)j σi0 ,...,ibj ,...,ip+1 |Ui0 ∩···∩Uip ,
j=0

where as usual î means remove that index. This gives rise to a cochain complex
δ0 δ1 δp
C 0 (U, F) −−→ C 1 (U, F) −−→ · · · −−→C p (U, F) −−→ C p+1 (U, F)−−→ · · ·
and we define the cohomology groups H p (U, F) by
kerδp
H p (U, F) = .
Imδp−1
If V is a cover of X refining U, there are homomorphisms ψ : H p (U, F)−−→H p (V, F).
Thus, we can take the direct limit of the cohomology groups H p (U, F) as the cover
becomes finer and finer, so we define the cohomology group with coefficients in the
sheaf F by
(1.1.5) H p (X, F) = lim
−→
H p (U, F).
U
p
Note that generally the groups H (U, F) depend on the cover, but there are certain
special covers known as acyclic covers for which we have the following theorem of
Leray
Theorem 1.1.9: If the cover U is acyclic in the sense that H q (Ui0 ∩ · · · ∩ Uip , F)
vanishes for all q > 0 and all i0 , . . . , ip , then H ∗ (M, F) ≈ H ∗ (U, F).
 1.1. SHEAVES AND SHEAF COHOMOLOGY 13

In practice, one always computes H ∗ (U, F) for an acyclic cover and then uses
the above theorem.
P3
Example 1.1.10: Consider the sphere S 2 = {(x1 , x2 , x3 ) ∈ R3 | i=1 (xi )2 = 1} and
the constant sheaf F = R, together with two covers. The first is the ‘stereographic
cover’ defined by two open sets U± = S 2 − {(0, 0, ±1)}. This cover is not acyclic
since U+ ∩ U− has the homotopy type of a circle. The second cover is a cover
consisting of a cover of the lower hemisphere by 3 open sets together with the
upper hemisphere. This gives an acyclic cover from which one can compute the
cohomology groups H i (S 2 , R) of S 2 . See Example 9.3 of [BT82].
The sheaf cohomology groups have some nice properties. For any sheaf F the
cohomology group H 0 (X, F) is the space of global sections F(X) of F, and a sheaf
morphism f : A−−→B induces a group homomorphism fq : H q (X, A)−−→H q (X, B).
Moreover, a short exact sequence of sheaves,

0−−−→A−−−→B−−−→C−−−→0

induces a long exact sequence in cohomology,

(1.1.6)
δ δ
0−→H 0 (M, A)−→H 0 (M, B)−→H 0 (M, C) −→ H 1 (M, A)−→ · · · −→ H q (M, A)−→ · · · ,

where δ is the well-known connecting homomorphism.

To proceed further we need some definitions.
Definition 1.1.11: A sheaf F is soft if for any closed subset S ⊂ X, the restriction
map ρX U : F(X)− −→F(S) is surjective. F is fine if for any locally finite open cover
{Ui } of X there is a partition
P of unity of F, i.e. there is a family of sheaf morphisms
υi : F−−→F such that υi = 1, and υi (Fx ) = 0 for all x in some neighborhood of
X − Ui .
The importance of soft sheaves is the following
Proposition 1.1.12: If F is a soft sheaf, then H q (X, F) = 0 for all q > 0.
Fine sheaves are special cases of soft sheaves, viz.
Proposition 1.1.13: Fine sheaves are soft.
Example 1.1.14: The structure sheaf EM of a real smooth manifold is fine, thus
soft. The structure sheaf OM of a complex manifold is neither fine nor soft, nor are
the constant sheaves.
Definition 1.1.15: Let F be a sheaf on X. A resolution of F is an exact sequence
of sheaves F i of the form

0−−→F−−→F 0 −−→F 1 −−→ · · · −−→F k −−→ · · · .

The resolution is acyclic if H q (X, F i ) = 0 for all q > 0 and all i ≥ 0.

A resolution of a sheaf F is conveniently written in the shorthand notation
0−−→F−−→F ∗ . By Proposition 1.1.12 if the sheaves F i of a resolution are soft
sheaves, then the resolution is acyclic. Thus, a resolution by soft or fine sheaves is
acyclic. A resolution of sheaves gives rise to a cochain complex C ∗ = C ∗ (F ∗ ) on
global sections
d0 d1 dk−1 dk
(1.1.7) 0−−→F(X)−−→F 0 (X) −−→ F 1 (X) −−→ · · · −−→ F k (X) −−→ · · · .
14 1. STRUCTURES ON MANIFOLDS

In general this sequence is exact only at F(X), but we do have dk+1 ◦ dk = 0. The
cohomology group
ker dk
H k (C ∗ ) =
im dk−1
of this complex is called the k th derived group of C ∗ . We are now ready for the
Abstract de Rham Theorem:
Theorem 1.1.16: Let F be a sheaf over a paracompact Hausdorff space X, and
let 0−−→F−−→F ∗ be an acyclic resolution of F. Then there is a natural isomorphism
ker dk
H q (X, F) ≈ H q (C ∗ ) = .
im dk−1
This is a very powerful theorem of which the usual de Rham theorem is a special
case. We give this as an example below. The Dolbeault Theorem is another special
case which will be given in Chapter 3.
Example 1.1.17: de Rham’s Theorem. Let M be a smooth manifold and let
E p denote the sheaf of germs of sections of the exterior bundles Λp M with E 0 = E.
Let F = R the constant sheaf on M. Then by the well-known Poincaré Lemma we
get a resolution of the constant sheaf R, viz
0−−→R−−→E 0 −−→E 1 −−→ · · · −−→E n−1 −−→E n −−→0.
Since the sheaves E p are fine this resolution is acyclic, so by the abstract de Rham
Theorem 1.1.16 we have an isomorphism
µ ¶
k k ker dk k
H (M, R) ≈ H = HdeRh (M, R).
im dk−1
Moreover, H k (M, R) can be identified with the k th singular cohomology group by
taking an acyclic resolution of the constant sheaf R by sheaves of singular cochains.
See [Wel80].

1.2. Principal and Associated Bundles

We begin by considering principal fibre bundles. The main reference here is
the classic text of Kobayashi and Nomizu[KN63, KN69].
Definition 1.2.1: A principal bundle P over M with Lie group G consists of:
(1) A pair of smooth manifolds P and M together with a smooth surjection π : P →
M.
(2) A smooth free action of the Lie group G on P , R : P × G → P given by the
“right” action (u, a) 7→ Ra u = ua such that the quotient is M , i.e., M = P/G.
(3) P is locally trivial. More precisely, every point p ∈ M has a neighborhood U
and a diffeomorphism χ : π −1 (U ) → U × G defined by χ(u) = (π(u), φ(u)), where
φ : π −1 (U ) → G satisfies the compatibility condition φ(ua) = φ(u)a.
As usual the manifold P is called the total space, the manifold M is called the base
space, and for p ∈ M , π −1 (p) is called the fibre at p. If we fix a point u0 ∈ π −1 (p)
the map sending a ∈ G to u0 a ∈ π −1 (p) identifies π −1 (p) with G. Moreover, one
easily sees that this map is a diffeomorphism from G to π −1 (p) whose inverse is φ.
When we want to emphasize the base space and the group we write P (M, G) for P ,
or we refer to P as a principal G-bundle over M. Notice that for any open subset
U ⊂ M, π −1 (U ) is a principal G-bundle over U. Another example is the “trivial
G-bundle” when P is just the product M × G.
 1.2. PRINCIPAL AND ASSOCIATED BUNDLES 15

Now let {Uα } be an open cover of M such that for each α the map χα :
π −1 (Uα )−−→Uα × G is a diffeomorphism, and satisfies the compatibility condition
of Definition 1.2.1. Then if u ∈ π −1 (Uα ∩ Uβ ) we see that φβ (ua)(φα (ua))−1 =
φβ (u)(φα (u))−1 for all a ∈ G. Hence, we can define a smooth map ψβα : Uα ∩
Uβ −−→G by

(1.2.1) ψβα (π(u)) = φβ (u)(φα (u))−1 .

The smooth maps ψβα are called transition functions for the principal bundle
P (M, G), and one easily sees that for each p ∈ Uα ∩ Uβ ∩ Uγ they satisfy the
‘cocycle condition’
(1.2.2) ψγα (p) = ψγβ (p)ψβα (p).
The appellation ‘cocycle condition’ comes from the fact that equivalence classes of
transition functions are elements of the sheaf cohomology set H 1 (M, G), where G
denotes the sheaf of germs of maps from opens sets of M to G. Here two transition
0
functions ψβα and ψβα are equivalent if there are maps ψα : Uα −−→G such that
ψβα (p) = ψβ (p)ψβα (p)(ψα (p))−1 , and the inverse element in G is given by ψαβ (p) =
0

(ψβα (p))−1 . The reader is referred to Section 1.1 below for more details on sheaf
theory and its cohomology. We then have the well-known
Theorem 1.2.2: Let {Uα } be an open cover on the smooth manifold M and let
G be a Lie group. Suppose further that on each nonempty intersection Uα ∩ Uβ
there exist smooth maps ψβα : Uα ∩ Uβ −−→G satisfying (1.2.2). Then there is a
principal G-bundle P (M, G) with transition functions ψβα . Moreover, there is a
bijective correspondence between isomorphism classes of principal G-bundles and
elements of H 1 (M, G).
The first statement in the theorem says that the transition functions determine
the principal bundle, and the last statement says that the set H 1 (M, G) classifies
principal G-bundles over M. The definition of an isomorphism of principal bundles
will be given below. It is only in the case that G is Abelian, for example, G = S 1 ,
the circle, that H 1 (M, G) has itself the structure of an Abelian group, and more
standard techniques such as the exponential sequence can be used to relate this to
the integral cohomology of M. We will discuss this later in more detail.
Perhaps the most important example of a principal bundle is:
Example 1.2.3: The linear frame bundle L(M ). Recall that a frame u =
(X1 , . . . , Xn ) at a point p ∈ M is a basis of the tangent space Tp M. We let L(M )
denote the set of all frames at all points of M. Then π : L(M )−−→M is a principal
bundle on M with group GL(n, R). The action of GL(n, R) on L(M ) is just given
by matrix multiplication from the right, that is (u, A) 7→ uA for u ∈ L(M ) and
A ∈ GL(n, R). The local triviality can be seen by choosing a local coordinate chart
(U ; x1 , . . . , xn ) on M and writing the vectors Xi of the frame u in local coordinates
as
X j ∂
Xi = Xi ,
j
∂xj

where the Xij are the components of a non-singular matrix (Xij ) of smooth functions
on U. Then the diffeomorphism ψ : π −1 (U )−−→U × GL(n, R) is given by ψ(v) =
(π(v), (Xij (v))). We remark that a point u ∈ L(M ) can be viewed as a linear map
16 1. STRUCTURES ON MANIFOLDS

u : Rn −−→Tp M with p = π(u) by u(ei ) = Xi , where {e1 , . . . , en } denotes the

standard basis of Rn .
Exercise 1.2: Let G be a Lie group and H a closed Lie subgroup. Show that the
Lie group G can be viewed as a principal H-bundle G(G/H, H) over the homoge-
neous manifold G/H.
Now suppose we are given a principal G-bundle P (M, G) over M , and a prin-
cipal H-bundle Q(N, H) on N. A homomorphism of principal bundles consists of a
smooth map fˆ : Q−−→P together with a Lie group homomorphism h : H−−→G such
that for all u ∈ Q and a ∈ H, fˆ satisfies fˆ(ua) = fˆ(u)h(a). This condition implies
that fˆ maps fibres to fibres; hence, there is a smooth map f : N −−→M such that
the diagram
fˆ
Q
 −−−−−→ P

 
(1.2.3)  
πQ πP
y y
f
N −−−−−→ M
commutes. If in addition fˆ is an embedding of smooth manifolds and h is a group
monomorphism then we say that fˆ : Q−−→P is an embedding of principal bundles.
This implies that the map f : N −−→M is also an embedding. In the case that
N = M and f is the identity map on M, then Q(M, H) is called a subbundle
or a reduction of P (M, G) to the group H. We also say that Q(M, H) is a reduced
subbundle of P (M, G). It may not be possible to reduce a principal bundle P (M, G)
to a given subgroup H. In general, there are topological obstructions to doing so. We
shall see many examples of this below. If also h is an isomorphism then Q(M, H)
and P (M, G) are isomorphic as principal bundles. We also say that P (M, G) is
trivial if it is isomorphic to the product M × G. Observe that the local triviality
conditions says that every point p ∈ M has a neighborhood U such that π −1 (U ) is
trivial. Concerning the transition functions we have
Proposition 1.2.4: A principal G-bundle P (M, G) can be reduced to a Lie sub-
group H if and only if there is an open cover {Uα } of M with transition functions
ψβα taking their values in H.
Proof. Suppose that P (M, G) can be reduced to a Lie subgroup H. Then there
is a principal H-bundle Q(M, H) together with a smooth bundle map f : Q−−→P
such that the diagram (1.2.3) commutes and f (ua) = f (u)a for all a ∈ H ⊂ G. On
−1
the open set πQ (Uα ) ⊂ Q(M, H) the maps φQ P Q P
α and φα are related by φα = φα ◦ f.
It follows that for any v ∈ P there exists an a ∈ G and u ∈ Q such that v = f (u)a
and φP Q P
α (v) = φα (u)a. Hence, the transition functions ψβα satisfy
P
(1.2.4) ψβα (v) = φP P
β (v)(φα (v))
−1
= φQ
β (u)a · a (φα (u))−1 = φQ
−1 Q Q
β (u)(φα (u))
−1
,
and thus, have their values in H.
P
Conversely, given transition functions ψβα : Uα ∩ Uβ −−→G which take their
P
values in the Lie subgroup H, a standard result says that ψβα is smooth as a
map into H. Thus, by Proposition 1.2.2 there is a principal H-bundle Q(M, H)
P
with transition functions ψβα . To construct the bundle map f we define maps
fα : πQ (Uα )−−→πP (Uα ) by putting fα = χ−1
−1 −1
P ◦ χQ . One easily sees that fα = fβ
 1.2. PRINCIPAL AND ASSOCIATED BUNDLES 17

on Uα ∩ Uβ , and so defines a global bundle map f : Q−−→P with the requisite

properties. ¤

Let P (M, G) be a principal bundle and F a G-manifold, that is, a manifold

together with a smooth action of the Lie group G. We denote this action by ‘left
multiplication’ as x 7→ a−1 x, where x ∈ F and a ∈ G. Then we have the “product
action” on P × F defined by (u, x) 7→ (ua, a−1 x). The quotient space (P × F )/G
by this action is a smooth fibre bundle E(M, F, G, P ) called the bundle associated
to P (M, G) with fibre F. Note that the projection map πE is defined as follows: let
[u, x] denote equivalence class of the pair (u, x) ∈ P × F, where (u, x) is equivalent
to (u0 , x0 ) if there is an a ∈ G such that (u0 , x0 ) = (ua, a−1 x). Then define the
projection map πE : E−−→M by πE ([u, x]) = πP (u). One easily checks that this
is well defined. In the case that F is a vector space, say Rk , and G = GL(k, R)
the associated bundle E = E(M, Rk , GL(k, R), P ) is called a vector bundle (of rank
k) associated to the principal bundle P (M, GL(k, R)), or just a real vector bundle.
Similarly, if the fibre F is Ck with group G = GL(k, C), then E is called a complex
vector bundle. We shall often combine these an refer to a real or complex vector
bundle as an F-bundle, where F = R or C. A vector bundle of rank one is called a
line bundle.
Exercise 1.3: Use πE to define a differential structure on E(M, F, G, P ) that
makes it a smooth fibre bundle with projection map πE and fibre F.
Exercise 1.4: Show that the tangent bundle T M of M is a vector bundle associ-
ated to the principal bundle L(M ). More generally show that the tensor bundles
Tsr M, exterior bundles Λp M, and symmetric bundles S p M are vector bundles as-
sociated to the principal bundle L(M ).
One often describes structures on manifolds by tensor fields. Examples are
complex structures, symplectic structures, Riemannian structures. These arise as
smooth sections of certain bundles over M which are associated to the frame bundle
L(M ) or its subbundles. Recall that a section of a fibre bundle πE : E−−→M
is a (smooth) map s : M −−→E such that πE ◦ s = 1lM . Unless otherwise stated
sections will be smooth. Vector bundles always have sections (e.g., the zero section);
whereas, a principal bundle has a section if and only if it is trivial. This is easy to
verify directly, but also follows from Theorem 1.2.5 below. By a local section of E
−1
we shall mean a section of the bundle πE (U ) for some open set U in M.
Theorem 1.2.5: A principal G-bundle P (M, G) is reducible to a closed subgroup H
if and only if the associated bundle E(M, G/H, G, P ) admits a section s : M −−→E.
Moreover, there is a bijective correspondence between such sections s and subbundles
Q(M, H) of P (M, G).

Proof. Suppose P (M, G) is reducible to a closed subgroup H and let f :

Q−−→P be the subbundle. The associated bundle E can be identified with the
quotient space P/H by mapping the equivalence class [u, gH] ∈ E to the equivalence
class [ug] ∈ P/H. Let ρ : P −−→E = P/H denote the natural projection. Then it is
easy to see that ρ ◦ f is constant on the fibres of Q. Thus, we can define a section
s : M −−→E by setting s(x) = ρ(f (u)), where πP (f (u)) = x.
Conversely, let s : M −−→E = P/H be a section, and consider P as a principal
H-bundle over E = P/H with transition functions hβα : Vα ∩Vβ −−→H, where {Vα } is
an open cover of E. Then we can define transition functions for P (M, G) with values
18 1. STRUCTURES ON MANIFOLDS

in H by setting ψβα = hβα ◦ s. Thus, by Proposition 1.2.2 P (M, G) is reducible to

the subgroup H. The correspondence between sections and H-subbundles can be
seen to be 1-1. ¤
Let P (M, G) be a principal bundle on a smooth manifold M with G = GL(k, R),
and let E be an associated real vector bundle of rank k. Reductions of P (M, G) to
subgroups G ⊂ GL(k, R) correspond to adding certain structures to its associated
vector bundle E. In general, there are obstructions to be able to do this. For exam-
ple, a reduction to the group GL(k, R)+ corresponds to choosing an orientation on
E, which can be done if and only if the bundle E is orientable. Such obstructions are
often given in terms of so-called characteristic classes [MS74]. Obstruction theory
for general fibre bundles is expounded in the classic text of Steenrod [Ste51] and
also in [Hus66]. E is orientable if and only if its first Stiefel-Whitney class w1 (E)
vanishes. See Appendix A. Many more examples of such obstructions are discussed
for G-structures in Section 1.4. Another example of such a reduction is the reduc-
tion of P (M, GL(k, R)) to its maximal compact subgroup O(k, R). Since GL(k, R)
is homotopy equivalent to the orthogonal group O(k, R) there are no obstructions
to performing this reduction. This corresponds to choosing a Riemannian metric
on E. If E is also orientable then we can reduce further to the special orthogonal
group SO(k, R). The corresponding principal bundle is now P (M, SO(k, R)).
Another type of structure comes from lifting instead of reduction. Thus, since
π1 (SO(k, R)) = Z2 the group SO(k, R) has a two-sheeted universal covering group
called Spin(k). So given a P (M, SO(k, R)) or one of its associated vector bundles E
one can ask whether the bundle P (M, SO(k, R)) can be lifted to a covering bundle
P (M, Spin(k)) with group Spin(k)? When such a covering exists the bundle E is
said to have a spin structure. We refer to [LM89] for a full treatment of spin
structures. The exact sequence of Lie groups 0−−→Z2 −−→Spin(k)−−→SO(k)−−→0
induces the coboundary map w2 : H 1 (M, SO(k))−−−→H 2 (M, Z2 ) whose image is
just the second Stiefel-Whitney class w2 (E) described briefly in Appendix A. So
an oriented vector bundle E admits a spin structure if and only if w2 (E) = 0.
Moreover, the distinct spin structures on E are in one-to-one correspondence with
the elements of H 1 (M, Z2 ). We say that a smooth oriented manifold M is a spin
manifold if T M admits a spin structure. So M is spin if and only if w1 (M ) =
w2 (M ) = 0. Spin manifolds admit certain vector bundles that do not exist on non-
spin manifolds, namely, those whose fibres are representations of Spin(n) that are
not representations of SO(n) or GL(n, R). Such vector bundles are called spinor
bundles and its sections are called spinor fields. Evaluation at a point gives a
spinor. Notice that spinor bundles depend on a choice of Riemannian metric. A
study of these representations involves Clifford algebras which will not be treated
in any detail in this text, and will not appear until Chapter 14. See [LM89] for a
thorough treatment.

1.3. Connections in Principal and Associated Vector Bundles

In this section we briefly review the fundamentals of the theory of connections.
As the reader undoubtedly knows there are various settings and definitions for a
connection. We begin with the most general as well as most abstract, namely that
of connections in a principal fibre bundle. We then discuss connections in associated
vector bundles, and the relationship to the former. Both formulations have their
advantages and will be used in the sequel. This brief discussion will also allow us to
 1.3. CONNECTIONS IN PRINCIPAL AND ASSOCIATED VECTOR BUNDLES 19

set our notation and terminology for the remainder of the text. Again the standard
reference for much of this material is the classic text of Kobayashi and Nomizu
[KN63, KN69].
Let P be a principal bundle over M with group G. Let u ∈ P and consider the
tangent space Tu P at u. Let Gu be the vertical subspace of Tu P consisting of all
vectors that are tangent to the fibre π −1 (π(u)). The differential of the restriction
of the map ψ to the fibre through u identifies Gu with the tangent space Te G at
the identity e ∈ G, and thus with the Lie algebra g of G.
Definition 1.3.1: A connection Γ in P is an assignment to each u ∈ P a com-
plimentary subspace Hu to Gu ⊂ Tu P , called the horizontal subspace, that for all
u ∈ P satisfies:
(i) Tu P = Gu ⊕ Hu .
(ii) Hua = (Ra )∗ Hu for all a ∈ G.
(iii) Hu depends smoothly on u.
In words a connection is a G-equivariant choice of compliment to the vertical
that varies smoothly with u. If X ∈ Tu P then vX and hX denotes the vertical
and horizontal components of X, respectively. We can describe a connection in
“fancier”terminology
F as follows. We define the vertical
F subbundle of T P by VP =
u∈P G u , and the horizontal subbundle by HP = u∈P Hu . Then we have an exact
sequence of G-modules
0 −→ VP −→ T P −→ Q −→ 0.
Then a connection in P is a splitting of this exact sequence as G-modules, and thus
gives a decomposition of the tangent bundle T P as G-modules, viz.
T P = VP ⊕ HP.
For a Lie group G acting smoothly on a manifold M there is well-known homo-
morphism from the Lie algebra g of G to the Lie algebra X (M ) of smooth vector
fields on M, and if G acts effectively (which we assume) this is a monomorphism.
Given A ∈ g we let A∗ denote its image in X (M ). If we specialize to the case of the
right action of G on a principal G-bundle P , the vector field A∗ is tangent to the
fibres π −1 (π(u)) at each point u ∈ P , and is called the fundamental vertical vector
field on P associated to A ∈ g. Evaluation of A∗ at a point u ∈ P gives a vector
space isomorphism of the Lie algebra g with the vertical tangent space Gu at u.
Definition 1.3.2: Given a connection Γ on P , we define the connection form
associated to Γ to be the 1-form ω on P with values in the Lie algebra g by setting
ω(X) equal to the unique A ∈ g such that (A∗ )u is the vertical component of X.
Clearly, ω satisfies ω(X) = 0 if X is horizontal. In the sequel we are particularly
interested in the case of a circle bundle. In this case the Lie algebra g ≈ R, and
it is common to take 1 as a generator of g. Actually a connection 1-form defines a
connection as is seen by the following
Proposition 1.3.3: The connection form ω satisfies
(i) ω(A∗ ) = A for all A ∈ g.
(ii) Ra∗ ω = ada−1 ω for all A ∈ g.
Conversely, given a g valued 1-form ω on P which satisfies conditions (i) and (ii),
there is a unique connection Γ in P whose connection form is ω.
20 1. STRUCTURES ON MANIFOLDS

We shall often refer to a connection 1-form ω as simply a connection. By

localizing and using partitions of unity one has
Theorem 1.3.4: Every principal bundle P (M, G) admits a connection.
Let P (M, G) be a principal bundle and V be a finite dimensional vector space
over F = R, C or the quaternions1 Q. Let Ωr (P, V ) denote the set of smooth r-forms
on P with values in V, that is the set of smooth sections of the bundle Λr (P )⊗V. This
set has some natural algebraic structures. It is an infinite dimensional vector space
over F as well as a C ∞ (P ) ⊗R F module. Consider a representation ρ : G → Aut V
on V . A pseudotensorial form of degree r and of type (ρ, V ) is r-form on P with
values in V such that
Ra∗ φ = ρ(a−1 ) · φ.
φ is called tensorial if φ is pseudotensorial and satisfies φ(X1 , . . . , Xr ) = 0 if at
least one of the vector fields X1 , . . . , Xr is vertical. In particular, a pseudotenso-
rial 0-form, which is automatically tensorial, is just a smooth G-equivariant map
from P to V. We denote by T r (P, V ) the closed subspace of tensorial r-forms. Now
suppose that Γ is a connection in P (M, G). Then the connection form ω of Γ is
pseudotensorial of type (ad, g), but not tensorial. Notice, however, Proposition
1.3.3 implies that the difference ω 0 − ω of any two connections is tensorial. Given
any pseudotensorial form φ we can define a tensorial form by taking its horizontal
projection, i.e., φh(X1 , . . . , Xr ) = φ(hX1 , . . . , hXr ). Generally dφ is only pseu-
dotensorial, even if φ is tensorial. However, the exterior covariant derivative Dφ
defined as the horizontal projection
(1.3.1) Dφ(X1 , . . . , Xr+1 ) = dφ(hX1 , . . . , hXr+1 )
is always tensorial. In particular, if ω is a connection 1-form, Dω is a tensorial 2-
form of type (ad, g) called the curvature 2-form of ω and usually denoted by Ω. It is
a smooth section of Λ2 (P ) ⊗ g, and satisfies the famous Cartan structure equations
(1.3.2) Ω = Dω = dω + 21 [ω, ω],
as well as the well-known Bianchi identities
(1.3.3) DΩ = 0.
The meaning of the bracket in the Equation (1.3.2) is [ω, ω](X, Y ) = [ω(X), ω(Y )].
Generally, the exterior covariant derivative of any tensorial form is
Dφ = dφ + [ω, φ].
A connection Γ or ω is said to be flat if its curvature Ω vanishes.
Tensorial forms can be described alternatively in terms of associated bundles.
Let E = P ×G V be the F-vector bundle on M with standard fibre V and associated
to the F-representation ρ of G. Then there is a 1-1 correspondence between tensorial
r-forms T r (P, V ) of type (ρ, V ) and smooth sections of the bundle Λr (M ) ⊗ E as
follows: By fixing a point u ∈ P the natural projection P × V −−→E gives an F-
linear isomorphism from the vector space V to the fibre Eπ(u) . Then if X1 , . . . Xr
are tangent vectors at π(u) ∈ M we define the section φ̃ of Λr (M ) ⊗ E by
φ̃(X1 , . . . , Xr ) = uφ(X1∗ , . . . , Xr∗ ),
1We shall need all three types of vector spaces in the sequel. Some care must be taken
when working with quaternionic vector spaces, due to the noncommutativity of Q. For example,
quaternionic vector spaces are considered by multiplication from the left.
 1.3. CONNECTIONS IN PRINCIPAL AND ASSOCIATED VECTOR BUNDLES 21

where X ∗ denotes any vector on P that projects to X on M. One checks that φ̃

is independent of the choices made. In particular, smooth G-equivariant functions
P −−→V correspond to smooth sections of E.
Now a connection Γ in P induces a connection ∇ in the vector bundle E and,
more generally, in the vector bundles Λr (M ) ⊗ E. We denote the C ∞ (M )-module
of smooth sections of Λr (M ) ⊗ E by Ωr (M, E).
Definition 1.3.5: A (Koszul) connection ∇ on an F-vector bundle E is an F-
linear map on sections: ∇ : Γ(E) → Λ1 (M, E) satisfying the Leibnitz rule ∇(f s) =
f ∇s + df ⊗ s, where f ∈ C ∞ (M ) and s is a smooth section of E.
Given a connection Γ on P (M, G) and an associated vector bundle E = P ×G V,
we can define a connection ∇ on E by setting
∇ = uDu−1 .
Conversely, suppose we have an F-vector bundle E with associated principal bundle
P (M, G) and F-representation ρ of G on the standard fibre V so that E = P ×G V.
Then given a connection ∇ on E, we obtain a connection Γ on P as follows: Let
s = (s1 , . . . , sk ) be a G-frame of local sections of E. If X ∈ Tx M, then the subspace
Hu of Pu with π(u) = x defined by
Hu = {s∗ X | ∇X s = 0}
is a G-equivariant complement to the vertical subspace Gu . This defines the con-
nection Γ in P. The Koszul connection ∇ in the associated bundles is often referred
to as the covariant derivative, and a section s of E is said to be covariantly constant
if ∇s = 0.
As in Proposition 1.3.3 it is convenient to express the connection ∇ in terms of
a 1-form. However, this can only be done locally in terms of a local trivialization
of the vector bundle E. Given a local trivialization E|U ' U × V, we can write
(1.3.4) ∇|U = d + AU ,
where the exterior derivative d represents the flat connection on U and AU is a g-
valued 1-form on U. Given another such trivialization on the open set W ⊂ M, and
transition functions g : U ∩ W −−→G relating the two trivializations in the overlap,
one obtains the relation
(1.3.5) AW = g −1 AU g + g −1 dg.
Conversely, given a cover of M by open sets {Uα } with g-valued 1-forms AU on each
open set in the cover satisfying Equation (1.3.5) in the overlaps, one can reconstruct
the connection ∇ on E. This formulation is essentially Cartan’s definition of con-
nection, whereas, the formulation given in Proposition 1.3.3 is due to his student
Ehresmann.
There is a natural extension of Equation (1.3.4) to the bundles Λr (M ) ⊗ E. We
shall use the notation in [DK90] and write the exterior covariant derivative as
dA = d + A : Ωr (M, E)−−→Ωr+1 (M, E).
Here the symbol A denotes a family of g-valued 1-forms on open sets of M satisfying
Equation (1.3.5) which act linearly via the representation ρ∗ : g−−→End V on the
local sections of E obtained from a local trivialization of E. The curvature form
22 1. STRUCTURES ON MANIFOLDS

Ω = Dω on P corresponds by the isomorphism u to the smooth section F A = dA A

of Λ2 (M ) ⊗ g. The Cartan structure Equations (1.3.2) then take the form

(1.3.6) F A = dA A = dA + A ∧ A,

where here again we follow the convention in [DK90] using the wedge product in
lieu of the brackets to emphasize A as an endomorphism of V via local trivializa-
tions.
Exercise
P 1.5: Show that if we write P the g-valued connection 1-form as A =
Ai dxi and its curvature 2-form F A = Fij dxi ∧ dxj in a local coordinate chart
(U ; x), Equation (1.3.6) can be written as

∂Aj ∂Ai
Fij = − + [Ai , Aj ].
∂xi ∂xj

A connection Γ in P (M, G) allows us to define the notion of parallel translation

of a fibre of P along any piecewise smooth curve in M . This is done as follows: Let
x0 , x1 ∈ M and let γ : [0, 1]−−→M be a curve in M with γ(0) = x0 and γ(1) = x1 .
Now at each point u ∈ P there is a vector space isomorphism ι : Tπ(u) M −−→Hu .
So fixing u0 ∈ π −1 (x0 ) we can lift γ horizontally to a unique piecewise smooth
curve γ ∗ in P such that γ ∗ (0) = u0 and γ̇ ∗ (t) = ι(γ̇(t)) for all t ∈ [0, 1]. This gives
an isomorphism of fibres π −1 (γ(t)) ≈ π −1 (x0 ), called parallel translation along γ.
Now suppose that γ is a loop at x0 , then parallel translation gives automorphisms
of the fibre π −1 (x0 ). By composing loops and running the loop backwards, we see
that the set of such automorphisms form a group. Moreover, if we fix a point
u ∈ π −1 (x0 ) this group can be identified with a connected Lie subgroup of G,
called the holonomy group through u and denoted by Holu . If we restrict ourselves
to loops at π(u) that are null homotopic, then we obtain a normal subgroup Hol0u
of Holu known as the restricted holonomy group through u. The groups Hol0u and
Holu enjoy the following properties:
(i) If u, v ∈ P can be joined by a horizontal curve then Holu = Holv and
Hol0u = Hol0v ,
(ii) If v = ua for a ∈ G, then Holu = Ada Holv and H0 (u) = Ada Hol0u .
A fundamental result in the theory of holonomy groups is the so-called Reduc-
tion Theorem which we now state. Of course, we refer to [KN63] for its proof:
Theorem 1.3.6: Let M be a smooth connected manifold and P (M, G) be a prin-
cipal G-bundle with a connection Γ. Let u ∈ P be an arbitrary point and let P (u)
denote the subset of points in P that can be joined to u by a horizontal curve. Then
P (u) is a reduced subbundle of P with structure group Holu , and the connection Γ
restricts to a connection on P (u).
The subbundle P (u) is called the holonomy bundle through u, and we call such
a connection reducible. It has its values in the Lie algebra holu of Holu . A well-
known theorem of Ambrose and Singer [AS53] characterizes the Lie algebra holu
as precisely the Lie algebra spanned by the curvature Ωv (X, Y ), where v ∈ P (u)
and X, Y are arbitrary horizontal vectors at v. Conversely, if Q(M, H) ⊂ P (M G)
is principal subbundle corresponding to a Lie subgroup H ⊂ G, a connection Γ in
a principal bundle Q(M, H) can be extended to a connection in P (M, G).
 1.4. G-STRUCTURES 23

1.4. G-Structures
In this section we describe structures on manifolds from the unifying viewpoint
of G-structures. There are several texts where this point of view is expounded
[Kob72, Mol77, Sal89, Ste83]. Here we consider only first order G-structures,
that is, reductions of the bundle of linear frames on M. Here is the precise definition.
Definition 1.4.1: Let G ⊂ GL(n, R) be a subgroup, then a G-structure on M is
a reduction of the frame bundle L(M ) to the subgroup G. The G-structure is said
to be integrable if every point of M has a local coordinate chart (U ; x) such that
the local section
³ ∂ ∂ ´
, . . . ,
∂x1 ∂xn
of L(M ) is a local section of the reduced bundle P (M, G). Such a coordinate chart
is called admissible.
Let us return to the bundle of linear frames L(M ) of a smooth manifold M.
On L(M ) there is a canonical Rn -valued 1-form θ defined as follows: as seen in
Example 1.2.3 we can view any point u ∈ L(M ) as a vector space isomorphism
u : R−−→Tπ(u) M. So θ can be defined by

(1.4.1) hθ, Xi = u−1 π∗ X,

where X ∈ Tu L(M ) and h·, ·i denotes the natural pairing between the tangent bun-
dle to the bundle of linear frames T L(M ) and its dual cotangent bundle T ∗ L(M ).
If P (M, G) is a G-structure on M, i.e., a subbundle of L(M ), then we can restrict θ
to P (M, G) giving a canonical 1-form on the G-structure. The canonical 1-form on
a G-structure P (M, G) behaves functorially under the action of the general linear
group GL(n, R) on L(M ). Indeed we have
Lemma 1.4.2: For any a ∈ G ⊂ GL(n, R) the transformation rule holds:
Ra∗ θ = a−1 θ.

Proof. For any vector field X on P (M, G), we have

hRa∗ θ, Xi = hθ, Ra∗ Xi = (ua)−1 π∗ Ra∗ X = a−1 u−1 π∗ X = ha−1 θ, Xi. ¤

Generally, there are topological obstructions to the existence of G-structures.

To see that such a reduction does not always exist let G = {e} the identity group.
Then an {e}-structure on M is nothing but a global frame or parallelism of M. But
it is well-known that global frames do not always exist, that is that M may not
be parallelizable. For example, the 2-sphere S 2 does not even have one nowhere
vanishing vector field let alone a global frame. Even if there is a G-structure
on M there may not be an integrable G-structure. For example, as seen below
every manifold M admits many O(n, R)-structures, but a compact M admits an
integrable O(n, R)-structure only if M is covered by a torus (see Example 1.4.7
below). On the other hand every GL(n, R) structure on M is integrable. The
following proposition is evident.
Proposition 1.4.3: A G-structure P (M, G) is integrable if and only if there is an
³ (α) ´
∂xi
atlas of coordinate charts {(Uα ; x(α) )}α∈I on M whose Jacobian matrices (β)
∂xj i,j
lie in G at all points of Uα ∩ Uβ .
24 1. STRUCTURES ON MANIFOLDS

Recall from Section 1.3 that associated to any connection Γ in a principal bundle
there is the fundamental curvature 2-form. So for any connection Γ in L(M ), usually
called a linear connection, we have its curvature Ω = Dω. However, since the linear
frame bundle L(M ) has a canonical 1-from θ associated to it, we have another
2-form associated to any linear connection Γ, namely Θ = Dθ, called the torsion
2-form. In the case of linear connections we can add to Cartan’s structure Equation
(1.3.2) the so-called First Structure Equation
(1.4.2) dθ + ω ∧ θ = Θ.
We can also add to the Bianchi identities (1.3.3), the First Bianchi identities
(1.4.3) DΘ = Ω ∧ θ.
Exercise 1.6: Show that for linear connections the usual expressions for the torsion
and curvature tensors
T (X, Y ) = ∇X Y − ∇Y X − [X, Y ], R(X, Y )Z = ∇X ∇Y Z − ∇Y ∇X Z − ∇[X,Y ] Z
are related to the corresponding Cartan expressions by
T (X, Y ) = u(2Θ(X ∗ , Y ∗ )), R(X, Y )Z = u(2Ω(X ∗ , Y ∗ ))(u−1 Z),
where X ∗ , Y ∗ are the horizontal lifts of the vectors X, Y ∈ Tπ(u) M , respectively,
and u ∈ L(M ) is any point.
This entire discussion holds for any G-structure P (M, G) ⊂ L(M ) with a con-
nection Γ. We shall often refer to a linear connection Γ in a G-structure P (M, G) as
a G-connection. A linear connection Γ with Θ = 0 is said to be torsion-free. Clearly
the notions of parallel translation and holonomy apply to the case of G-structures.
We now want to put Theorem 1.2.5 to work by seeing how certain natural
tensor fields define G-structures on M. Suppose that T0 is an element of the tensor
algebra T (Rn ) over Rn and that G is the largest closed Lie subgroup of GL(n, R)
that leaves T0 invariant. Viewing a point u ∈ L(M ) as a vector space isomorphism
u : Rn −−→Tπ(u) M, we obtain an induced isomorphism u∗ : T (Rn )−−→T (Tπ(u) M ) of
the tensor algebras. Then the image T = u∗ T0 is a section of the tensor algebra
bundle T (M ). If T0 is a tensor of type (r, s) then T is a section of the tensor bundle
Tsr M. In any case because of the invariance of T0 under G, the tensor field T defines
a section of the associated bundle L(M )/G. In this case we say that P (M, G) is a
G-structure defined by the tensor T0 .
Proposition 1.4.4: Let P (M, G) be a G-structure defined by the tensor T0 . Then
P (M, G) is integrable if and only if there exists an atlas of charts {(Uα ; x(α) )}α∈I
on M such that the corresponding tensor field T = u∗ T0 has constant components
on U.
Proof. (⇒) Let P (M, G) be integrable and let (U ; x) be a coordinate chart,
∂
then the frame u = ( ∂x 1
, . . . , ∂x∂n ) belongs to P (M, G). Let {ei } denote the standard
n j ∂
basis for R and {e } its dual basis. Then ∂x i
= u(ei ), and dxj = u(ej ). So if
T0 = tji11···i j1
···js ei1 ⊗ · · · ⊗ eir ⊗ e ⊗ · · · ⊗ e
r js

is a tensor of type (r, s), we have

∂ ∂
T = u∗ T0 = tji11···i
⊗ ··· ⊗
···js
r
⊗ dxj1 ⊗ · · · ⊗ dxjs .
∂xi1 ∂xir
Hence, T has constant components on U.
 1.4. G-STRUCTURES 25

(⇐) Conversely, suppose that (U ; x) is a coordinate chart such that T = u∗ T0

has constant components. So T is a constant section of the associated bundle
L(U )/G. Then there is a linear transformation of coordinates on U to new coor-
dinates (y1 , . . . , yn ) such that the frame u = ( ∂y∂ 1 , . . . , ∂y∂n ) belongs to P (M, G).
Thus, the G-structure P (M, G) is integrable. ¤
One easily sees that if a G-structure P (M, G) is defined by a tensor field T and Γ
is a G-connection, then T is covariantly constant with respect to Γ, or equivalently,
∇T = 0. Generally, the ‘integrability condition’ of Definition 1.4.1 can be very
restrictive, and we will discuss various ‘levels of integrability’. First, it is easy to
show that
Proposition 1.4.5: An integrable G structure admits a torsion-free connection.
It is now convenient to consider a much less restrictive definition of integrability.
Definition 1.4.6: A G-structure P (M, G) is said to be 1-integrable if it admits
a torsion-free G-connection.
So an integrable G-structure is automatically 1-integrable. The failure of the
existence of torsion-free G-connection can be seen as the first order obstruction to
integrability, hence, the name 1-integrable. The question of uniqueness of a torsion-
free connection, assuming one exists, is related to prolongations of G-structures
which we shall briefly treat below (See Definition 1.6.8). We now wish to consider
many examples of G-structures. Our first example is a good example where non-
integrable G-structures are of more interest than the integrable ones.
Example 1.4.7: Riemannian metrics. We consider a reduction of the frame
bundle L(M ) to the orthogonal group
O(n, R) = {A ∈ GL(n, R) | At A = 1ln } .
The reduced bundle O(M ) ⊂ L(M ) is called the orthonormal frame bundle of M.
Theorem 1.2.5 says that such a reduction is equivalent to a choice of section of the
associated bundle E(M, GL(n, R)/O(n, R), L(M )) = L(M )/O(n, R). We show that
such a section is just a Riemannian metric on M. As mentioned previously each
point u ∈ L(M ) gives an isomorphism of the standard vector space Rn with the
tangent space Tπ(u) M. Let h·, ·i denote the Euclidean inner product on Rn , and let
X, Y ∈ Tπ(u) M. Then
gu (X, Y ) = hu−1 X, u−1 Y i
defines an inner product on Tπ(u) M. Furthermore, if a ∈ O(n, R) we have
gua (X, Y ) = h(ua)−1 X, (ua)−1 Y i = ha−1 u−1 X, a−1 u−1 Y i
= hu−1 X, u−1 Y i = gu (X, Y ),
where the second to the last equality holds by the invariance of h·, ·i under O(n, R).
This shows that gu is constant along the fibres of O(M ), and thus, is a section of
L(M )/O(n, R). Thus, the associated bundle L(M )/O(n, R) can be identified with
a subbundle of the vector bundle Sym2 T ∗ M of symmetric covariant 2-tensors on
M. The choice of u modulo O(n, R) corresponds to a choice of Riemannian metric
on M. Since manifolds are paracompact the standard partition of unity argument
shows that such sections and hence, such reductions always exist.
When are O(n)-structures integrable? According to Proposition 1.4.4 this oc-
curs when the metric tensor g has constant components in some coordinate chart
26 1. STRUCTURES ON MANIFOLDS

(U ; x), and then by a change of coordinate, say (V ; y) the metric can be brought
to the form X
g= (dyi )2
i
in V. By the Second Fundamental Theorem of Riemannian geometry this happens
precisely when the Riemann curvature tensor vanishes. For example, if M is a com-
pact 2-dimensional manifold, then it is either a torus or a Klein bottle depending
on whether it is orientable or not. This is quite restrictive. On the other hand
the First Fundamental Theorem of Riemannian geometry says that there exists
a unique torsion-free Riemannian connection, denoted ∇g ≡ ∇, called the Levi-
Civita connection. So all Riemannian G-structures are 1-integrable. Similarly, a
reduction to the group SO(n, R) = O(n, R) ∩ GL+ (n, R) corresponds to oriented
Riemannian geometry. In particular, one can consider the parallel translation de-
fined by the Levi-Civita connection and its associated holonomy group which is a
subgroup of the structure group O(n, R) (SO(n, R) in the oriented case). Since this
connection ∇g is uniquely associated to the metric g, we denote it by Hol(g), and
refer to it as the Riemannian holonomy group or just the holonomy group when
the context is clear. Indeed, it is precisely this Riemannian holonomy that plays
an important role in this book. Now on a Riemannian manifold (M, g) there is a
canonical epimorphism π1 (M )−−−→Hol(g)/Hol0 (g), in particular, if π1 (M ) = 0 then
Hol(g) = Hol0 (g). In 1955 Berger proved the following theorem [Ber55] concerning
Riemannian holonomy:
Theorem 1.4.8: Let (M, g) be an oriented Riemannian manifold which is neither
locally a Riemannian product nor locally symmetric. Then the restricted holonomy
group Hol0 (g) is one of the following groups listed in Table 1.4.1.

Table 1.4.1: Berger’s Riemannian Holonomy Groups

Hol0 (g) dim(M ) Geometry of M Comments
SO(n) n orientable Riemannian generic Riemannian
U (n) 2n Kähler generic Kähler
SU (n) 2n Calabi-Yau Ricci-flat Kähler
Sp(n) · Sp(1) 4n quaternionic Kähler Einstein
Sp(n) 4n hyperkähler Ricci-flat
G2 7 G2 -manifold Ricci-flat
Spin(7) 8 Spin(7)-manifold Ricci-flat

We will encounter all the geometries listed in this table throughout this book.
Most of them will already be introduced in this chapter as G-structures. Orig-
inally, Berger’s list included Spin(9) but Alekseevsky proved that any manifold
with such holonomy group must be symmetric [Ale68]. In the same paper Berger
also claimed a classification of all holonomy groups of torsion-free affine (linear)
connections that act irreducibly. He produced a list of possible holonomy rep-
resentations up to what he claimed was a finite number of exceptions. But his
classification had some gaps discovered 35 years later by Bryant [Bry91]. An infi-
nite series of exotic holonomies was found in [CMS96] and finally the classification
in the non-Riemannian affine case was completed by Merkulov and Schwachhöfer
[MS99]. We refer the reader to [MS99] for the proof, references and the history of
the general affine case. In the Riemannian case a new geometric proof of Berger’s
 1.4. G-STRUCTURES 27

Theorem is now available [Olm05]. An excellent review of the subject just prior
to the Merkulov and Schwachhöfer’s classification can be found in [Bry96]. We
should add that one of the first non-trivial results concerning manifolds with the
exceptional holonomy groups of the last two rows of Table 1.4.1 is due to Bonan
[Bon66] who established Ricci-flatness of manifolds with parallel spinors.
It is clear in the Riemannian case that non-integrable structures are much
more interesting than integrable ones, and there is a big gap between the integrable
structures and the 1-integrable structures. The obstructions to integrability occur
at order two, namely the Riemannian curvature. In order to study these further one
needs to study certain invariants of the G-structure which involves a detailed study
of the Riemannian curvature. For a general G-structure these invariants involve
Spencer cohomology to which we refer the reader to the literature [Fuj72, KS72,
SS65, Ste83]. In the case of O(n, R)-structures the relevant Spencer cohomology
group consists of the Riemann curvature tensor. In our next example the integrable
case is very rich.
Example 1.4.9: Almost complex structures. A complex structure on a real
vector space V is an endomorphism J of V that satisfies J 2 = −1l, where 1l denotes
the identity endomorphism of V. We can give V the structure of a complex vector
space by defining scalar multiplication with scalars in C by
(a + ib)v = av + bJv.
Conversely, if V is a complex vector space we define the endomorphism J on V by
Jv = iv.
Let {v1 , . . . , vn } be basis for V as a complex vector space, then it is easy to check
that {v1 , . . . , vn , Jv1 , . . . , Jvn } is a basis for V as a real vector space. In particular,
this implies that any real vector space with a complex structure has even real
dimension. For example, by writing zj = xj + iyj for j = 1, . . . , n, the complex
vector (z1 , . . . , zn ) in the complex vector space Cn is identified with the real vector
(x1 , . . . , xn , y1 , . . . , yn ) in the real vector space R2n . The canonical complex structure
J0 on V is given in block form with respect to the standard basis of R2n by
µ ¶
0 1ln
J0 = ,
−1ln 0
where 1ln denotes the n × n identity matrix. We can identify the complex linear
group GL(n, C) with the subgroup of matrices in GL(2n, R) that commute with J0 .
Thus, the complex n × n matrix A + iB is identified with the real 2n × 2n matrix
µ ¶
A B
.
−B A
On a manifold M an almost complex structure is an endomorphism Jx of the
tangent space Tx M that varies smoothly with x and satisfies Jx2 = −1lx at each
point x ∈ M, or in other words, a smooth section J of the endomorphism bundle
End T M satisfying J 2 = −1l. As in the Riemannian case such a section can be
obtained from the canonical complex structure J0 on R2n by identifying u ∈ L(M )
with a linear map u : R2n −−→Tπ(u) M. Thus, we define

Ju = uJ0 u−1 .
28 1. STRUCTURES ON MANIFOLDS

The invariance of J0 under the subgroup GL(n, C) ⊂ GL(2n, R) shows that Ju is

constant along the fibres of the bundle CL(M ) of complex linear frames on M and
thus, gives a section of the associated bundle E(M, GL(2n, R)/GL(n, C), L(M )) =
L(M )/GL(n, C) which is the subbundle of End T M consisting of endomorphisms
that satisfy J 2 = −1l. Unlike in the Riemannian case it is not always possible to find
such a section, or equivalently to obtain a reduction of the real frame bundle L(M )
to the complex linear group GL(n, C). In general, this is a fairly deep topological
question. The existence of an almost complex structure J on M is equivalent to
the splitting of the complexified tangent bundle T M ⊗ C = T 1,0 M ⊕ T 0,1 M. If
one has such a splitting, J can be defined to be multiplication by i on T 1,0 M
and multiplication by −1 on T 0,1 M. Conversely, given J one easily sees that it is
diagnonalizable over T M ⊗ C with eigenvalues ±i.
By Proposition 1.4.3 a GL(n, C)-structure is integrable when there is an atlas
(α) (α) (α) (α)
of coordinate charts {Uα ; x1 , . . . , xn , y1 , . . . , yn }α∈I whose Jacobian matrix
(α)
lies in GL(n, C) ⊂ GL(2n, R). This corresponds to the complex coordinates zj =
(α) (α) (β)
xj +iyj being independent of the complex conjugate coordinates z̄i , that is the
transition functions are holomorphic or equivalently satisfy the Cauchy-Riemann
equations. An integrable almost complex structure is called a complex structure,
and a smooth manifold M with a complex structure is called a complex manifold.
The structure tensor of the G-structure that measures the failure of integrability is
the Nijenhuis torsion tensor defined by
(1.4.4) NJ (X, Y ) = [JX, JY ] − [X, Y ] − J[X, JY ] − J[JX, Y ].
The vanishing of this tensor field is equivalent to the integrability of the almost com-
plex structure. If one assumes that M is real analytic this follows from a Frobenius
integrability argument (cf. [KN69], Appendix 8). However, in the C ∞ case this is
a much deeper result due to Newlander and Nirenberg [NN57]. Summarizing an
almost complex structure is integrable if and only if the Nijenhuis torsion tensor
NJ vanishes. This is indeed the torsion of an almost complex connection, so the
only obstruction to integrability occurs at level one, that is, in this case 1-integrable
implies integrable.
Following up on the remark at the end of the second paragraph of Example
1.4.9, one can ask why a partition of unity argument doesn’t work in the case of
almost complex structures. The point is that one cannot guarantee the condition
J 2 = −1l at every point of the manifold by using partitions of unity. A better under-
standing is obtained from the point of view of G-structures. An alternative proof
of the existence of Riemannian metrics on any manifold is as follows: by Theorem
1.2.5 giving a Riemannian metric is equivalent to giving a reduction of the frame
bundle to the orthogonal group O(n, R). This reduction can be accomplished (gen-
erally over any paracompact space) since as topological spaces the general linear
n(n+1)
group GL(n, R) decomposes as O(n, R) × R 2 , the so-called Iwasawa decompo-
sition. Thus, GL(n, R) is homotopy equivalent to O(n, R). Then paracompactness
allows one to construct this homotopy equivalence at each point of M giving the
reduction. However, in the case of reducing L(M ) to the complex linear group
GL(n, C) there is no such homotopy equivalence. Indeed, GL(2n, R) is homotopy
equivalent to O(2n, R), whereas, again the Iwasawa decomposition for GL(n, C)
says that GL(n, C) is homotopy equivalent to the unitary group U (n), that is, the
subgroup of GL(n, C) that leaves invariant the standard Hermitian form h·, ·i on
 1.4. G-STRUCTURES 29

Cn . It is well-known that O(2n, R) and U (n) are not homotopy equivalent, e.g.
they have different homology groups. So, generally, there are obstructions to such
reductions. A general theory of obstructions is expounded in the classic text of
Steenrod [Ste51]. A well-known example of an even dimensional manifold which
admits no almost complex structure is the 4-sphere S 4 . This can be shown either by
obstruction theory or using the Atiyah-Singer index theorem. Another interesting
example is S 6 which is known to admit an almost complex structure, but it is still
an open question whether S 6 has an integrable almost complex structure, i.e., a
complex structure.
Exercise 1.7: Let G = GL+ (n, R) be the subgroup of matrices in GL(n, R) whose
determinant is positive. Show that M admits such a G-structure if and only if M
is orientable. Show that a GL+ (n, R)-structure on M coincides with a choice of
orientation of M. What about integrability?
Exercise 1.8: Show that an almost complex structure on M determines an orien-
tation on M, hence, any almost complex manifold is orientable.
Notice that an oriented Riemannian manifold gives a reduction of L(M ) to the
special orthogonal group SO(n) = {A ∈ O(n) | det(A) = 1}, and that SO(n) =
GL+ (n, R) ∩ O(n). We will frequently use G-structures that occur as a combination
of other G-structures or, alternatively, a G-structure defined by more than one
tensor. Another important example of this is:
Example 1.4.10: Almost Hermitian structures. Let G = U (n) the unitary
group defined by
U (n) = {A ∈ GL(n, C) | Āt A = 1ln }.
This is precisely the group that leaves the standard Hermitian form hu, vi = u¯1 v1 +
· · · + u¯n vn in Cn invariant. Since U (n) = GL(n, C) ∩ O(2n), a U (n)-structure or
almost Hermitian structure consists of an almost complex structure J together with
a Riemannian metric g on M satisfying the compatibility condition g(JX, JY ) =
g(X, Y ). The two tensor fields g and J give rise to a 2-form ω called the Kähler
form defined by ω(X, Y ) = g(X, JY ). The structure is called almost Kähler if ω is
closed, Hermitian if J is integrable, and it is called Kähler if both J is integrable
and ω is closed. A U (n)-structure is integrable if both the almost complex structure
and the Riemannian structures are integrable which is very restrictive. The only
compact examples are covered by a complex torus. A less restrictive notion is that
of being 1-integrable. A U (n)-structure is 1-integrable if and only if it is Kähler.
This is equivalent to the U (n) bundle coinciding with the holonomy bundle. We
shall discuss these structures in much more detail in Chapter 3.
These ideas can also be applied to indefinite or pseudo-Riemannian and pseudo-
Hermitian metrics2. For example,
Example 1.4.11: Pseudo-Riemannian structures. Let O(p, q) denote the sub-
group of GL(n, R) that leaves the quadratic form u21 + · · · + u2p − u2p+1 − · · · − u2p+q
invariant. Then an O(p, q)-structure on M is the same as a choice of pseudo-
Riemannian metric g on M of signature (p, q). If p or q equals 1, then this is called
a Lorentzian structure and g a Lorentzian metric. Similarly, one has almost pseudo-
Hermitian structures by considering the pseudo-unitary groups U (p, q) which is the

2One should be cautioned that the terminology pseudo-Hermitian structure is used in an

entirely different sense in [Web78]
30 1. STRUCTURES ON MANIFOLDS

group that leaves invariant an indefinite Hermitian form of signature (p, q). Clearly,
U (p, q) = GL(p + q, C) ∩ O(2p, 2q). The integrability question is similar to that of
Example 1.4.7.
Unlike Riemannian metrics (structures), pseudo-Riemannian metrics do not
always exist on M. See for example [HE73, O’N83] for the existence of Lorentzian
metrics on M. In fact, it is easy to see that M admits a Lorentzian metric if and
only if it admits a nowhere vanishing vector field.
Exercise 1.9: Define the group
CO(n) = {A ∈ GL(n, R) | At A = c1ln , c > 0} .
Show that a CO(n)-structure on M coincides with a conformal class of Riemannian
metrics on M. Recall that a conformal class is an equivalence class of Riemannian
metrics, where the Riemannian metrics g and g 0 are equivalent if there exists a
positive function λ on M such that g 0 = λg.
A CO(n)-structure on M is called a conformal structure on M ; however, CO(n)
is usually not called the conformal group, a name that is usually reserved for the
automorphism group of a conformal structure, and this group is generally larger
than CO(n). This phenomenon is related to prolongations of G-structures which is
briefly touched upon below. The reader is referred to [Kob72, Ste83] for complete
discussions of prolongations of G-structures. Conformal structures can be defined
similarly in the case of pseudo-Riemannian structures. Our next example will also
be of fundamental importance to us.
Example 1.4.12: Almost symplectic structures. Consider the non-degenerate
antisymmetric bilinear form
ω0 (u, v) = u1 vn+1 + · · · + un v2n − un+1 v1 − · · · − u2n vn
on the vector space R2n . Notice that ω0 can also be written in terms of the standard
Euclidean metric h·, ·i and canonical complex structure J0 as
ω0 (u, v) = hu, J0 vi.
The subgroup of GL(2n, R) that leaves ω0 invariant is called the real symplectic
group and denoted3 by Sp(n, R). It is easy to see that this group can be defined by
Sp(n, R) = {A ∈ GL(2n, R) | At J0 A = J0 }.
An almost symplectic structure on a manifold M is then given by transporting
the antisymmetric form ω0 to the manifold as before by the linear isomorphism
u : R2n −−→Tp M, viz. for X, Y ∈ Tp M we define
ω(X, Y ) = ω0 (u−1 X, u−1 Y ).
ω is a section of the bundle E(M, GL(2n, R)/Sp(n, R), L(M )) = L(M )/Sp(n, R)
which can be identified with the subbundle of the exterior bundle Λ2 M bundle con-
sisting of non-degenerate 2-forms. By Theorem 1.2.5 this corresponds to a reduction
of L(M ) to the real symplectic group Sp(n, R). The appearance of the tensor J0 sug-
gests that there may be a connection between almost complex structures and almost
symplectic structures. This is indeed the case for Sp(n, R) ∩ O(2n, R) = U (n) =
GL(n, C) ∩ O(2n, R) Since, an O(2n, R)-structure exists on any 2n-dimensional
3The notation for this group is by no means standard. It is sometimes written as Sp(2n, R),
and is written as Sp(2n) in [MS98]. We shall employ the notation Sp(n) for the ‘compact
symplectic group’ which is different.
 1.4. G-STRUCTURES 31

manifold, we see that M 2n admits an almost symplectic structure if and only if it

admits an almost complex structure. Of course, the two structures are different.
By Proposition 1.4.4 an Sp(n, R)-structure is integrable if and only if the tensor
field ω has constant components with respect to some local coordinates. In this
case there are local coordinates (x1 , . . . , xn , y1 , . . . , yn ) on M such that ω takes the
form
X n
ω= dxi ∧ dyi .
i=1
A well-known theorem of Darboux says that an Sp(n, R)-structure is integrable if
and only if the tensor field ω is closed, that is dω = 0. An integrable Sp(n, R)-
structure on M is called a symplectic structure. In this case the pair (M, ω) is
called a symplectic manifold.
Let (M, ω) be a symplectic manifold and consider the group S(M, ω) of sym-
plectomorphisms on M, i.e., the subgroup of Diff M that leaves ω invariant. The
‘Lie algebra’ s(M, ω) of S(M, ω) is the subalgebra of vector fields X on M such that
£X ω = 0. This equation implies that the 1-form X ω is closed. If it is also exact
then the there is a smooth function H on M such that X ω = dH. In this case
the function H is called a Hamiltonian function and the X a Hamiltonian vector
field. Notice that H is only defined up to a constant and if H 1 (M, R) = 0 then
every X ∈ s(M, ω) is Hamiltonian. We refer to [LM87] for further development.
From the discussion above about Lie groups, it should be clear that an almost
Hermitian structure gives naturally an almost symplectic structure. So we recover
Example 1.4.10 as a particular case of an almost symplectic structure with an added
compatible Riemannian structure.
Exercise 1.10: Define the conformal symplectic group by
CSp(n, R) = {A ∈ GL(2n, R) | At J0 A = cJ0 , c ∈ R∗ } ,
where R∗ denotes R − {0}. Show that a CSp(n, R)-structure on M coincides with
a choice of equivalence class of non-degenerate 2-forms, where the 2-forms ω and
ω 0 are equivalent if there is a nowhere zero function λ on M such that ω 0 = λω.
Such a structure can be called a conformal almost symplectic structure. Show that
if the CSp(n, R)-structure is integrable, then the non-degenerate 2-forms are closed
and λ is a constant. The resulting structure is known as a conformal symplectic
structure.
Example 1.4.13: p-dimensional distribution A p-dimensional distribution on
an n-dimensional manifold M is a choice of p-dimensional subbundle E of the
tangent bundle T M. The group G that stabilizes E is given by
nµ ¶ o
A 0
GL(p, q; R) = | A ∈ GL(p, R), B ∈ GL(q, R), C ∈ Mq,p ,
C B
where Mp,q denotes the p by q real matrices, and p + q = n. An integrable
GL(p, q; R)-structure on M is just a foliation F of M, that is, every point of M
has a ‘foliated coordinate chart’ (U ; φ) with coordinates (x1 , . . . , xp ; y1 , . . . , yq ) such
∂
that ( ∂x 1
, . . . , ∂x∂ p ) spans E|U, and the Jacobian matrix of the change of foliated
charts lies in GL(p, q; R). An atlas U = {(Uα ; φα )}α∈I of such foliated charts is
called a foliated atlas for M. Two foliated atlases U and V of the same codimen-
sion q correspond to the same foliation F if and only if they satisfy the coherence
property that the disjoint union U t V is a foliated atlas for F.
32 1. STRUCTURES ON MANIFOLDS

Globally, the foliation F gives a partition of M into disjoint p-dimensional

immersed submanifolds Lα , called the leaves of F. By the classical Frobenius The-
orem, a p-dimensional distribution is integrable if and only if it involutive, that is
the Lie bracket of any sections of the subbundle E is also a section of E. This has a
dual formulation as follows: let E 0 denote the subbundle of T ∗ M that annihilates
E. Then the Frobenius Theorem says that E is integrable if and only if the ideal
I(E 0 ) generated by the sections of E 0 is closed under exterior differentiation, i.e.,
I(E 0 ) is a differential ideal. Of particular interest in foliation theory is a study
of the transverse geometry and this is studied in more detail in Chapter 2. The
transverse geometry is the geometry of the space of leaves of the foliation defined
to be the quotient space defined by the equivalence relation that x ∼ y if x and
y lie on the same leaf. However, this can be very crudely defined as a topological
space, so one studies the transverse geometry directly on the foliated manifold.
Alternatively, a foliation can be described by patching together local submer-
sions. This is the point of view advanced by Haefliger [Hae70] and will be discussed
in Section 2.2. A (global) submersion (hence, a fibre bundle) is a special case of a
foliation. It is often called also a simple foliation. More development of foliation
theory is given below in Chapter 2.
There are various special cases of Example 1.4.13 of special interest.
Example 1.4.14: Almost product structures. This is Example 1.4.13 with
C = 0. In this case the tangent bundle to M splits as a direct sum of two subbun-
dles T M = D+ ⊕ D− . By Proposition 1.4.3 an integrable almost product structure
implies the integrability of the complementary subbundles. In particular, the uni-
versal covering M̃ is actually a product of two manifolds, one of dimension p and
the other of dimension q. One can easily see that an almost product structure is
equivalent to the existence of an endomorphism P ∈ End(T M ) such that P 2 = 1l.
Namely, given such P we can introduce the projection operators
1
P± = (1l ± P ), P+ ◦ P− = P− ◦ P+ = 0, P+ + P− = 1l.
2
We get the desired spitting T M = D+ ⊕ D− by taking D± = P± (T M ). As in
the almost complex case the Nijenhuis tensor NP (X, Y ) is the obstruction to the
integrability of the two distributions D± . A Riemannian almost product structure
is a further reduction of the structure group to O(p, R) × O(q, R) which is not
obstructed, i.e., if a manifold admits an almost product structure it also admits an
almost product Riemannian metric structure. It follows that a Riemannian almost
product structure is a triple (M, Π, g), where Π is an almost product structure and
g is a Riemannian metric with the property
g(P X, P Y ) = g(X, Y ).
A Riemannian almost product structure is said to be a locally Riemannian product
structure if ∇P = 0, i.e., Π is parallel with respect to the Levi-Civita connection of
g. The condition ∇P = 0 implies the integrability NP (X, Y ) = 0. Note also that if
a manifold M admits an almost product structure with group GL(p, R) × GL(q, R)
then it also admits an O(p, q) pseudo-Riemannian structure.
Exercise 1.11: Let (M n , P, g) be an almost product structure. Show that P is a
local O(n, R) matrix with all eigenvalues ±1 and Tr(P ) = 2q −n, where q = rk(D+ )
is the number of positive eigenvalues. Furthermore, show that one can choose a
local chart and a basis so that P, P± take the following simple forms
 1.4. G-STRUCTURES 33

µ ¶ µ ¶ µ ¶
1lq 0 1lq 0 0 0
P = , P+ = , P− = .
0 −1ln−q 0 0 0 1ln−q

We are especially interested in almost product structures when p = 2m and

q = 1 which is related to almost contact structures treated in detail in Chapter 6.
Unfortunately, the terminology with regard to contact and almost contact struc-
tures is very misleading. For example, a contact structure is generally not an almost
contact structure in contrast to the usual use of this terminology; moreover both
of these structures are defined in [Gra59]! This was noted and fixed in [Sto74], at
least for topological purposes, but we would like consistency with geometric struc-
tures as well. We shall refer to our structure as an almost contact G-structure to
distinguish it from an almost contact structure whose definition is by now too well
ensconced in the literature.
Example 1.4.15: Almost contact G-structure. Consider the group G defined
as follows:
nµ ¶ o
A 0
G= | A ∈ CSp(m, R), b ∈ GL(1, R), c ∈ M1,2m .
c b
This G-structure picks out a codimension 1 subbundle with a conformal almost
symplectic structure on it. Here an integrable G-structure is related to what could
be called a conformal cosymplectic structure.
In the case of Example 1.4.15 we shall be more interested in the non-integrable
case. In Chapter 6 we shall describe its relation with almost contact and contact
structures.
Another structure that is related to both Examples 1.4.9 and 1.4.13 is a CR-
structure. Here the concept of an almost CR-structure is not standard. Indeed
the definition of a CR-structure varies as well. Some authors require that a CR-
structure be defined by a codimension one subbundle (cf. [Jac90]), the idea being
that of modelling a real hypersurface in Cn+1 . We shall be more general.
Example 1.4.16: Almost CR-structures. An almost CR-structure of codimen-
sion q is a G-structure with G given by
nµA 0 ¶ o
G= | A ∈ GL(m, C), B ∈ GL(q, R), C ∈ Mq,2m ,
C B
where GL(m, C) is realized as a subgroup of GL(2m, R) as described in Example
1.4.9. This G picks out a 2m-dimensional subbundle E of T M with an almost
complex structure, that is a smooth section J of the endomorphism bundle End E
satisfying J 2 = −1lE . We say that (E, J) defines a CR-structure on M if for any
smooth sections X, Y of E the vector field [JX, Y ] + [X, JY ] is a section of E and
the Nijenhuis tensor NJ , as defined in Example 1.4.9, vanishes. We are interested
in the case of codimension one, i.e., q = 1. If M is orientable then there exists a
1-form η on M such that E = ker η. The symmetric bilinear form L(X, Y ) on E
defined by L(X, Y ) = dη(X, JY ) is called the Levi form of the CR-structure. The
CR-structure is said to be strictly pseudoconvex if L is either positive or negative
definite. For an extensive treatment of CR-structures we refer the reader to the
recent book [DT06].
34 1. STRUCTURES ON MANIFOLDS

Example 1.4.17: f -structures. An f -structure of corank q is a special kind of

an almost product structure. It is a G-structure with G given by
nµA 0 ¶ o
G= | A ∈ GL(n, C), B ∈ GL(q, R) .
0 B
f -structures were first introduced and studied by Yano [Yan63]. One can easily see
that the reduction implies an existence of a non-vanishing section f of the bundle
End(T M ) with constant rank such that
f 3 + f = 0.
Given f we define two commuting projections P+ = −f 2 and P− = f 2 + 1l, which,
in turn, give the splitting T M = D+ ⊕ D− . Clearly, f 2 (f X) = −f (X) so that f
defines and almost complex structure on the subbundle D+ = Im(f ) ⊂ T M . In
particular, further reduction to the subgroup U (n) × O(q, R) can be achieved by
choosing a locally product Riemannian metric g on M which is compatible with f
in the following natural sense
g(f X, Y ) = −g(X, f Y ).
Clearly, such a metric always exists and the triple (M, f, g) is said to be a metric f -
structure. When q = 0, an f -structure is simply an almost complex structure, and
with a compatible metric it is an almost Hermitian structure of two earlier examples.
When q = 1 an f -structure is a special case of an almost contact G-structure. In
particular, when q = 1 and the structure group reduces further to GL(n, C) × 1 an
f -structure is a (strict) almost contact structure discussed in Chapter 6. It follows
that there exist a nowhere vanishing vector field ξ and its dual 1-form η such that
[Bla76a]
(1.4.5) f = −1l + η ⊗ ξ.
Note that in the latter case picking a compatible Riemannian metric, i.e., the
metric g which satisfies g(f X, f Y ) = g(X, Y ) − η(X)η(Y ) is always possible and
gives further reduction of the structure group to U (n) × 1 which is the standard
definition of an almost contact (metric) structure due to Gray [Gra59]. When q ≥ 2
one can consider the so-called globally framed (metric) f -structure as the further
reduction if the structure group to U (n) × 1lq . The definition implies that the
bundle D− = Ker(f ) is parallelizable, i.e., there exist global frame field {ξ1 , . . . , ξs }
spanning D− = Ker(f ). Such structures were introduced and studied by Goldberg
and Yano [GY71, GY70] and Blair [Bla70]. In some literature they are also
called f.pk-structure (see the recent book [FIP04] for more details and references).
Example 1.4.18: Quaternionic structures. Quaternionic G-structures ap-
peared early on in the work of Libermann [Lib54], Berger [Ber55], Obata [Oba56],
and Martinelli [Mar59], and continued in [Ale68, Bon67, Gra69a, Wol65] to
cite a few. Our brief treatment here follows that of Salamon [Sal86] with further
development in Chapter 12. Let Hn be a quaternionic vector space with n > 1.
By writing each quaternionic coordinate as ui = u0i + u1i i + u2i j + u3i k we natu-
rally identify Hn ' R4n . Let GL(n, H) ⊂ GL(4n, R) be a group of non-singular
quaternionic n × n matrices and consider the product GL(n, H) × GL(1, H). For
any (A, λ) we can define GL(n, H) × GL(1, H)-action on Hn by (u; (A, λ)) 7→ A · uλ.
The central R∗ acts trivially and we denote the quotient group by GL(n, H)Sp(1),
 1.4. G-STRUCTURES 35

where Sp(1) ⊂ GL(1, H) is the subgroup of unit quaternions. In particular, it fol-

lows that G = GL(n, H)Sp(1) ¡ is a subgroup
¢ of GL(4n, R), and one easily sees
that GL(n, H)Sp(1) = GL(n, H) × Sp(1) /Z2 . An almost quaternionic struc-
ture on a manifold M is a reduction of the frame bundle L(M ) to the subgroup
GL(n, H)Sp(1). Theorem 1.2.5 gives a one-to-one correspondence between sections
of the associated vector bundle E(M, GL(4n, R)/G and reductions Q of the frame
bundle L(M ) to the group G. Moreover, in analogy with Example 1.4.9 we see that
sections of E(M, GL(4n, R)/G) correspond to the existence of a triple of local sec-
tions {I1 , I2 , I3 } of a rank 3 vector subbundle Q ⊂ End(T M ) satisfying the algebra
of the quaternions. Hence, one can consider either a reduction to the bundle Q
or the existence of the subbundle Q as an almost quaternionic manifold. In the
world of quaternionic geometry the usual notion of integrability is much too strong.
For example, the only simply connected compact manifold which admits a quater-
nionic atlas whose transition functions are in GL(n, H)Sp(1) is the quaternionic
projective space HP(n) [Mar70, Som75]. More interesting quaternionic geome-
tries emerge by considering 1-integrable (recall Definition 1.4.6) almost quaternionic
G-structures. A 1-integrable almost quaternionic structure is called a quaternionic
structure. In the case of quaternionic structures torsion-free connections are not
unique. Such a connection [Opr77, Opr84] is called an Oproiu connection in
[AM96b] or a quaternionic connection.
One can always equip M with a Riemannian metric g which preserves Q, i.e.,
g(JY, JY ) = g(X, Y ) for any local section J of Q. The triple (M, Q, g) is then
called almost quaternionic Hermitian manifold. This corresponds to a further re-
duction of the bundle Q to its maximal compact subgroup Sp(n)Sp(1). An almost
quaternionic (Hermitian) manifold with a torsion-free connection preserving Q is
called a quaternionic (Hermitian) manifold [Sal86]. In dimension greater than 4 an
almost quaternionic Hermitian structure is 1-integrable if and only if it is quater-
nionic Kähler [AM96b]. The usual definition of a quaternionic Kähler structure
(in dimension greater than 4) is that the Riemannian holonomy lies in Sp(n)Sp(1).
Quaternionic structures will be described in more detail in Chapter 12. In dimen-
sion 4 the definitions of quaternionic and quaternionic Kähler involve ‘duality’ of
Weyl’s conformal curvature and will be defined in Chapter 12.
Example 1.4.19: Hypercomplex structure. A GL(n, H)-structure on a mani-
fold M is called an almost hypercomplex structure. Thus, an almost hypercomplex
manifold is the special case of an almost quaternionic manifold, where the sub-
bundle Q ⊂ End(T M ) admits a trivialization by global sections I ≡ {I1 , I2 , I3 }
satisfying quaternion algebra. Notice that the quaternionic relations on Q allows
one to define a two-sphere’s worth of almost complex structures

(1.4.6) I(τ ) = τ1 I1 + τ2 I2 + τ3 I3 ,

where τ = (τ1 , τ2 , τ3 ) ∈ S 2 . So M together with I(τ ) is usually called an almost

hypercomplex manifold. A GL(n, H)-connection satisfies ∇I(τ ) = 0 for all τ ∈
S 2 . An almost hypercomplex structure that is 1-integrable is called hypercomplex.
In this case torsion-free GL(n, H)-connections are unique. Such a connection is
called the Obata connection . Compact four dimensional manifolds admitting a
hypercomplex structure were classified in [Boy88a].
One can always equip M with a Riemannian metric g which preserves I(τ ),
i.e., g(I(τ )Y, I(τ )Y ) = g(X, Y ) for any τ ∈ S 2 . The triple (M, I, h) is then called
36 1. STRUCTURES ON MANIFOLDS

an almost hyperhermitian manifold and provides further reduction of the structure

group to Sp(n). A hypercomplex structure with a compatible Riemannian metric
is called a hyperhermitian structure. An almost hyperhermitian structure is 1-
integrable if and only if it is hyperkähler. This is equivalent to the usual definition
of a hyperkähler structure as one in which the Riemannian holonomy is contained
in Sp(n). Both hypercomplex and hyperkähler structures are treated in more detail
in Chapter 12.
Example 1.4.20: Quaternionic f -structures. A G-structure with the group
G = GL(n, H)Sp(1) × 1lm ⊂ GL(4n + m, R) is called an almost quaternionic f -
structure of corank m. It can be seen that such a reduction is equivalent to the
existence of two subbundles Q ⊂ End(T M ) and E ⊂ T M of rank 3 and m respec-
tively, such that Q admits local sections {J1 , J2 , J3 } satisfying quaternionic algebra
relations, and E is parallelizable. One can choose global sections {ξ1 , . . . , ξm } of E
so that
Xm
Ji (ξα ) = 0, ηα ◦ Ji = 0, Ji2 = −1l + ηα ⊗ ξα ,
α=1
where α = {1, . . . , m}, i ∈ {1, 2, 3} and {η1 , . . . , ηm } is the dual frame in E ∗ . When
the sections {J1 , J2 , J3 } are global, i.e., the bundle Q is trivial we call (M, Q, E) a
hyper f -structure of corank m. Such structure is a G-structure with G = Sp(n)×1lm .
When m = 0 these definitions reduce to the usual notions of almost quaternionic
and hypercomplex geometry in the two previous examples. When m = 3 a hyper f -
structure is called an almost hypercontact structure introduced by Kuo [Kuo70] as
an almost contact (metric) 3-structures. The geometry of quaternionic and hyper
f -structures of corank 3 was studied by [Her96] and we will come back to some of
these examples in Chapter 13.

1.5. Pseudogroup Structures

We have seen that the notion of G-structure is a unifying concept for study-
ing geometric structures. Another unifying concept is the idea of a pseudogroup
of transformations. Roughly speaking a pseudogroup is a ‘group of transforma-
tions’, where there may be some problems with domains of definition. The concept
goes back to Lie and Cartan, but the precise notion was defined more recently by
Ehresmann [Ehr53]. The ideas were then developed much further by Kodaira and
Spencer among others, cf. [GS66, Kod60b, KS58, Spe62]. Here is the definition
following [Kob61]
Definition 1.5.1: A pseudogroup of transformations on Rn is a set Γ of local
diffeomorphisms which satisfy
(i) Each f ∈ Γ is a diffeomorphism of an open subset of Rn , called the
domain of f , onto an open subset of Rn , called the range of f.
(ii) If U = ∪i Ui , where each Ui is an open subset of Rn , then a diffeomor-
phism f with domain U belongs to Γ if and only if its restriction to each
Ui belongs to Γ.
(iii) For every open subset U of Rn , the identity transformation on U belongs
to Γ.
(iv) If f ∈ Γ then so is f −1 .
(v) If f : U −−→V and f 0 : U 0 −−→V 0 are elements of Γ and V ∩ U 0 6= ∅, then
the diffeomorphism f 0 ◦ f : f −1 (V ∩ U 0 )−−→f 0 (V ∩ U 0 ) belongs to Γ.
 1.5. PSEUDOGROUP STRUCTURES 37

Γ is said to be transitive if for every pair of points p, q ∈ Rn there is an f ∈ Γ

such that f (p) = q.
More generally, one can replace Rn by a different ‘model space’ [Kob61], for
example a smooth manifold M. The importance of pseudogroups for our purpose
lies in its ability to describe different types of atlases of coordinate charts.
Definition 1.5.2: Let Γ be a transitive pseudogroup, and M a second countable,
Hausdorff topological space. A Γ-atlas for M is a family of pairs {(Uα , φα )}α∈I ,
called charts, such that
(i) The collection {Uα }α∈I of open sets covers M, i.e., each Uα is an open
subset of M, and ∪α Uα = M.
(ii) Each φα is a homeomorphism of Uα onto an open subset φα (Uα ) ⊂ Rn
such that whenever Uα ∩ Uβ 6= ∅, the map
φβ ◦ φ−1
α : φα (Uα ∩ Uβ ) : −
−→φβ (Uα ∩ Uβ )
belongs to Γ.
A Γ-atlas on M is said to be maximal if it is not properly contained in any
other Γ-atlas on M.
Clearly, if Γ is the pseudogroup of all local diffeomorphisms of Rn , this definition
coincides with the usual definition of an atlas of coordinate charts for a differentiable
manifold. As in the usual case, it is easily seen that every Γ-atlas is contained in
a unique maximal Γ-atlas. Thus, a given Γ-atlas defines uniquely a Γ-structure on
M. The relation between transitive pseudogroups and G-structure should now be
evident from Proposition 1.4.3.
Proposition 1.5.3: Let G be a Lie subgroup of GL(n, R), and let ΓG be the pseu-
dogroup of local diffeomorphisms of Rn whose Jacobian matrix lies in G at each
point of its domain. Then the ΓG -structures on M are in 1 − 1 correspondence with
the integrable G-structures on M.
It should be clear that the pseudogroup ΓG is transitive since it contains the
translations. Notice that not every Γ-structure is a ΓG -structure for some G ⊂
GL(n, R). We will see an important example of this later. Of course, as with
integrable G-structures, or just G-structures, for a given Γ there may not exist any
Γ-structure at all on M. That is, generally, there are topological obstructions to
the existence of a Γ-structure. Two important ΓG -structures are for G = GL(n, C)
and G = Sp( n2 , R). It is clear from the two examples 1.4.9 and 1.4.12 that the
ΓGL(n,C) -structures are just the complex structures, while the ΓSp( n2 ,R) -structures
are the symplectic structures. The example of the six sphere S 6 is interesting in
this context. As mentioned in the paragraph after Example 1.4.9 S 6 is known to
admit an almost complex structure, i.e., a GL(3, C)-structure, but it is not known
whether it has a complex structure, i.e., a ΓGL(3,C) -structure. From the symplectic
point of view, S 6 also admits an almost symplectic structure, i.e., an Sp(3, R)-
structure, since it admits an almost complex structure; however, it cannot admit
a symplectic structure, i.e., a ΓSp(3,R) -structure, since the closed 2-form ω would
provide a non-vanishing element of H 2 (S 6 , R) by Stokes Theorem.
Following [SS65] we relate pseudogroups to collections of vector fields. Let
{Xα } be a collection of smooth vector fields each defined on an open set Uα ⊂ M.
Each Xα generates a local1-parameter group φα (t) of local diffeomorphisms defined
for small t and for some open subset of Uα . The collection {φα } defines a family
38 1. STRUCTURES ON MANIFOLDS

of local diffeomorphisms and thus generates a pseudogroup P on M. Fix a point

x ∈ M and consider the subpseudogroup P0 of P generated by those vector fields
in the collection that vanish at x. Then any local 1-parameter group φα (t) ∈ P0
satisfies φα (t)(x) = x for all t in its domain. The differential φα (t)∗ of any such
local 1-parameter group is a linear map on the vector space Tx M. Running through
all such local 1-parameter groups in P0 , determines a Lie subalgebra g0 of gl(n, R).
We call g0 the linear isotropy algebra of the pseudogroup P. This definition differs a
bit from the one in [SS65], but they should be equivalent. The connected subgroup
G0 of GL(n, R) determined by the Lie algebra g0 is called the linear isotropy group
of P.
Let Γ be a pseudogroup of local transformations4 on a smooth manifold M. We
can consider germs of elements f ∈ Γ, that is two maps f, g : U −−→V in Γ sending
x ∈ M to y ∈ M are equivalent if there is an open set W ⊂ U such that f |W = g|W .
The equivalence class fx is the germ of the map f at x. Note that fx can only be
evaluated at x in which case fx (x) = y. Thus, we have a source map s(fx ) = x,
a target map t(fx ) = fx (x) = y, and a composition defined as follows: if fx is a
germ of a map from x ∈ M to y ∈ M, and gy is a germ of a map from y to z, then
the composition gy ◦ fx is well defined and is the germ at x of the map g ◦ f from
x to z. One easily sees that the set G(Γ) of germs of maps of the pseudogroup Γ
forms a groupoid (cf. Appendix A), called a groupoid of germs of the pseudogroup
Γ. Furthermore, we can give G(Γ) a topology that makes the source map s a local
homeomorphism. So s : G(Γ)−−→M becomes an étale space over M. Thus, G(Γ)
is an étale groupoid constructed from germs, or in the terminology of Haefliger
[Hae58], a faisceau de groupoids (sheaf of groupoids). Conversely given a groupoid
of germs G on a manifold M, we can reconstruct a pseudogroup as follows: let
σ : U −−→G be a local section of the source map of G. Composing σ with the target
projection t : G−−→M gives a local diffeomorphism of M, and it is easy to check that
the set of all such diffeomorphisms forms a pseudogroup of local transformations of
M. Summarizing we have
Proposition 1.5.4: Let M be a smooth manifold and Γ a pseudogroup of local
transformations on M. Then the set G(Γ) can be given the structure of a étale
groupoid. Conversely given a groupoid of germs G over M, the set
{t ◦ σ | t is the target projection, σ : U −−→G is a local section of the source map s}
is a pseudogroup of local transformations on M.
This proposition says that the notion of pseudogroups of transformation on M
and that of groupoids of germs on M are equivalent.
Exercise 1.12: Check the details of Proposition 1.5.4. That is, given a pseu-
dogroup of transformations on M, show that passing to germs one obtains an étale
groupoid, and conversely, given a groupoid of germs on M, the set defined above
forms a pseudogroup.

1.6. Group Actions on Manifolds

In this section we will review some basic properties of group actions on mani-
folds. This is a classical subject treated in several texts [MZ55, Bre72, Kob72,
4By a pseudogroup of local transformations on M we mean any subpseudogroup of the
pseudogroup of local diffeomorphisms of M. More generally one could take M to be a topological
space and Γ a subpseudogroup of the pseudogroup of local homeomorphisms of M .
 1.6. GROUP ACTIONS ON MANIFOLDS 39

DK00] to name a few. We mainly follow the approach and the notation of Duis-
termaat and Kolk and refer the reader to their modern textbook for more details
[DK00]. Recall that a topological group is a group G endowed with a topology
such that the multiplication map µ : M × M −→M , µ(g1 , g2 ) = g1 g2 and the inverse
map ι : M −→M , ι(g) = g −1 are continuous. A discrete group is a topological group
with the discrete topology. For a topological manifold M we denote by Hom(M )
the group of homeomorphisms of M. If M is a smooth manifold the group of diffeo-
morphisms is denoted by Diff(M ). We give Diff(M ) the compact-open C ∞ topology.
This is a good topology when M is compact, but in the non-compact case it does
not control the behavior at infinity. Since we deal mainly with compact manifolds
we do not concern ourselves with this failure. We refer the reader to [Ban97] for
further discussion of diffeomorphism groups.
Definition 1.6.1: An action A of a topological group G on a topological manifold
M is a homomorphism A : G −→ Hom(M ) of G into the group of homeomorphisms
of M . We say that the action A is continuous if the mapping
(1.6.1) (g, x) 7→ g · x = A(g)(x) : G × M −−→ M,
also denoted by A : G × M −→ M, is continuous. We say that A is a proper
action if the associated map G × M −→ M × M given by
(1.6.2) (g, x) 7→ (x, g · x)
is proper, i.e, the inverse image of any compact set is compact. If M is a smooth
manifold and G is only a topological group we say that A is smooth if A(g) ∈
Diff(M ) for each g ∈ G. However, if G is a Lie group a smooth action A means
that the map (1.6.1) is smooth.
Let A be an action of G on M and let e ∈ G be the identity element. For any
x ∈ M the subgroup Gx = {g ∈ G | g · x = x} is called the isotropy subgroup of the
action at the point x ∈ M , or the stabilizer of x ∈ M under the action. Then A is
called effective if Ker(A) = {1l}. The action A is free if Gx = {e} for all x ∈ M
and locally free if Gx is a finite group for all x ∈ M .
For each x ∈ M the orbit through x is defined as
(1.6.3) G · x = {y ∈ M |y = g · x, g ∈ G} ⊂ M.
The orbit G · x through x is an immersed submanifold of M, and there is a natural
identification of G · x with the coset space G/Gx .
It is clear that the relation x ∼ y if y ∈ G · x is an equivalence relation
partitioning M into orbits. We shall denote the quotient space M/ ∼ by M/G,
and call it the space of orbits. The map π : M −→M/G sending x 7→ G · x is the
canonical projection. We equip M/G with the quotient topology, i.e., V is open in
M/G if and only if π −1 (V ) is open in M .
Exercise 1.13: Let M be a manifold with group action A : G × M −→ M . Show
that M/G is Hausdorff if and only if the set {(x, y) ∈ M × M | y ∈ G · x} is a
closed subset of M × M .
In this book we will mainly be concerned with actions which yield Hausdorff
quotients. The importance of proper group actions is realized through the following
Proposition 1.6.2: If the action A of G on M is proper then the quotient M/G
is Hausdorff.
40 1. STRUCTURES ON MANIFOLDS

Example 1.6.3: A typical example of a non-proper group action is the so-called

irrational flow on a torus. Let M = T 2 = SO(2)×SO(2) which is an Abelian group.
So for any α ∈ R we can consider a subgroup G(α) = {(A(t), A(αt), t ∈ R}, where
µ ¶
cos t − sin t
A(t) = .
− sin t cos t
Then G(α) acts on M by left (or right) multiplication. For α ∈ Q the orbits are
all closed in M . In this case G(α) is homeomorphic to S 1 and so is the quotient
M/G(α). However, for α ∈ R \ Q the orbits are not closed. They are immersed but
not embedded submanifolds homeomorphic to R.
The closure of any orbit equals to M and the quotient M/G(α) is non-Hausdorff.
There are also examples with non-Hausdorff quotients with all orbits closed.
Consider, for instance, G = R acting on R2 via
¡ ¢
t, (x1 , x2 ) 7→ (x1 + tx2 , x2 ), t ∈ R, (x1 , x2 ) ∈ R2 .
Here the orbits are indeed closed: they are either lines parallel to the x1 -axis, or
points on the x1 -axis. It is easy to see that the quotient is non-Hausdorff.
Definition 1.6.4: A G-manifold is a smooth manifold M together with a smooth
action A of a group G.
The importance of proper and free actions is immediately realized by the fol-
lowing classical result.
Theorem 1.6.5: Let G be a Lie group acting smoothly, freely, and properly on
M . Then the orbit space M/G is a smooth manifold of dimension dim(M ) −
dim(G) with a unique differentiable structure having the following property: if
π : M −→ M/G is the canonical projection then for every a ∈ M/G there exists an
open neighborhood U 3 a and a diffeomorphism
τ = (χ, s) : π −1 (U) −→ G × U
such that, for each x ∈ π −1 (U), g ∈ G we have
(1.6.4) π(x) = s(x), τ (g · x) = (gχ(x), s(x)).
In fact, it follows that the map π is a smooth submersion and M = P (M/G, G)
is a principal G-bundle over M/G, and the map τ is a local trivialization. If we add
a bit more structure the hypothesis on the action of Theorem 1.6.5 can be weakened
to be only locally free. This is of great importance to us in this book, and will be
discussed further in the forthcoming chapters. See, for example, Theorem 2.5.11
below.
A case of particular interest to us is when a group G leaves invariant some
geometric structure. Following [Kob72] we have
Definition 1.6.6: Let P be a G-structure over M. A diffeomorphism f : M −−→M
is an automorphism of the G-structure if the induced map f∗ : P −−→P is an
isomorphism.
The automorphisms of P form a subgroup of Diff(M ) and is denoted by Aut(P ).
In particular, if the G-structure P is defined by a tensor field T then f : M −−→M
is an automorphism of P if and only if it leaves T invariant. There are many
such examples of interest to us throughout this book. We shall often denote the
automorphism group by Aut(T ), but in some cases there is more common standard
terminology. For example, if (M, g) is a Riemannian manifold the automorphism
 1.6. GROUP ACTIONS ON MANIFOLDS 41

group Aut(g) is called the isometry group and denoted by Isom(M, g). We shall
come across other examples as we proceed. Since Aut(P ) a subgroup of Diff(M ) it
is endowed with the compact-open topology, and we are interested in when it is a
Lie group. Recall [Kob72] that a Lie subalgebra g ⊂ gl(n, R) is said to be elliptic
if g contains no matrix of rank one. Then we have
Proposition 1.6.7: Let P be a G-structure on a compact manifold M with group
G and Lie algebra g. If g is elliptic then Aut(P ) is a Lie group with respect to the
compact-open C ∞ topology.
Examples of elliptic Lie algebras are the orthogonal Lie algebras o(n, R), the
complex Lie algebra gl(n, C) ⊂ gl(2n, R), and any of their subalgebras. Thus,
the automorphism groups of a compact Riemannian manifold, or of a compact
almost complex manifold are Lie groups. In general compactness is a necessary
hypothesis for Aut(P ) of an elliptic G-structure to be a Lie group. For example the
complex automorphism group Aut(Cn ) is not a Lie group. However, for Riemannian
and certain other G-structures compactness is not necessary. In order to develop
more we need the notion of prolongation. Since this is amply treated in several
books [Ste83, Kob72] we give only a very brief treatment here. Let V be an
n-dimensional vector space and g a Lie subalgebra of gl(n, R) = End(V ).
Definition 1.6.8: The first prolongation g(1) of g is defined by
g(1) = {T ∈ Hom(V, g) | T (u)v = T (v)u for all u, v ∈ V }.
(1)
The k th prolongation g(k) is defined inductively by g(k+1) = g(k) . A Lie algebra
g is said to be of finite type of order k if for some k, g(k) = 0, but g(k−1) 6= 0
otherwise g has infinite type.
Similarly one can define the prolongations of a Lie group G ⊂ GL(n, R). Such
prolongations give rise to higher order G-structures for which we have no real need.
We say the a G-structure is of finite type of order k if its Lie algebra g is of finite
type of order k. It is easy to see that if h ⊂ g and g(k) = 0 then h(k) = 0. It
is equally easy to show that o(n, R)(1) = 0, and that gl(n, R) and gl(n, C) are of
infinite type. However, gl(n, H) is of finite type, in fact gl(n, H)(1) = 0. Many of the
G-structures that are of importance in this book are of infinite type, in particular
those of Examples 1.4.9, 1.4.12 1.4.13 1.4.15, and 1.4.16. G-structures that are of
finite type, but with g(1) 6= 0 are also of interest. This occurs in two important
cases for us, the conformal G-structures with G = CO(n) and the quaternionic G-
structures with G = GL(n, H)Sp(1). These both have g(1) 6= 0, but g(2) = 0. When
a G-structure is 1-integrable, the vector space g(1) parameterizes the torsion-free
connections, and if g(1) = 0 the torsion-free connection is unique. We now have
Theorem 1.6.9: Let P (M, G) be a G-structure of finite type of order k, Then
Aut(P ) is a Lie group of dimension at most dim M + dim g + dim g(1) + · · · +
dim g(k−1) .
The proof of this general theorem is given in [Ste83, Kob72]. It, of course,
has many precursors the most famous of which is the somewhat stronger result
essentially due to Myers and Steenrod Theorem [MS39]:
Theorem 1.6.10: Let (M, g) be a Riemannian manifold with finitely many con-
nected components. Then the group of isometries Isom(M, g) is a finite-dimensional
Lie group. The action of Isom(M, g) on M is proper and its Lie group topology
coincides with the compact-open C ∞ topology as a subgroup of Diff(M ).
42 1. STRUCTURES ON MANIFOLDS

In particular, this theorem says that the isometry group Isom(M, g) of a com-
pact manifold is compact. We discuss many other examples of G-structures of
finite type throughout the book. The G-structures dealt with in this book for
which Aut(P ) is not a Lie group include the symplectic and contact structures.
Definition 1.6.11: Let G be a group acting on a manifold M . We say that the
action is transitive if for any x, y ∈ M there exists g ∈ G such that y = g · x. In
such a case we say that M is G-homogeneous. If Aut(P ) is the automorphism
group of a G-structure P (M, G) we say that M is a homogeneous G-structure
if Aut(P ) acts transitively on M. If G is a Lie group which acts transitively on M
we say that M is a homogeneous manifold.
When a manifold admits a Lie group G acting transitively and is a subgroup
of the Aut(P ) for some G-structure P (M, G), we say that M is a homogeneous (G-
structure) manifold, for example a homogeneous complex or homogeneous contact
manifold, etc.
Example 1.6.12: Let G be a Lie group and H any subgroup. The H acts on G
from the left and from the right:

(1.6.5) AL (h)g = hg,

(1.6.6) AR (h)g = gh−1 .

The corresponding quotient spaces are denoted by H\G and G/H. Both of these
actions are free but they may or may not be proper. If H is a closed subgroup of
G then the H-action is proper and the quotient G/H is a homogeneous smooth
G-manifold under the natural left G-action.
More generally, we will be interested in proper actions which are not necessarily
free. In such case the space of orbits is not always a manifold; nevertheless, it has
a very tractable structure, namely that of an orbifold which we discuss in Chapter
4.
Consider the differential (Ax )∗ : g −→ Tx M , where g ' Te G is identified with
the tangent space of G at the identity g = e.
Definition 1.6.13: Let A : G × M −→ M be a smooth action of a Lie group on a
manifold M . A smooth slice at x0 ∈ M for the action A is a smooth submanifold
S ⊂ M , x0 ∈ S such that
(i) Tx0 M = (Ax0 )∗ (g) ⊕ Tx0 S and Tx0 M = (Ax0 )∗ (g) + Tx0 S for all x ∈ S,
(ii) S is Gx0 -invariant,
(iii) if x ∈ S, g ∈ G, and A(g)(x) ∈ S, then g ∈ Gx0 .
It follows that the inclusion S ,→ M induces a bijective map Gx0 · x 7→ G · x
from the space S/Gx0 of Gx0 orbits in S onto an open neighborhood of G · x0 in
M/G.
Definition 1.6.14: The action A : G × M −→ M is said to be proper at x0 if
for any convergent sequence xi −→x0 in M and a sequence {gi }i∈N in G such that
gi · xi −→x0 , there exists a convergent subsequence of {gi }i∈N .
Naturally, if A : G × M −→ M is proper then it is proper at every point x ∈ M .
But the converse is in general not true. However, if A is proper at every point of
M and M/G is Hausdorff then A is proper.
Exploring the Variety of Random
Documents with Different Content
prize guinea fowls. He is not sure, but he thinks that you may have
been one of them. How about it?"
Bobby looked uncomprehending for a moment while he covertly
studied Patrick. The man's air was apologetic; his accusation was
evidently based upon suspicion rather than proof.
"I went crabbing with Bert Holliday this morning," said Bobby.
"Ah!" his father's face cleared, though he still maintained his stern
tone. "I gave you strict orders, you remember, never to touch my
revolver when I was not with you?"
"Yes, father."
"You never have touched it?"
"No." Bobby's tone was barely audible.
"Speak up! I can't hear you."
"No!" snapped Bobby.
"Don't act that way. I am not accusing you of anything. I merely
wish to know the truth." Mr. Carter turned to Patrick, who was
nervously fumbling with his hat. "You see, Patrick, you were
mistaken. Tell Mr. Jasper that I am sorry about the guinea fowl, but
that Master Bobby had nothing to do with the shooting."
He dismissed the man with a nod, and mounted and rode away.
Peter watched him out of sight, then he turned and crossed the lawn
to the tennis-court. Bobby was back on his bench again engaged in
carving his name on the handle of a racket, though his face, Peter
noted, did not reflect much pleasure in the work. He glanced up
carelessly as Peter approached, but as he caught the look in his eye,
he flushed quickly, and with elaborate attention applied himself to
shaping a "C."
Peter sat down on the end of the bench and regarded him soberly.
He was uncertain in his own mind how he ought to deal with the
case, but that it must be dealt with, and drastically, he knew. Peter
was by no means a Puritan. The boy could accomplish any amount
of mischief—go crabbing instead of to Sunday-school, play fox and
geese over the newly sprouted garden, break windows and hotbeds,
steal cake from the pantry and peaches from Judge Benedict's
orchard, and Peter would always shield him. His code of morals was
broad, but where he did draw the line he drew it tight. Bobby's sins
must be the sins of a gentleman, and Peter's definition of
"gentleman" was old fashioned and strict.
Bobby grew restless under the silent scrutiny.
"What do you want?" he asked crossly. "If you don't look out you'll
make me cut my hand."
He closed the large blade with an easy air of unconcern, and
opening a smaller one, fell to work again. The knife was equipped
with five blades and a corkscrew; it was one of the dignities to which
Bobby had attained on his recent birthday. Peter stretched out his
hand and, taking possession of the knife, snapped it shut and
returned it.
"Put it in yer pocket an' pay attention to me."
"Oh, don't bother, Pete. I'm busy."
"Your father will be home before long," said Peter, significantly.
"Well, fire ahead. What do you want?"
"Ye told a lie—two o' them, to be accurate. Ye were one o' them
boys that shot the chicken an' ye did have the pistol."
"I didn't shoot his old chicken; it was Bert Holliday. And anyway he
didn't mean to; it flew straight in front of the target just as he fired."
"He had no business to be firin'. But it's not the chicken I'm mournin'
about; it's the lie."
"It's none of your business," said Bobby, sullenly.
"Then I'll make it me business! Either ye goes to yer father an' tells
him ye lied, or I will. Ye can take yer choice."
"Peter," Bobby began to plead, "he'll not give me the mustang—you
know he won't. I didn't mean to touch the revolver, but Bert forgot
his air rifle, and the boys were waiting to have a shooting match. I
won't do it again—honest, Peter—hope to die."
"It ain't no use, Master Bobby. Ye can't wheedle me. Ye told a lie an'
ye've got to be punished. Gentlemen don't tell lies—leastways, not
direct. They hires a lawyer like Judge Benedict to do it for them. If
ye keep on ye'll grow to be like the Judge yerself."
Bobby smiled wanly. The Judge, as Peter knew well, was his chiefest
aversion, owing to an unfortunate meeting under the peach trees.
"You've told lots of lies yourself!"
"There's different kinds o' lies," said Peter, "an' this is the kind that I
don't tell. It ain't that I'm fond o' carrying tales," he added, "but that
I wants to see ye grow up to be a thoroughbred."
Bobby changed his tactics.
"Father'll feel awfully bad; I hate to have him find it out."
Peter suppressed a grin.
"Boys ought always to be considerate o' their fathers' feelin's," he
conceded.
"And you know, Pete, that you want me to have the mustang. You
said yourself that it was a shame for a big boy like me to be riding
Toddles."
Peter folded his arms and studied the distance a moment with
thoughtful eyes; then he faced his companion with the air of
pronouncing an ultimatum.
"I'll tell ye what I'll do, Master Bobby, since ye're so anxious to save
yer father's feelin's. I'll agree not to mention the matter, an' ye can
take yer punishment from me at the end of a strap."
Bobby stared.
"Do you mean," he gasped, "that you want to whip me?"
"Well, no, I can't say as I want to, but I think it's me dooty. If ye was
a stable-boy and I caught ye in a lie like that, I'd wallop ye till ye
couldn't stand."
"I never was whipped in my life!"
"The more reason ye need it now. I've often thought, Master Bobby,
that a thorough lickin' would do ye good."
Bobby sprang to his feet.
"Tell him if you want. I don't care!"
"Just as ye please. He's over to Shannon Farms now buyin' the
mustang. When he gets back an' finds his son is a liar and a coward,
he'll be returnin' that horse by telephone."
Bobby's flight was suspended while he hung wavering between
indignation and desire.
"There it is," said Peter. "I won't go back on me word. Either ye
keeps a whole skin an' rides Toddles another year, or ye takes yer
lickin' like a man an' gets the horse. Ye can have an hour to think it
over."
He rose and sauntered unconcernedly toward the stables. Bobby
stared after him, several different emotions struggling for supremacy
in his freckled face; then he plunged his hands deep into his pockets
and turned down the lane with an attempt at a swagger as he
passed the stable door. At the paddock gate Toddles poked his
shaggy little head through the bars and whinnied insistently. But
Bobby, instead of bestowing the expected lump of sugar, shoved him
viciously with his elbow and scuffed on. He seated himself
precariously on the top rail of the pasture fence and fell to digging
holes in the wood with his new knife, cogitating meanwhile the two
alternatives he had been invited to consider.
They appealed to him as equally revolting. Only that morning he had
carelessly informed the boys that his father was going to buy him a
mustang—a brown and white circus mustang that was trained to
stand on its hind legs. The humiliation of losing the horse was more
than he could face. Yet, on the other hand, to be beaten like a
stable-boy for telling a lie! He had boasted to the Hartridge boys,
who did not enjoy such immunity, that he had never received a
flogging in his life. He might have stood it from his father—but from
Peter! Peter, who had always been his stanchest ally, who, on
occasion, had even deviated from the strict truth himself in order to
shield Bobby from justice. The boy already had his full quota of
parents; he did not relish having Peter usurp the rôle.
For thirty minutes he balanced on the fence, testing first one then
the other of the horns of his dilemma. But suddenly he saw, across
the fields where the high-road was visible, a horse and rider
approaching at a quick canter. He slid down and walked with an air
of grim resolution to the stables.
Peter was in the harness-room busily engaged in cleaning out the
closet. The floor was a litter of buckles and straps and horse
medicine.
"Well?" he inquired, as Bobby appeared in the door.
"You can give me that licking if you want," said Bobby, "but I tell you
now, I'll pay you back!"
"All right!" said Peter, cheerfully, reaching for a strap that hung
behind the door. "I'm ready if you are. We'll go down in the lower
meadow where there won't be no interruption."
He led the way and Bobby followed a dozen paces behind. They
paused in a secluded clump of willows.
"Take yer coat off," said Peter.
Bobby cast him one appealing glance, but his face was adamant.
"Take it off," he repeated.
Bobby complied without a word, his own face growing white.
Peter laid on the strap six times. He did not soften the blows in the
slightest; it was exactly the same flogging that a stable-boy would
have received under the same circumstances. Two tears slipped
down Bobby's cheeks, but he set his jaw hard and took it like a man.
Peter dropped the strap.
"I'm sorry, Master Bobby. I didn't like it any better than you, but it
had to be done. Are we friends?" he held out his hand.
"No, we're not friends!" Bobby snapped. He turned his back and put
on his coat; then he started for the house. "You'll be sorry," he threw
over his shoulder.
During the next few days Bobby ignored Peter. If he had any
business in the neighbourhood of the stables he addressed himself
ostentatiously to one of the under men. The rupture of their
friendship did not pass unmarked, though the grooms soon found
that it did not pay to be facetious on the subject. Billy, in return for
some jocular comments, spent an afternoon in adding a superfluous
lustre to already brilliant carriage lamps.
The mustang arrived, was christened Apache, and assigned to a box
stall. He possessed a slightly vicious eye and a tendency to buck, as
two of the grooms found to their cost while trying to ride him
bareback in the paddock. Peter shook his head dubiously as he
watched the unseating of the second groom.
"We'll put a curb bit on that horse. I don't just like his looks for a
youngster to ride."
"Huh!" said Billy, "Master Bobby ain't such a baby as everybody
thinks; he can manage him all right."
Word came out from the house that afternoon that Bobby was to try
the new mustang. Billy saddled the horses—Apache, and Blue Gypsy
for Miss Ethel, and a cob for Peter—and led them out, while Peter in
his most immaculate riding clothes swaggered after. The maids were
all on the back porch and the family at the porte-cochère to watch
the departure. Bobby would accept no assistance, but mounted from
the ground with a fine air of pride. Apache plunged a trifle, but the
boy was a horseman and he stuck to his saddle.
"Be careful, Bobby," his mother warned.
"You needn't worry about me," Bobby called back gaily. "I'm not
afraid of any horse living!"
Blue Gypsy never stood well, and Miss Ethel was already off. Bobby
started to follow, but he wheeled about to say:
"You come, Billy; I don't want Peter."
"Bobby, dear," his mother expostulated, "you don't know the horse;
it would be safer——"
"I want Billy! I won't go if Peter has to come tagging along."
Peter removed his foot from the stirrup and passed the horse over to
the groom. The cavalcade clattered off and he walked slowly back to
the stables. He felt the slight keenly. He could remember when he
had held Bobby, a baby in short dresses, on the back of his father's
hunter, when he had first taught the little hands to close about a
bridle. And now, when the boy had his first horse, not to go! Peter's
feeling for Bobby was almost paternal; the slight hurt not only his
pride but his affections as well.
He spent an hour puttering about the carriage room, whistling a
cheerful two-step and vainly pretending to himself that he felt in a
cheerful frame of mind. Then suddenly his music and his thoughts
were interrupted by the ringing of the house telephone bell, long
and insistently. He sprang to the instrument and heard Annie's voice,
her words punctuated by frightened sobs.
"Oh, Pete! Is that you? Something awful's happened. There's been
an accident. Master Bobby's been throwed. The doctor's telephoned
to get a room ready and have a nurse from the hospital here. You're
to hitch up Arab as fast as you can and drive to the hospital after
her. Oh, I hope he won't die!" she wailed.
Peter dropped the receiver and ran to Arab's stall. He led him out
and threw on the harness with hands that trembled so they could
scarcely fasten a buckle.
"Why can't I learn to mind me own business?" he groaned. "What
right have I to be floggin' Master Bobby?"
The young woman whom Peter brought back decided before the end
of the drive that the man beside her was crazy. All that she could get
in return for her inquiries as to the gravity of the accident was the
incoherent assertion:
"He's probably dead by now, ma'am, and if he is it's me that done
it."
As they turned in at the Willowbrook gate Peter strained forward to
catch sight of the house. A strange coupé was drawn up before the
porte-cochère. He involuntarily pulled Arab to a standstill and looked
away, but the nurse reached out and grasped the reins.
"Here, man, what is the matter with you? Hurry up! They may want
me to help get the boy in."
Peter drove on and sat staring woodenly while she sprang to the
ground and hurried forward. Mrs. Carter and the maids were
gathered in a frightened group on the steps. He could hear Miss
Ethel inside the carriage calling wildly:
"Do be quick! His head has commenced to bleed again."
The driver climbed down to help the doctor lift him out. They jarred
him going up the steps and he moaned slightly. Peter cursed the
man's clumsy feet, though not for worlds could he himself have
stirred to help them. The boy's head was bandaged with a towel,
and he looked very limp and white, but he summoned a feeble smile
at sight of his mother. They carried him in and the servants crowded
after in an anxious effort to help.
Peter drove on to the stables and put up Arab. In a few minutes Billy
returned leading the two horses. He was frightened and excited; and
he burst into an account of the accident while he was still half way
down the drive.
"It wasn't my fault," he called. "Miss Ethel said it wasn't my fault. We
met a mowing-machine and Apache bolted. He threw the boy off
against a stone wall, and by the time I reached 'em, Apache was
eating grass in the next field and Master Bobby lying in the ditch
with 'is head cut open."
"I don't want to hear about it," Peter returned shortly. "Put them
horses up and get out."
He himself removed Apache's new saddle and bridle and drove him
with a vicious whack into the stall. Billy took himself off to find a
more appreciative audience, while Peter dropped down on a stool
inside the stable door, and with his chin in his hands sat watching
the house. He saw the nurse fling wide the blinds of Bobby's room
and roll up the shades; he wondered with a choking sensation what
they were doing to the boy that they needed so much light. He saw
Annie come out and hang some towels on the line. The whole aspect
of the place to Peter's sharpened senses wore an air of tragic bustle.
No one came near to tell him how the boy was doing; he had not
the courage to go to the house and ask. He sat dumbly waiting for
something to happen while twilight faded into dusk. One of the
stableboys came to call him to supper and he replied crossly that he
didn't want any supper. Presently he heard a step scrunching on the
gravel, and he looked up to find Annie coming toward him.
"Is—is he dead?" he whispered.
"He's not goin' to die. He's feelin' better now; they've sewed up the
hole in his head. The doctor did it with a thread an' needle just like
you'd sew a dress. He took ten stitches an' Master Bobby bled awful.
He never cried once, though; he just got whiter an' whiter an'
fainted away. Don't feel so bad, Pete, he's goin' to get well."
She laid her hand caressingly on his hair and brushed it back from
his forehead. He caught her hand and held it.
"It's me that's to blame for his gettin' hurt. He won't never speak to
me again."
"Yes, he will; he's wantin' to speak to you now. They sent me out to
fetch you."
"Me?" he asked, shrinking back. "What's he wantin' with me?"
"He's been out of his head an' callin' for you; he won't go to sleep till
he sees you. The doctor said to fetch you in. Come on."
Annie's manner was insistent and Peter rose and followed her.
"Here he is," she whispered, pushing him ahead of her into the
darkened room.
Bobby made a half movement to turn as the door creaked, but a
quick pain shot through his shoulder and he fell back with a little
gasp.
"Take care, Bobby," the nurse warned. "You mustn't move or you will
hurt that bad arm." Her greeting to Peter was stern. "You may stay
five minutes, and mind you don't get him excited!" She bent over
the boy to loosen the bandage about his shoulder.
"You go out," said Bobby, querulously. "I want to see Peter alone."
"Yes, dear," she patted the bedclothes indulgently. "Remember, five
minutes!" she added as she closed the door.
The two left alone stared at each other rather consciously for a
moment. They both felt that the occasion demanded something
heroic in the way of a reconciliation, but it was the natural instinct of
each to fly from sentiment. The sight of Bobby's pale face and
bandaged head, however, had their effect on Peter's already
overwrought nerves.
"I'm a blunderin' fool!" he groaned. "I don't know why I can't never
learn to attend to me own affairs. If I'd told yer father, as was me
dooty, he'd never uv given ye that spotted devil of a horse."
"You aren't to blame, Pete. I guess I was hurt for more punishment
'cause I didn't take the first in the right spirit." He fumbled under his
pillow and drew out the new five-bladed knife. "This is for a
remembrance, and whenever you use it you will think 'it was me that
cured Bobby Carter of telling lies.'"
Peter received the gift with an air of hesitation.
"I don't like to take it," he said, dubiously, "though I have a feelin'
that perhaps I ought, for with five blades to choose from ye'll be
cuttin' yer blamed young throat—I'd hate to be the cause of any
more accidents." He balanced it thoughtfully in his palm. "But I'm
thinkin," he added softly, "that the corkscrew might be doin' as much
damage to me as the five blades to you."
Bobby grinned appreciatively, and held out his uninjured left hand.
"Pete," he said, "if I promise never, never to tell any more lies, will
you promise never, never to use that corkscrew?"
"It's a bargain!" said Peter, grasping the boy's hand. "And I'm glad
that we're friends again."
They stared at each other solemnly, neither thinking of anything
further to add, when Peter suddenly became aware of the ticking of
the clock.
"Holy Saint Patrick!" he ejaculated. "Me five minutes was up five
minutes ago. I must be takin' me leave or that commandin' young
woman will come back and eject me."
He moved toward the door, but paused to throw over his shoulder:
"I'd already promised the same to Annie, so ye needn't be takin' too
much credit to yerself fer me conversion."

MRS. CARTER AS FATE

 IX
MRS. CARTER AS FATE

As the summer wore to an end, the course of affairs between Peter

and Annie became a matter of interested comment among the other
servants. They had all seen Peter recover from many incipient
attacks of love, but this they unanimously diagnosed as the real
thing. Joe and his wife talked the matter over upon his return from
the hospital, and decided that the time had definitely come for the
livery stable; Peter, in all fairness, had served as groom long enough.
They would move out of the coachman's cottage the following
spring, and give the young people a chance. Thus was the way open
for a happy conclusion, and everyone was preparing to dance at the
wedding, except Peter and Annie themselves. They alone were not
certain as to the outcome. Neither was quite comfortably sure that
the other was in earnest, and when it came to the point they were
both a little shy. Annie, with laughing eyes, tempted Peter at every
point, but when he showed a disposition to control matters himself,
she precipitously fled.
The two were standing on the back veranda one moonlight night,
and Annie was engaged in pointing out to Peter the lady in the
moon. Peter was either stubborn or stupid; he frankly declared that
he saw no "loidy," and didn't believe there was one. In her zeal in
the cause of astronomy, Annie unwarily bent her head too near, and
while her eyes were turned to the moon, Peter kissed her. She
slapped him smartly, as a well-brought-up young woman should, and
fled into the house before he could catch her. Peter, strong in his
new-found courage, waited about in the hope that she would
reappear; but she did not, and he finally took himself off to his room
over the carriage-house, where he sat by the window gazing out at
the moonlight for two hours or more before he remembered to go to
bed. The slap had hurt neither him nor his feelings; he liked her the
better for it. She wasn't really mad, he reflected happily, for she had
laughed as she banged the door in his face.
The next morning Peter went about his work with a singing heart
and many a glance toward the kitchen windows. He swashed water
over the stable floor and rubbed down the horses with a mind
happily intent upon what he would say to Annie when he saw her.
About ten o'clock Mrs. Carter ordered the victoria, but as the
carriage horses were at the shop being shod, Joe sent Peter in to
ask if Trixy and the phaeton would do as well.
Peter dropped his sponge and started for the house at exactly the
wrong moment for his future peace of mind. He arrived at the
kitchen door just in time to see the man from the grocery put his
packages on the table and his arms around Annie, and kiss her with
a smack that resounded through the room and would, to Peter's
outraged senses, resound through all time. Annie turned with a
startled cry, and as her gaze fell upon Peter, her face paled before
the look in his eyes. Without a word he whirled about and strode
back to the stables with white lips and clenched fists, and murder in
his heart for the grocer's man. He did not hear what Annie said to
him, nor did he know that she locked herself in her room and cried;
what he did know was that she had been making a fool of him, and
that she flirted with every man who came along, and that that
wasn't the kind of a girl he wanted to do with.
Several days before, as Peter was driving Mr. Lane, who was visiting
at Willowbrook again, and Master Bobby to the village, Annie had
been sweeping the front veranda as they passed, and had thrown a
friendly smile in the direction of the cart. The smile was intended for
Peter, but Mr. Lane had caught it, and had remarked to Bobby:
"That's a deuced pretty maid you've got there."
"Annie's the bulliest maid we ever had," Bobby had returned
appreciatively. "She swipes cake for me when Nora isn't looking."
But Peter had frowned angrily, as he longingly sized up Mr. Lane,
and wished he were not a gentleman so that he could punch him. It
was none of Mr. Lane's business whether Annie was pretty or not.
At that time Annie could do no wrong, and Peter had not thought of
blaming her for Mr. Lane's too-open admiration, but now he
wrathfully accused her of trying to flirt with gentlemen, than which,
in Peter's estimation, she could do no worse. As he could take it out
of neither of them in blood—which his soul thirsted for—he added it
to the grocer's score, and his fingers fairly itched to be at work. The
grocer was just the sort of man that he most enjoyed pummelling—
big and florid, with curling hair, a black moustache, and a dimple in
his chin.
Annie, after her contretemps with the grocer, passed a miserable
day. In vain she tried to get a word with Peter; he was not to be
seen. Billy was the groom who came to the house on all further
errands from the stables. That evening she put on her prettiest frock
and sat for two hours on the top step of the back veranda with her
eyes turned expectantly toward the carriage-house, and then she
went to bed and cried. Had she but known it, Peter was in a vacant
lot back of Paddy Callahan's saloon, blissfully remodelling the
features of the grocer's man.
Annie passed a wakeful night, and the next morning she swallowed
her pride and went to the stables in the hope of seeing Peter alone.
Peter, too, in spite of his victory of the evening, had kept vigil
through the night. He was listlessly currying one of the carriage
horses when he saw Annie leave the house and come slowly down
the walk toward the stables. His heart suddenly leaped to his mouth,
but a moment later he was bending over the horse with his back to
the door, whistling as merrily as though he had not a care in the
world. He heard Annie's hesitating step on the threshold, and he
smiled grimly to himself and whistled the louder.
"Pete, I'm wantin' to speak to you, if ye're not busy."
Peter glanced up with a well-assumed start of surprise. He looked
Annie over, slowly and deliberately, and then turned back to the
horse.
"Aw, but I am busy," he returned. "Lift up!" he added to the horse,
and he solicitously examined her foot.
Annie waited patiently, struggling between a sense of pride which
urged her to go back and never speak to Peter again, and a sense of
shame which told her that she owed him an explanation.
"Pete," she began, and there was a little catch in her voice which
went to Peter's heart; in his effort to resist it and mete out due
punishment for all the misery she had caused him, he was harder
than he otherwise would have been. "Pete, I wanted to be tellin' ye
that it wasn't my fault. He—he niver kissed me before, and I didn't
know he was goin' to then."
Peter shrugged.
"Ye needn't be apologizin' to me. I ain't interested in yer amoors. If
ye wants to be apologizin' to any one go an' do it to his wife."
"His wife?" asked Annie.
"Aye, his wife an' his three childern."
"I didn't know he was married," said Annie, flushing again, "but 'tis
no difference, for it weren't my fault. I niver acted a bit nicer to him
than to anny other man, an' that's the truth."
"Oh, ye're a lovely girl, ye are! Flirtin' around with other women's
husbands, and lettin' every fool that comes along kiss ye if he wants
to."
"Ye needn't talk," cried Annie. "Ye did it yerself, an' ye're no better
than the grocer man."
"An' do ye think I'd a-done it if I hadn't knowed ye was willin?"
Annie backed against the wall, and with flushed cheeks and blazing
eyes, stared at him speechlessly, angry with herself at her
powerlessness to say anything that would hurt him enough. As she
stood there, Master Bobby and Mr. Lane came in on their way to visit
the kennels. Mr. Lane looked curiously from the angry girl to the
nonchalant groom, who had resumed his work, and was softly
whistling under his breath. Master Bobby, being intent only upon
puppies, passed on without noticing the two, but Mr. Lane glanced
back over his shoulder at Annie's pretty flushed face, and paused to
ask:
"My dear girl, has that fellow been annoying you?"
"No, no!" Annie said wildly. "Go away, Mr. Lane, please."
Mr. Lane glanced from one to the other with a laugh. "Ah, I see! A
lovers' quarrel," and he followed Master Bobby.
Peter echoed his laugh, and in a tone which would have justified Mr.
Lane in knocking him down had he heard.
"So ye're his dear girl too, are ye? He's a nice gentleman, he is! Ye
ought to be proud o' him."
Annie straightened herself with her head thrown back.
"Peter Malone," she burst out, "I came here to 'pologize, 'cause,
without meanin' any harm, I thought as I'd hurt yer feelin's an' was
owin' an explanation. I niver had anything to do with that groc'ry
man nor any other man, an' ye know it as true as ye're standin'
there. Instead o' believin' what I say like a gentleman would, ye
insult me worse than anybody's iver done in the whole o' me life, an'
I'll niver speak to ye again as long as I live." She choked down a
sob, and with head erect turned and walked back to the house.
The two had had differences before, but never anything like this.
Peter, his arms dropped limply at his side, stood watching her go,
while the words she had spoken rang in his ears. Suddenly a lump
rose in his throat, and he leaned his head against the horse's neck.
"Lord!" he whispered. "What have I done?"

The week which followed was one of outward indifference and

inward misery to both. Annie mourned when alone, but under the
eyes of the stables she flirted openly and without conscience with
one of the painters who was opportunely engaged in re-staining the
shingle roof of the Jasper house. Peter watched her with a heavy
heart, and formed a brave determination never to think of her again,
and ended by thinking of her every minute of the day. He made one
awkward attempt at reconciliation which was spurned, whereupon
he, too, plunged into a reckless flirtation with Mary, the
chambermaid, who was fat, and every day of thirty-five. As neither
Peter nor Annie had any means of knowing how wretched this
treatment was making the other, they got very little comfort from it.

Annie sat at the kitchen table polishing silver with a sober face. It
was six days since the grocery man's historic visit, and the war
clouds showed no sign of lifting. There was a houseful of company
at Willowbrook, and the work was mercifully distracting. Mary, this
morning, had hung a long row of blankets and curtains on the line to
air, for the sole purpose, Annie knew, of being near the stables.
Peter was visible through the open window, greasing harness in the
carriage-house doorway, and exchanging jocular remarks with Mary.
Annie's eyes were out of doors oftener than upon her work. Nora,
who was sitting on the back veranda shelling peas, remarked on
Peter's newly awakened interest in the chambermaid, but as Annie
did not answer, she very wisely changed the subject.
"I guess that Mr. Lane what's visitin' here has got a heap o' money,"
she called in tentatively.
"I guess he has," Annie assented indifferently.
"He seems to be pretty taken up with Miss Ethel. That was an awful
becomin' pink dress she had on last night. Mrs. Carter would be
pleased all right."
Annie received this remark in silence, but Nora was not to be
discouraged. She felt that this new freak of taciturnity on Annie's
part was defrauding her of her rights. A maid whose duties call her
to the front part of the house is in a position to supply more
accurate gossip than it is given a cook to know, and it is her
business to supply it.
"Mr. Harry would feel awful, havin' growed up with her like," Nora
continued. "He's a sight the best lookin' o' the two, and I'm thinkin'
Miss Ethel knows it. It ud be convenient, too, havin' the places
joined. The Jaspers has got money enough, an' him the only son. I
guess they wouldn't starve if she did marry him. I've always noticed
'tis the people who has the most money as needs the most. I don't
think much o' that Mr. Lane," she added.
Annie suddenly woke up.
"I don't neither. 'Tis too fresh he is."
"That's what I'm thinkin' meself," said Nora, cordially. "An' I guess so
does Mr. Harry. I'm after observin' that he hasn't been around much
since Mr. Lane's been here."
Annie's mind had wandered again. Her own affairs were requiring so
much attention lately that Miss Ethel's were no longer a source of
interest. Out in the stable Peter was proclaiming, in tones calculated
to reach the kitchen, "There's only one girl in this world for me."
Annie's lip quivered slightly as she heard him; a week before she
had laughed at the same song, but as affairs stood now, it was
insulting.
The peas finished, Nora gathered the yellow bowl under her arm and
returned to the kitchen, where she concentrated her attention upon
Annie and the silver.
"I'm thinkin' ye must be in love!" she declared. "Ye've cleaned that
same spoon three times while I've been watchin', an' ye didn't count
the plates right last night for dinner, an' ye forgot to give 'em any
butter for breakfast."
Annie blushed guiltily at this damning array of evidence, and then
she laughed. "If it's in love I am whiniver I forget things, then I must
a-been in love since I was out o' the cradle."
"An' there's him as would be in love with you, if ye'd only act dacent
to him—and I'm not meanin' the painter."
Annie chose to overlook this remark, and Nora's sociability was
suppressed by the entrance of Mrs. Carter.
"We have decided to have a picnic supper at the beach to-night,
Nora," she said. "You will not have to get dinner for anyone but Mr.
Carter."
"Very well, ma'am."
"I am sorry that it happens on your afternoon out, Annie," she
added, turning to the maid, "but I shall need you at the picnic to
help about serving."
"Certainly, ma'am," said Annie. "I don't care about goin' out
anyway."
"We shall start early in the afternoon, but I want you to wait and
help Nora with the sandwiches, and then Peter can drive you out
about six o'clock in the dog-cart."
Annie's face clouded precipitously.
"Please, ma'am," she stammered, "I think—that is, if ye please——"
she hesitated and looked about desperately. "I'm afraid if ye're after
wantin' coffee, I can't make it right. I'm niver sure o' me coffee two
times runnin', and I should hate to be spoilin' it when there's
company. If ye could take Nora instead o' me, ma'am, I could just be
gettin' the lovely dinner for Mr. Carter when he comes."
"Why, Annie," she remonstrated, "you've always made excellent
coffee before, and Nora doesn't wait on the table. Is it because you
want to go out this afternoon? I am sorry, but you will have to wait
until Miss Ethel's guests have gone."
"No, ma'am," said Annie, hastily, "I'm not wantin' the afternoon, an'
it's willin' I am to help Miss Ethel, only—only—will you tell Peter,
ma'am, about the cart?" she finished lamely, "'cause if I tell him he's
likely to be late."
Mrs. Carter passed out of the kitchen door and crossed the lawn
toward the stables, casting meanwhile a sharp eye about the
premises to be sure that all was as it should be. Mary was shaking
blankets with an air of deep absorption; Peter was industriously
cleaning the already clean harness, and Joe could be heard inside
officiously telling Billy to grease the other wheel and be quick about
it. Unless Mrs. Carter approached very quietly indeed, she always
found her servants oblivious to everything but their several duties.
As she drew near the doorway, Peter rose from the harness and
respectfully touched his cap with a very dirty hand, while the
coachman, with a final order over his shoulder to a brow-beaten
stable-boy, came forward hastily, and stood at attention.
"Joe, we are going to have a picnic at the beach this afternoon, and
I want you to have the horses ready at three o'clock. Miss Ethel, Mr.
Lane, and Master Bobby will ride, and you will drive the rest of us in
the waggonette."
"Very well, ma'am," said Joe.
"And Peter," she added, turning to the groom, "I want you to bring
out the supper with Trixy and the dog-cart at five o'clock."
"All right, ma'am," said Peter, saluting.
"Be sure to be on time," she warned. "Stop at the kitchen for Annie
and the hampers promptly at five."
Peter's face suddenly darkened. He drew his mouth into a straight
line, and looked sullenly down at the harness. "Beggin' yer pardon,
ma'am," he mumbled, "I don't think—that is——" He scowled
defiance at Joe, who grinned back appreciatively. "If it's just the
same to ye, ma'am, I'd like to drive the waggonette an' let Joe fetch
the lunch. If I'm to be coachman, ma'am, I'd sort o' like to get used
to me dooties before he goes."
Mrs. Carter was frankly puzzled; she could not imagine what had
suddenly got into her servants this morning. A lady who has a grown
daughter, of some attractions and many admirers, to chaperone,
cannot be expected to keep au courant of her servants' love affairs.
"You have had a month in which to get used to your duties while Joe
was in the hospital; that is sufficient for the present. Joe will drive
the waggonette and you will follow with the supper—I wish you to
help Tom put new netting in the screen-doors this afternoon."
Her tone precluded argument. As soon as she was out of hearing,
Joe remarked softly, "Now, if she'd only said Mary instead of Annie I
'spose——"
"Aw, let up," Peter growled, and he fell to rubbing in the grease with
unnecessary vehemence. His misunderstanding with Annie was a
subject he would stand no fooling about, even from his chief.
At five o'clock, Peter, in a spotless top-hat and shining boots, looking
as stiff as if he were clothed in steel armour, drew up before the
kitchen door and piled the hampers and pails he found on the back
veranda onto the seat beside him. He climbed to the box again with
an air of finality, and gathering his reins together made a feint of
starting.
"Peter!" Nora called from the kitchen window. "Where is it ye're
goin'? Wait for Annie."
"Annie?" Peter looked as if he had never heard the name before.
"Yes, Annie. Did ye think ye was to cook the supper yerself?"
"I didn't think nothin'," said Peter. "Me orders was to stop for the
lunch at five o'clock, an' I done it. If she wants to come along she'll
have to sit on the back seat. I ain't a goin' to change these baskets
again."
Annie appeared in the doorway in time to hear this ungracious
speech; she clambered up to the somewhat uncomfortable
footman's seat in silence, and they drove off back to back, as stiff as
twin ramrods.
The cart rolled along over the smooth roads, past country clubs and
summer cottages, and the only sign either of the two gave of being
alive was an occasional vicious crack of the whip from Peter when
patient little Trixy showed signs of wishing to take a quieter pace. At
such times Annie would instinctively stretch out a deterring hand
and form her mouth as if to say, "Please, Pete, don't whip her; she's
doin' her best," and then suddenly remembering that formidable
vow, would straighten up again and stare ahead with flushed
cheeks.
The beach was five miles away, and there is an element of
ludicrousness in the spectacle of two people in one small dog-cart
riding five miles without speaking. Annie's sense of humour was
keen; it struggled hard with her sense of wrong. She was never an
Indian to cherish vengeance; her anger could be fierce at the
moment, but it rarely lasted. And Peter was sorry for what he had
said, she reminded herself; he had already tried to make up. By the
end of the second mile two dimples appeared in her cheeks. At the
third mile she shut her mouth tight to keep a laugh from escaping.
At the fourth mile she spoke.
"Say, Pete, why don't ye talk to me? Are ye mad?"
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com