An Alternative Approach To Lie Groups and Geometric Structures Ercument H Ortacgil Full Chapter

An Alternative Approach to Lie Groups
and Geometric Structures Ercüment H.

Ortaçgil
Visit to download the full and correct content document:
https://ebookmass.com/product/an-alternative-approach-to-lie-groups-and-geometric-
structures-ercument-h-ortacgil/
an alternative approach to lie
groups and geometric structures
An Alternative Approach to Lie Groups
and Geometric Structures
E RC Ü M E N T H . O RTA Ç G İL
1
3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Ercüment H. Ortaçgil 2018
The moral rights of the author have been asserted
First Edition published in 2018
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2017960678
ISBN 978–0–19–882165–6
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Dedicated to the memory of my beloved mother
for her lifetime service, devotion and love
FOREWORD
When I received the first few chapters of this book from the author, I was both surprised and
intrigued, asking myself why anyone should write a whole book on such a seemingly naive
and trivial subject as local Lie groups. The next day, however, I changed my mind entirely.
Since the late 1930s, we have suffered from a rather nasty affliction, the “Syndrome of
Globality,” under the leadership of Général Nicolas Bourbaki (as played mainly by André
Weil, who never raised his sword in Sevastopol at the end of the Crimean War)—although
I have always thought that the name of Eudoxus of Cnidus would have been a much better
choice for the main actor in this saga. It has seemed as if the only acceptable approach to any
mathematical problem or setting is a global one. Of course, this has its charm and beauty,
a global Lie group, for instance, being a much more expressive structure than any local
counterpart, even possessing a topology and a differentiable structure and no longer being
just some finite set of numerical parameters difficult to look for and pinpoint—Humphrey
Bogart versus Charlie Chaplin. Then I suddenly and unexpectedly noticed in these chapters
an aspect of a most contradictory nature, namely, the definition of a global object called a
local Lie group. With the help of parallelism and a covariant derivative, the author of this
book manages to produce a very useful global object that transcends locality. It not only
helps by providing a better understanding and greater precision to much of Lie and Cartan’s
work, but also fits like a glove with many practical applications in physics and technology. It
cannot be denied that we live and thrive in a completely local spatial context and hence must
adapt our thoughts to this. For many years, the present author has had a great admiration for
global objects and for their harmonious interlacing and coexistence, the most spectacular
example being Leonardo da Vinci’s Last Supper, where renaissance art came to coexist for the
first time with projective geometry (perspective). I should also mention that other authors
have recently focused their attention on locality within a global context, an example being
Ronald Brown’s approach in his writings on the philosophy of mathematics.
As for the book itself, it is a very agreeable read, containing excellent ideas, with thorough
argumentation concerning its purposes, presented mostly through complete proofs. As a
final suggestion, I believe that similar consideration should be given to the infinite groups of
transformations of Lie and Cartan that nowadays show up in the disguise of Lie groupoids
of higher-order Ehresmannian jets or Lie pseudogroups of local transformations in the
terminology of Matsushima and Kuranishi.
Antonio Kumpera
A CK N O W L E D G M E N TS
I am grateful to Antonio Kumpera, who became my distant teacher with his book years ago
and my close friend later, for kindly accepting to write the foreword upon my request and
for his many suggestions to improve my exposition.
I am also grateful to Peter Olver for his encouraging taps on my back over the years.
Without his support, this book would not be in print.
I express my hearty thanks to Anthony Blaom for reading the whole book and making
valuable suggestions. Over the years, he generously shared his ideas and preprints with me,
from which I benefited greatly.
While I was stumbling with the first ideas for this book many years ago, some people
gave me invaluable support and protection, and I feel a particular debt of gratitude to
them. They are my venerable teacher Cahit Arf, Ergün Toğrol (Rector of BU, 1982–92),
Yalçın Koç (Dean of Arts and Sciences, 1990–92), and my friends Alp Eden (Head of
the Mathematics Department, 1999–2001) and Teoman Turgut (Head of the Feza Gürsey
Institute, 2005–2011).
Gregor Weingart pointed out to me the mistake mentioned in Chapter 16 and Jan
Draisma helped me with its correction. My friends Ender Abadoğlu, Yılmaz Akyıldız, Yücel
and Aydan Eğecioğlu, Alihan Neşeliler, İlhan Özemek and George Simpson made various
suggestions and corrections. I thank them all very much.
I also thank Jean-François Pommaret for our intense correspondence in 1994, from which
I learned a great deal about jets.
Last, but not least, I express my hearty thanks to Daniel Taber, my editor at OUP, for
his friendly and professional assistance during the process of evaluation and preparation
of this book.
Ercüment Ortaçgil
ortacgile@gmail.com
Bodrum (Halicarnassus) 2017
CO N T E N TS
PART I Fundamental Concepts
0 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1 Parallelizable Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 The Nonlinear Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Local Lie Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4 The Centralizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5 ε-Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6 The Linear Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7 The Structure Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
PART II Some Consequences

8 The Nonlinear Spencer Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
9 Deformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
10 The de Rham Cohomology of an LLG . . . . . . . . . . . . . . . . . . . . . . . . . 77
11 The Linear Spencer Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
12 The Secondary Characteristic Classes . . . . . . . . . . . . . . . . . . . . . . . . . 93
13 The Homogeneous Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
14 The Van Est Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
15 The Symmetry Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
PART III How to Generalize?

16 Klein Geometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
17 The Universal Jet Groupoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
18 Embeddings of Klein Geometries into Universal Jet Groupoids . . . . . 161
xii | contents
19 The Definition of a Prehomogeneous Geometry (PHG) . . . . . . . . . . . 171

20 Curvature and Generalized PHGs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Appendix: Torsion-Free Connections 201

References 209
Index 213
PA RT I
Fundamental Concepts
Everything should be made as simple as possible

but not simpler
Einstein
0
• • • • • • •
Introduction
The theory of Lie groups is one of the most important mathematical themes of the last
century and belongs at the center of modern differential geometry. Starting with their
discovery by Sophus Lie in 1872, Lie groups have evolved into a vast subject with links
to many branches of mathematics and with many applications in various fields. There are
dozens of books (possibly more) on Lie groups and Lie algebras, with varying aims and
scopes. When an author writes a book, the most serious question he or she faces is “Why
another book?” This question becomes more challenging if the subject is classical and well
known like Lie groups. In our case, however, this question has an easy answer: To our
knowledge, none of the previous books (including those on transformation groups) have
approached the subject from the standpoint of this one. Furthermore, we believe that this
new approach is more in accordance with the original work of Sophus Lie and Felix Klein
than the present modern theory.
We would like to explain the unique feature of this book briefly in somewhat technical
terms. Let us first recall the well-known question: When is a smooth manifold M a Lie
group? More precisely, is it possible to define a structure on M in such a way that when the
curvature R of this structure vanishes, M becomes “locally a Lie group?” This question has
a well-known answer: For a given M, let G be a Lie group with dim G = dim M and with
Lie algebra g. The structure we seek is a g-valued 1-form ω on M with trivial kernel, called
the Maurer–Cartan form, and the curvature R is defined using the exterior derivative dω.
In this classical approach, the key fact is that the Lie group G is fixed beforehand as a model.
Naturally, the question arises as to whether it is possible to dispense with the model. It is
a remarkable fact that the answer is affirmative, and indeed such a model-free approach to
geometric structures has been proposed by Blaom [B1, B2] in a very general framework
and has also been promoted independently [AO1, AO2]. In the above simplest case,
the problem involves starting with some structure on M, examples of which abound in
geometry, and then defining its curvature R in such a way that R = 0 if and only if M
“becomes locally some Lie group G.” We carry out this program in detail in Part I of this
book. It turns out that the structure we seek is an absolute parallelism ε on M, which we
study in detail in Chapter 1. In Chapter 2, we define the curvature R of the parallelizable
manifold (M, ε), and in Chapter 3, we call (M, ε) a local Lie group (LLG) if R = 0.
This is a global concept, in contrast to the common modern terminology of “local Lie
An Alternative Approach to Lie Groups and Geometric Structures, Ercüment H. Ortaçgil (2018).
© Ercüment H. Ortaçgil 2018. Published 2018 by Oxford University Press.
4 | fundamental concepts
group,” which is a local concept referring to the neighborhood of the identity of a global
Lie group. An LLG determines a transitive pseudogroup of local diffeomorphisms on M.
If M is compact and simply connected, then these local diffeomorphisms extend uniquely
to global diffeomorphisms, giving a transformation group G that acts simply transitively
on M. Taking the centralizer C( G ) of G inside Diff(M), we show in Chapter 4 that G
and C( G ) define a Lie group structure on M modulo two choices: (1) a base point that
serves as the identity; (2) whether G (or C( G )) should be left (or right) translations.
Therefore, a Lie group is a special (globalizable) LLG. In the words of Olver [O4], “This
fact reinstates the paradigm of local to global to its historical record,” as envisaged by him
in [O1]. The crucial fact here is that the Lie group M is not fixed beforehand, that is, this
new approach is “model-free.” In Chapters 5 and 6, we show that this construction can
be carried out equivalently on an infinitesimal level by replacing point transformations by
vector fields (called infinitesimal point transformations in the old literature). As a surprising
consequence, we show that the well-known tensor calculus, which emerged as an attempt to
formulate Riemann’s revolutionary ideas, emerges also from Lie theory, however with quite
different interpretations of the concepts of torsion and curvature. In Chapter 7, we compare
this new approach with the classical one via the Maurer–Cartan form.
The purpose of Part II is to show the usefulness of the theory developed in Part I by
establishing some surprising relations, raising some far-reaching questions, and proposing
new points of view. For instance, in Chapter 9, we define the analog of the Kodaira–Spencer
map well known from the deformation theory of complex structures. In Chapter 11, we
establish a concrete but intriguing relation between the cohomology groups of tensorial
and trivial representations. In Chapter 12, we recast Chern–Simons theory in a completely
different setup. In Chapter 13, we construct a geometric flow that is more elementary than
the Ricci flow and, we believe, will simplify the present proof of the Poincaré Conjecture.
In Chapter 14, we give a simple proof of the Van Est theorem using the seemingly unrelated
idea of horizontal cohomology of a PDE due to Vinogradov [Vn, V]. In Chapter 15, we
propose an algebraic prolongation theory of a Klein geometry in analogy with the well-
known prolongation theory of a single linear Lie algebra developed by Guillemin, Singer,
and Sternberg [KN2, SS]. This new prolongation theory, in which the original idea of
taking successive normalizers is due to Pommaret [P1], is considerably more intuitive and
geometric than the familiar prolongation theory. Chapters 8–11 and 15 form a logical
sequence, while 12–14 are independent of the rest, although all of these chapters require
a good understanding of Part I.
Our main purpose in Part III is to generalize Part I in the following way: Define a
geometric structure P and its curvature R on M in such a way that R = 0 if and only
if P is “locally homogeneous” in the sense that any p, q ∈ M have neighborhoods U, V
such that P|U can be “identified” with P|V . Naturally, we require this identification to be
induced by some local diffeomorphism f : U → V. Therefore, if R = 0, these identifica-
tions define a transitive pseudogroup on M as in Part I. Furthermore, if these local diffeomor-
phisms extend to global diffeomorphisms, then we get a transitive (but not necessarily simply
transitive as in Part I) transformation group of M, turning M into a homogeneous space
M = G/H. We fix beforehand only H and dim G = dim P = dim H + dim M, but not G
or its Lie algebra g; that is, our approach is model-free. Now P is an absolute parallelism if
introduction | 5
and only if H is trivial, in which case we recover Part I. At this stage, it is instructive to look at
a Riemannian structure. which is locally homogeneous in our sense if and only if the metric
has constant curvature. Since we require R = 0 to be equivalent to local homogeneity,
just from the outset we discover the surprising (and also annoying) fact that the curvature
R that we seek cannot be the Riemann curvature tensor! The uniformization theorem
tells us that a Riemannian structure with R = 0 can be locally homogeneous in only 3
distinct ways. For absolute parallelism, this number is ∞, because we do not impose any
condition on the emerging Lie group beforehand, even though such conditions are imposed
by the underlying structure of M as in the case of the Poincaré Conjecture in Chapter 13.
What is this number for some geometric structure P yet to be defined? A possible answer,
M = G/H, gives a representation of h on V = g/h with dimension equal to dim M, where
h and g are the Lie algebras of H and G. This fact gives a hint of the level of difficulty of this
question.
As far as we know, there exist essentially three approaches to geometric structures on
manifolds:
1. The classical approach via Cartan connections (which generalize the Maurer–
Cartan form) on principal bundles (see [CS] and references therein).
2. The more recent approach using groupoids induced by some auxiliary bundles on
M [CSa].
3. The intrinsic approach via jet groupoids on M [KS, Ku1, Ly, P1, P2].
In all these approaches, the model is fixed, and in the third, curvature is replaced by the
concept of formal integrability. We start our search for a “good” definition following the third
approach through the somewhat lengthy and leisurely but very instructive tour of Chapters
16–19. However, our ambitious plan of generalizing Part I to arbitrary geometric structures
in a model-free way, which works perfectly well for affine and Riemannian structures, comes
to an abrupt halt in Chapter 19 with the unexpected discovery that a projective structure
is always flat. The reason is that we need a technical assumption to define curvature, and
this assumption forces flatness of a projective structure. We formulate the final definition
in Chapter 20, which is inspired by an idea communicated to us by Blaom. This definition
depends on the ideas developed in Part I and the concept of a foliation that is compatible
with the absolute parallelism. The reader who feels comfortable with Part I can skip directly
to this definition in Chapter 20. However, there is an immense amount of structure behind
this definition, and it is not really possible to appreciate it without the motivational tour
of Chapters 16–19. We finish Chapter 20 by defining seemingly new characteristic classes
depending on higher-order jets using Chern–Simons forms, but the question of their
nontriviality remains open . . . and this short book ends, unfortunately, at a point where it
should actually start.
The prerequisites for this book are quite modest: We assume that the reader is comfort-
able with the definition of a smooth manifold and some basic structures on it, like vector
fields, Lie derivative, differential forms, tensors, etc., and we otherwise start from scratch
in Chapter 1. The only technical tool that we use in Part I is the well-known existence
and uniqueness theorem for first-order systems of PDEs with initial conditions, which is
proved, for instance, in the Appendix of [L]. We use some basic topological notions like
simple connectivity and the rudiments of covering spaces in Chapter 3. Therefore, a good
undergraduate student should have no difficulty in understanding Part I in depth. The level
of abstraction increases slightly in Part II, which requires more active involvement on the
part of the reader, but we believe that an undergraduate student with persistence and math-
ematical maturity can also follow Part II if he or she is willing to take a few facts for granted.
However, Part III is the beginning of an adventure in unknown territory and requires a
greater degree of sophistication. Our language here becomes more intuitive and descriptive,
and some unproved assertions are not entirely trivial, although their detailed proofs would
distract us from our main goal, which is to motivate the definition in Chapter 20. We should
remark here that the main purpose of this book is to attract the attention of curious and
courageous mathematical minds, because what we are doing here is no more than scratching
at the topsoil, and much remains to be done to build this new theory on solid ground.
The process of maturation of some ideas in this book was an adventurous, joyful, and
even ecstatic journey that spanned a period of more than 30 years, but was at times also
deeply solitary, painful, and even frustrating. We would like to remind the reader that the
alternative approach adopted in this book is unfortunately, but inescapably, not in much
accordance with the mainstream. From our experience over the years, we happened to learn
that clinging to our prior knowledge may obstruct the light of the new. So we kindly request
the reader to temporarily put aside his or her existing understanding (and hence conditioned
views) of some classical concepts (like Lie group, curvature, and representation) and read
this book with innocent eyes.
1
• • • • • • •
Parallelizable Manifolds
In this book, we always assume that M is a smooth manifold; that is, M is a topological
manifold (Hausdorff with a countable base) with an atlas whose transition functions are
differentiable up to any order. Since we will deal with certain integrability conditions, we
also assume that M is connected and n = dim M ≥ 2.
We recall that M is called parallelizable if it admits n independent vector fields. In more
detail, there exist some vector fields X1 , . . . , Xn whose values X1 (p), . . . , Xn (p) form a basis
of the tangent space Tp M for all p ∈ M. The purpose of this chapter is to formulate an
equivalent definition of parallelizability. As we proceed in this book, it will gradually become
clear that this equivalent formulation has some remarkable consequences.
So, let M be a smooth manifold and (a, b) and (c, d) be two elements of M × M. If b = c,
then we can operate on these two pairs according to the rule
def
(b, d) ◦ (a, b) = (a, d) (1.1)
The set M×M with the partial operation ◦ defined by (1.1) is called the pair groupoid on M.
Clearly, the pair groupoid can be defined on any set. For reasons to be made clear below, if M
is a smooth manifold, we call the pair (a, b) a 0-arrow from a to b. We call a the source and b
the target of the 0-arrow (a, b). The formula (1.1) states that two 0-arrows can be composed
if (and only if) the target of the first is equal to the source of the second, and the result is a
0-arrow from the source of the first to the target of the second. In this way, (1.1) assumes a
very intuitive geometric meaning.
A word of caution with our notation: In many works on groupoids, the pair (b, a) denotes
a 0-arrow from a to b, that is, the second argument denotes the source and the first argument
denotes the target. With this notation, (1.1) is written as
(c, b) ◦ (b, a) = (c, a) (1.2)
Both notations have their advantages and disadvantages, but we prefer (1.1). Note that (1.1)
gives
(b, a) ◦ (a, b) = (a, a) (1.3)
We define
(a, b)−1 = (b, a)

def
(1.4)
so that (a, b)−1 ◦ (a, b) = (a, a) and (a, b) ◦ (a, b)−1 = (b, b). Given the 0-arrow (a, b)
from a to b, we choose two coordinate systems (U, x) and (V, y) around a and b, respectively.
Suppose a = (ai ) and b = (bi ) in these coordinates, with 1 ≤ i ≤ n = dim M. Now (a, b)
has the coordinate representation (ai , bi ). Letting a and b vary inside U and V, we see that
any 0-arrow with source in U and target in V has the unique coordinate representation
(xi , yi ). Therefore, choosing coordinates around a, b, d, we can rewrite (1.1) as
(yi , zi ) ◦ (xi , yi ) = (xi , zi ) (1.5)
and (1.4) as
(xi , yi )−1 = (yi , xi ) (1.6)
in coordinates.
We observe that the coordinate expression (ai , bi ) may have another meaning: It may also
represent a 0-arrow whose source and target are in the same coordinate system; that is, it
may represent (a, b) where a, b ∈ (U, xi ). In this book, we will often let the target approach
the source as a limit in certain coordinate expressions, and this second interpretation will be
assumed in such arguments. These remarks apply also to the local expressions for 1-arrows
to be defined below.
We let U0 denote the set of all 0-arrows on M, so U0 = M × M. Now U0 is the first
element of a sequence of groupoids Uk , k ≥ 0 on M, and the definition of Uk for k ≥ 1
needs a smooth structure on M. In the first two parts of this book, we will need only U0 and
U1 , except in Chapter 15.
Before we proceed to the definition of U1 , we first define a 1-arrow on M and, as expected,
U1 as a set will be the collection of all 1-arrows on M. Furthermore, U1 will have the extra
structure of a smooth manifold like U0 = M × M.
In order to define a 1-arrow from a to b, we fix (a, b) ∈ U0 and consider all the local
diffeomorphisms of M that map a to b. We fix two coordinate systems (U, x) and (V, y)
around a and b, respectively, where a = (ai ) and b = (bi ) as above. Let f and g be two such
local diffeomorphisms. By choosing U sufficiently small, we may assume that f , g are both
defined on U. We declare f ∼ g if, in addition to f (a) = g(a), we also have
∂f i ∂g i
(a) = (a) (1.7)
∂x j ∂x j
Clearly, ∼ defined by (1.7) is an equivalence relation on the set of all local diffeomorphisms
of M that map a to b. We claim that this equivalence relation is independent of the choice
of the coordinate systems (U, x) and (V, y) around a and b. Indeed, let f be a local diffeo-
morphism that maps (U, x) onto (V, y) and let its derivative have the local representation
parallelizable manifolds | 9
(∂yi /∂x j ) with respect to these coordinates. Now we apply a coordinate change ( y) → (w).
The chain rule shows that the derivative (∂wi /∂x j ) of f with respect to the coordinates (x)
and (w) is given by
∂wi ∂w i ∂ya
= (1.8)
∂x j ∂y a ∂x j
We have used the Einstein summation convention in (1.8) and we will use it throughout
this book. Similarly, a coordinate change (x) → (z) transforms (∂yi /∂x j ) according to the
formula
∂yi ∂y i ∂xa
= (1.9)
∂z j ∂x a ∂z j
In particular, if the above two coordinate changes are applied simultaneously, we obtain
the transformation rule
∂wi ∂wi ∂ya ∂xb

= a b j (1.10)
∂z j ∂y ∂x ∂z
Now (1.10) shows that the equivalence relation defined by (1.7) is independent of coordi-
nates, because if (1.7) holds in some coordinates, then it holds in all coordinates according
to (1.10). A 1-arrow from a to b is simply a geometric name for an equivalence class. So, we
make the following important definition:
Definition 1.1 A 1-arrow from a to b is an equivalence class defined by the equivalence
relation (1.7).
If f is a representative for a 1-arrow from a to b, it is standard to denote this 1-arrow by
j1 ( f )a and call it the 1-jet of f with the source at a and the target at b = f (a). To be consistent
with our later notation, we will write j1 ( f )a and sometimes also j1 ( f )a,b , where b = f (a), in
order to emphasize the target b, even though this is clear from the notation j1 ( f )a . Of course,
we may have j1 ( f )a = j1 ( g)a but f = g as local diffeomorphisms; that is, an equivalence
class has many different representatives. Indeed, f and g may have different derivatives at a
of order k for some k ≥ 2, so f = g even though (1.7) may hold, so they will represent the
same 1-arrow. In fact, f and g may have the same derivatives of all orders at a but we may still
have f = g as local diffeomorphisms, but this can occur only in the smooth category and not
in the analytic category. These preliminary observations provide some intuition regarding
the definition of a k-arrow from a to b that we will return to in Chapter 15 and Part III. For
instance, it is not difficult at this stage to guess that a 2-arrow from a to b will be defined as
an equivalence class on the set of all the local diffeomorphisms that map a to b as follows:
We declare f ∼ g if, in addition to f (a) = g(a) and (∂f i /∂x j )(a) = (∂g i /∂x j )(a), we also
have (∂ 2 f i /∂x j ∂yk )(a) = (∂ 2 g i /∂x j ∂yk )(a). However, it becomes more and more tedious
to show the independence of coordinates directly, and one introduces the jet groups Gk (n)
for this and other purposes, as in Part III.
Having defined a 1-arrow from a to b, we let U1a,b denote the set of all the 1-arrows from
def
a to b, and we define the set U1 = ∪a,b ∈ M U1a,b . Then, U1 is the set of all 1-arrows on M.
We have the obvious projection map
π : U1 −→ U0 (1.11)
which projects a 1-arrow from a to b to the 0-arrow (a, b) from a to b.

To complete the definition of U1 , we need to do two more things: First, we will endow
U1 with the structure of a smooth manifold and, second, we will define a partial composition
as well as an inversion on the elements of U1 in a way compatible with the projection
(1.11). In more abstract terms, we will turn U1 into a differentiable groupoid in such a way
that (1.11) will become a homomorphism of groupoids. However, we will not need any
prior knowledge of groupoids and our elementary treatment will be self-contained. We refer
the interested reader to [MK1] and [MK2] for the general theory of groupoids and also to
the excellent expository articles [Wn2] and [Ku2].
So, let j1 ( f )a,b be a 1-arrow from a to b = f (a),
(U, x) and (V, y) be co-ordinate systems
i i
around a and b. We set f j = (∂yi /∂x j )(a), so f j is an invertible matrix and therefore an
element of GL(n, R). It follows that any 1-arrow from a to b defines
an element of GL(n, R)
i
once we fix the coordinates around a, b. Conversely, for any f j ∈ GL(n, R), there exists
i
some local diffeomorphism f satisfying f (a) = b and f j = (∂yi /∂x j )(a): For instance,
i
yi = bi + f j (x j − a j ) is such a local diffeomorphism. Now, let W be an open neighborhood
i
of f j = (∂yi /∂x j )(a) in GL(n, R). We declare the set U × V × W to be a coordinate system
i

around j1 ( f )a and the triplet ai , bi , f j to be the coordinates of j1 ( f )a in this system. It
follows that any 1-arrow with source in U and target in V is uniquely parametrized by some
(xi , yi , fji ) ∈ U × V × W.
Since we will often work with local coordinates in this book, we will identify a 1-arrow
with its coordinate representation (xi , yi , fji ) and will not explicitly refer to the coordinate
maps that perform this identification, as we did also for 0-arrows. This identification will
also allow us to avoid keeping track of the intersections of some open sets. Keeping this in
mind, notice that (1.11) is given by

xi , yi , fji −→ xi , yi (1.12)
in coordinates.
The relation (1.10) together with the local coordinates defined above imply that U1 is
a smooth manifold of dimension 2n + n2 = 2 dim M + dim GL(n, R). Indeed, the two
coordinate changes (x) → (z) and ( y) → (w) transform the local representation ( xi , yi , fji )
into (zi , wi , gji = (∂wi /∂yb )fab (∂xa /∂z j )) according to (1.10). Consequently, since (zi ) and
(wi ) depend smoothly on (xi ) and (yi ), respectively, and since (gji ) depends smoothly on
(xi ), (yi ), and ( fji ), we conclude that the transition functions between two such coordinates
systems are smooth and therefore that U1 is a smooth manifold. Furthermore (1.12) shows
that (1.11) is a smooth map.
It remains to define the partial composition and inversion. Consider therefore the 1-arrow
j1 ( f )a,b from a to b = f (a) with the representative f , and the 1-arrow j1 ( g)b,c from b to
g(b) = g(f (a)) = c with the representative g. We define the composition of these 1-arrows
by the formula
def
j1 ( g)b,c ◦ j1 ( f )a,b = j1 (g ◦ f )a,c (1.13)
Note that ◦ on the left-hand side of (1.13) refers to the defined composition, whereas on
the right-hand side it refers to the ordinary composition of two local diffeomorphisms. Also,
(1.13) defines the composition of two equivalence classes in terms of some representatives,
and so we must check that (1.13) is well-defined, but this is immediate using (1.10).
Therefore, we deduce from (1.13) the composition map
U1b,c × U1a,b −→ U1a,c (1.14)
Choosing coordinates around a, b, c and using the chain rule, (1.13) becomes

yi , zi , gji ◦ xi , yi , fji = xi , zi , gai fja (1.15)
We note that ◦ in (1.13) is a smooth operation. This means that the values on the right-hand
side of (1.15) depend smoothly on the values on the left-hand side.

Let I be the identity map of M and consider j1 (I)a , which is ai , ai , δji in terms of some
coordinates around a. Here δji is the Kronecker delta: δji = 1 if i = j and δji = 0 if i = j. With
an abuse of notation, we will simply write j1 (I)a = (ai , ai , δji ), such notation being used
henceforth. We now define
−1 f (a)
= j1 f −1
def
j1 ( f )a (1.16)
which is easily seen to be well-defined. So, the inverse of a 1-arrow from a to b is a 1-arrow
from b to a. Now, (1.16) defines the inversion map
U1a,b −→ U1b,a (1.17)
We have
−1 f (a)
j1 ( f )a ◦ j1 ( f )a = j1 f −1 ◦ j1 ( f )a
−1 a
= j1 f ◦ f
= j1 (I)a (1.18)
and similarly
−1
j1 ( f )a ◦ j1 ( f )a = j1 (I)f (a) (1.19)
The local formulas for (1.16) and (1.18) now become

−1
xi , yi , fji = yi , xi , gji , fai gja = gai fja = δji (1.20)
and
−1
xi , yi , fji ◦ xi , yi , fji = yi , xi , gji ◦ xi , yi , fji

= xi , xi , δji (1.21)
Clearly, the inversion (1.17) is a smooth map by (1.20). Also, (1.5) and (1.15) show that
the projection π in (1.11) preserves the composition: Composing and then projecting is
the same as projecting first and then composing. Similarly, (1.6) and (1.20) show that π
preserves inversion: Taking the inverse and then projecting is the same as projecting first
and then taking the inverse.
It is worth paying special attention to the structure of the 1-arrows with the same source
and target, namely, U1a,a for some a ∈ M. Now, (1.13) and (1.16) show that U1a,a is a group.
A choice of coordinates around a identifies U1a,a with the coordinate expressions (ai , ai , fji ),
which in turn can be identified with the group GL(n, R). Note, however, that there is no
such canonical identification.
Having completed the construction of the groupoids U0 and U1 as well as the smooth
projection homomorphism π in (1.11), we now come to a fundamental point on which this
whole book rests.
Suppose there exists a smooth map
ε : U0 −→ U1 (1.22)
satisfying
π ◦ ε = I U0 (1.23)
where I U0 denotes the identity map on U0 . In other words, ε assigns to any 0-arrow from a
to b a unique 1-arrow from a to b for any a, b ∈ M and this assignment is smooth. We further
assume that ε preserves the composition and inversion of arrows; that is, we require ε to
be a homomorphism of groupoids. We call ε a splitting. If M admits a splitting ε, denoted
by (M, ε), then ε( U0 ) ⊂ U1 is a subgroupoid with the obvious meaning. It will turn out in
Chapter 2 that ε( U0 ) is actually a globally defined first-order nonlinear PDE. As preparation
for understanding this PDE, we express ε in coordinates. Suppose ε(a, b) = j1 ( f )a with
f (a) = b. Choosing coordinates around a, b we can write

ε xi , yi = xi , yi , εji (x, y) (1.24)

Note that εji x, y is a shorthand notation for εji x1 , . . . , xn ; y1 , . . . , yn . The smoothness of
ε means that its components εji are smooth functions of their arguments. Since ε preserves
composition, (1.24) and (1.15) imply

yi , zi , εji (y, z) ◦ xi , yi , εji (x, y) = xi , zi , εai ( y, z)εja (x, y)

= xi , zi , εji (x, z) (1.25)
and the second equality holds if and only if we have

εji (x, z) = εai ( y, z)εja (x, y) (1.26)
Clearly, ε(a, a) = j1 (I)a for all a ∈ M: We apply ε to the identity (a, b) ◦ (a, a) = (a, b) and
cancel ε(a, b). So, we have the local formula
εji (x, x) = δji (1.27)
for all x ∈ U. Setting z = x in (1.26) and using (1.27), we also get the local formula
εai (y, x)εja (x, y) = δji (1.28)
So, (1.27) and (1.28) are consequences of (1.26). Equivalently, if ε preserves composition,
then it must also preserve inversion. We remark here that (1.26)–(1.28) are local expressions
for global formulas.
It is a truly remarkable fact that everything we do in Parts I and II of this book (without
exception!) is a direct consequence of (1.26), and it will then become a real challenge to
generalize (1.26) in the right way in Part III. The surprise that we will discover in Chapter 20
is that the right way to generalize (1.26) is to add some extra structure to it, this extra
structure being vacuous in the case of absolute parallelism.
However, the existence of a splitting puts a very strong condition on M. To understand
this condition, let us recall the definition of an absolute parallelism on M. As we have already
remarked, suppose M admits n smooth vector fields ξ(1) , . . . , ξ(n) on M, where n = dim M,
such that their values ξ(1) (p), . . . , ξ(n) (p) at p form a basis of the tangent space Tp M for all
p ∈ M. In this case, we say that these vector fields define an absolute parallelism on M, or that
M is parallelized by these vector fields. Without any reference to the vector fields, we also
say that M is parallelizable. Lie groups are the most important examples of parallelizable
manifolds. In fact, a Lie group can be parallelized in two ways: by left translations and by
right translations. These two parallelizations can be quite “different” if the Lie group is not
abelian. We will see in the next two chapters that Lie groups are very special parallelizable
manifolds. The key fact here is that we do not actually need the left (right) translations to
parallelize a Lie group, but only the 1-arrows of these translations. As another example, any
orientable 3-manifold is known to be parallelizable. This fact will be crucial in Chapter 13.
We will see in Chapter 20 that the total spaces of some principal bundles relevant for our
purposes are also parallelizable. Clearly, M can be parallelized in different ways, in the same
way as it can admit different splittings. We now have the following proposition:
Proposition 1.2 The following are equivalent:

(i) M admits a splitting;
(ii) M is parallelizable.
The proof is simple and depends on an equivalent formulation of a 1-arrow from p to q,
which is clarified by the following trivial lemma:
Lemma 1.3 There is a 1–1 correspondence between the following objects:
(i) 1-arrows from p to q;
(ii) linear isomorphisms Tp M → Tq M.
We will give the proof in some detail in order to fix our notation. Let j1 ( f )p,q = (pi , qi , fji )
be a 1-arrow from p to q and let ξp = ξ a (∂/∂xa )|p be a tangent vector at p that we write in
short as ξp = (ξ i ). We define the components (ηi ) of an object ηq by
def
ηi = fai ξ a (1.29)
Using the transformation rules for (pi , qi , fji ) and (ξ i ) under coordinate changes, we easily
check that (ηi ) are the components of a tangent vector at q, so ηq = ηa (∂/∂ya )|q . Since
the map ξp → ηq defined by (1.29) is linear, we conclude that j1 ( f )p,q defines a linear map
(using the same notation) j1 ( f )p,q : Tp M → Tq M, which is an isomorphism since the matrix
(fji ) is invertible. Note that (1.29) is equivalent to j1 ( f )p,q (∂/∂x j )|p = fja (∂/∂ya )|q , and
therefore (fji ) is the matrix of j1 ( f )p,q with respect to the coordinate bases (∂/∂x j )|p and
(∂/∂y j )|q of Tp M and Tq M, respectively. Conversely, given an isomorphism Tp M → Tq M,
let (fji ) be its matrix with respect to the coordinate basis, and it follows easily that (pi , qi , fji )
is a 1-arrow from p to q.
Now we can easily prove Proposition 1.2. Assume that ε is a splitting. Fixing some
basepoint e ∈ M and choosing a basis ξ(1) (e), . . . , ξ(n) (e) of Te M, we know, by Lemma 1.3,
that ε(e, x) defines a linear isomorphism ε(e, x) : Te M → Tx M for any x ∈ M. We define
def
the n vector fields ξ(1) , . . . , ξ(n) on M by the formula ξ(i) (x) = ε(e, x)(ξ(i) (e)), 1 ≤ i ≤ n.
These vector fields are smooth and define an absolute parallelism on M.
Conversely, assume that the vector fields ξ(1) , . . . , ξ(n) parallelize M. For p, q ∈ M, we
define a linear isomorphism Ep,q : Tp M → Tp M by sending the basis ξ(1) (p), . . . , ξ(n) (p)
of Tp M to the bases ξ(1) (q), . . . , ξ(n) (q) of Tq M, respectively. By Lemma 1.3, Ep,q defines a
1-arrow ε(p, q) from p to q. Clearly, we have Eq,r ◦ Ep,q = Ep,r , and therefore ε is a splitting.
Note that the composition of these linear maps corresponds to the multiplication of their
matrices, thus giving (1.26).
A digression: Our frequent use of local coordinates throughout this book may seem old
fashioned and therefore pointless. However, we hope we will be able to convince the reader
in this book that sometimes (though not always!) the use of local coordinates may be very
rewarding indeed.
There are two objects that emerge naturally from the groupoid U1 and play an important
role in the theory. The idea of the first is already hidden in the proof of Proposition 1.2:
def
We fix some base point e ∈ M and define U1e,• = ∪x ∈ M U1e,x ; that is, U1e,• is the set of all
1-arrows spreading out from e. The projection π : U1e,• → M maps the set U1e,x onto x. The
local representations of 1-arrows shows that π : U1e,• → M is a smooth and locally trivial
fiber bundle with fibers π −1 (x) = U1e,x . Further, the composition U1e,x × U1e,e → U1e,x of
1-arrows provides a smooth right action of the group U1e,e on the fiber U1e,x , and
π : U1e,• → M is a right principal bundle with structure group U1e,e . We recall that a choice
of coordinates in a neighborhood of e identifies the group U1e,e with GL(n, R). Fixing such
a coordinate system once and for all, we obtain a principal GL(n, R)-bundle. Let us call
π : U1e,• → M the (first-order) frame bundle of M with base point e. In much the same
way, we can define the left principal bundle π : U1•,e → M with structure group U1e,e , where
U1•,e = ∪x ∈ M U1x,e . We call this bundle the coframe bundle of M with base point e. We note
def
p,q e,q p,e
that these principal bundles determine the groupoid U1 , because U1 = U1 ◦ U1 for all
e,•
p, q ∈ M. The splitting ε determines a global section of U1 → M, namely, x → ε(e, x), so
this principal bundle is trivial. Therefore, this book is about some trivial principal bundles!!
def
We next define A1 = ∪x ∈ M U1x,x, and observe that the restriction π : A1 → M maps
the group U1x,x onto the point x. The local representation of a 1-arrow shows again that
π : A1 → M is a smooth and locally trivial group bundle, and we call a (smooth) section of
this bundle a (universal first-order) gauge transformation. The fiberwise group structure
allows us to compose two gauge transformations in the obvious way, and therefore the
set of gauge transformations forms a (very big) group. Gauge transformations will play a
fundamental role in Chapters 8, 9, and 13.
Finally, we should mention, for those readers familiar with the formalism of principal
bundles and their associated bundles, that the group bundle A1 → M and the tangent
and cotangent bundles TM → M and T ∗ M → M can be constructed as bundles associated
with the principal bundles U1e,• → M or U1•,e → M. However, we will not make use of this
general formalism until Chapter 16.
2
• • • • • • •
The Nonlinear Curvature
Let M be a smooth manifold with a splitting ε, which we fix once and for all. We denote
this data by (M, ε). By the definition of the 1-arrow ε(p, q) from p to q, there exists a local
diffeomorphism f defined near p with f (p) = q such that
j1 ( f )p, f (p) = ε(p, f (p)) (2.1)
The representative f in (2.1) defines also the 1-arrows j1 ( f )x,f (x) , for all x ∈ Dom( f ).
However, there is no reason why we should have
j1 ( f )x,f (x) = ε(x, f (x)), x ∈ Dom( f ) (2.2)
even though (2.2) holds at x = p by (2.1). The key fact here is that f in (2.1) depends on
(p, q) and the same f may not work for other (p, q)’s. The question arises whether there
exists a local diffeomorphism f defined near p satisfying
(i) f (p) = q;
(ii) j1 ( f )x,f (x) = ε(x, f (x)) for all x ∈ Dom( f ).
If (i) and (ii) hold, we say that f is a local solution of ε satisfying the initial condition
(p, q) ∈ U0 = M × M and that the 1-arrow ε(p, q) integrates to the local solution f . Clearly,
if f is a local solution of ε, then it satisfies any initial condition (r, f (r)) for r ∈ Dom( f ).
Equivalently, a local solution integrates all its 1-arrows. This fundamental concept is worth
a definition:
Definition 2.1 Let U ⊂ M be an open subset and f : U → f (U) be a diffeomorphism. Then f
is a solution of ε on U if
j1 ( f )x, f (x) = ε(x, f (x)), x∈U (2.3)
Some natural questions arise: Does ε admit solutions? If it does, where are the solutions
defined? Are they unique? Our ultimate purpose will be to find conditions that will imply
the existence of solutions of ε that are global diffeomorphisms of M.
To answer the above questions, we express (2.3) in coordinates. Now, the local repre-
sentation j1 ( f )x,f (x) = (xi , yi , (∂f i /∂x j )) and (1.24) in Chapter 1 show that (2.3) takes the
form (xi , f (x)i , (∂f i /∂x j )) = (xi , f (x)i , εji (x, f (x))) for all x ∈ U, which is equivalent to
∂f i
= εji (x, f (x)), x∈U (2.4)
∂x j
Now, (2.4) is a first order nonlinear system of PDEs for f . Assuming that f is a solution of
(2.4), we differentiate (2.4) with respect to xk , which gives
∂ 2f i ∂εji (x, f (x)) ∂εji (x, f (x)) ∂f a

= + (2.5)
∂xk ∂x j ∂xk ∂ya ∂xk
Substituting ∂f a /∂xk from (2.4) into (2.5) and alternating k and j, we get
i
∂εj (x, f (x)) ∂εji (x, f (x)) a
+ εk (x, f (x)) =0 (2.6)
∂xk ∂ya
[kj]

where [kj] means alternation; for instance, Aikj = Aikj − Aijk (omitting the factor 12 ).
[kj]
Now (2.6) gives the so-called integrability conditions for (2.4). We define
i
def ∂εj (x, y) ∂εji (x, y) a
Rjk (x, y) =
i
+ εk (x, y) (2.7)
∂xk ∂ya
[kj]
where x ∈ U, and y ∈ V = f (U). Therefore, Rikj (x, f (x)) = 0, x ∈ U, is a necessary condi-

tion for the existence of a solution f : U → f (U) = V. Following the old terminology, we
call ε completely integrable if for any (p, q) ∈ U0 , ε admits a local solution f satisfying
f (p) = q. Consequently, complete integrability of ε on (U, x) × (V, y) implies R(x, y) = 0
on (U, x) × (V, y).
It seems at first that Rijk (p, q) is defined on (U, x) × (V, y), where (U, x) and (V, y) are
any two coordinate neighborhoods, and therefore R has only a local meaning. However,
since ε is globally defined on U0 = M × M, R is defined consistently on the overlap
((U, x) ∩ (Z, z))× (V, y) ∩ (W, w) , and therefore R(p, q) is also defined globally on U0 .
What kind of an object is R? Equivalently, how does Rikj (x, y) transform if we apply the
coordinate changes (x) → (z) and ( y) → (w)?
Lemma 2.2 We have the transformation rule
∂xc ∂xb a ∂wi

Rikj (z, w) = R (x, y) (2.8)
∂z j ∂zk bc ∂ya
The proof of Lemma 2.2 is a somewhat tedious but straightforward verification that we
will leave to the reader. However, it is quite easy to guess (2.8): Since Rikj (x, y) = 0 does
the nonlinear curvature | 19
have a coordinate-free meaning and only first-order derivatives are involved, Rikj (x, y)
must transform tensorially. Consequently, a coordinate change (x) → (z) at the source
transforms Rikj (x, y) as a 2-form in the indices k, j, and a coordinate change ( y) → (w)
at the target transforms Rikj (x, y) as a tangent vector in the index i, which is the
content of (2.8).
The next proposition restates (2.8) in an elegant coordinate-free way:

Proposition 2.3 R(p, q) is an element of ∧2 Tp∗ ⊗ Tq .
Definition 2.4 R is the nonlinear curvature of (M, ε).

We have already observed that the complete integrability of ε on U0 = M × M implies
R = 0 on U0 . Our purpose is to find a local converse to this statement. This is contained in
the following proposition:
Proposition 2.5 Suppose Rikj (x, y) = 0 on (U, x) × (V, y). For any (p, q) ∈ U × V, there
exists a neighborhood U ⊂ U of p and a diffeomorphism defined on U ⊂ U, f (U) ⊂ V,
satisfying (i) f (p) = q and (ii) j1 ( f )x,f (x) = ε(x, f (x)) for all x ∈ U; that is, f is a solution
of ε on U with the initial condition f (p) = q. Further, f is unique: If g is another such solution
on W ⊂ U, then f = g on U ∩ W.
Proposition 2.5 is a direct consequence of the well-known local existence and uniqueness
theorem for systems of first-order PDEs with initial conditions applied to (2.4) (see, e.g.,
the Appendix of [L]).
Corollary 2.6 Suppose R = 0 on M × M. Then, for any (p, q) ∈ M × M, (2.4) admits a
unique local solution f with f (p) = q; that is, any 1-arrow ε(a, b) integrates uniquely to a
local solution.
Corollary 2.7 Suppose f , g are two solutions of ε on some connected open set U ⊂ M. If
f (p) = g(p) for some p ∈ U, then f ≡ g on U.
def
Proof We define A = {x ∈ U | f (x) = g(x)} ⊂ U. Clearly, A is closed in U but also open by
Proposition 2.5, so A = U.
An important property of R is given by the following proposition:

Proposition 2.8 R(p, p) = 0, p ∈ M; that is, R vanishes on the diagonal ⊂ M × M.
Proof Differentiation of (1.28) with respect to xk gives
∂εai (x, y) a ∂εja (y, x)

εj (y, x) + εa
i
(x, y) =0 (2.9)
∂xk ∂xk
Setting y = x and using (1.27), (2.9) becomes

∂εji (x, y) ∂εji (y, x)
+ =0 (2.10)
∂xk ∂xk
y=x y=x
Setting y = x in (2.7) and again using (1.27), we get

⎡ ⎤
∂εji (x, y) ∂εji (x, y)
Rijk (x, x) = ⎣ + ⎦
∂xk ∂yk
y=x y=x [kj]
= [0][kj] = 0 (2.11)
in view of (2.10).
Proposition 2.8 is not very surprising: The initial conditions (p, p) integrate to the
identity diffeomorphism, and therefore the integrability conditions should be vacuous in
this case.
For the reader familiar with the theory of connections on principal bundles and their
associated vector bundles, this seems to be a good place to comment on some intriguing
points: The trivialization ε of the principal bundle U1e,• → M defines an obvious “flat
connection” on the principal bundle U1e,• → M according to the general theory, whereas ε
now has a nonlinear curvature R that need not vanish! Even more intriguing is the fact that
R determines a “linear curvature” R that need not vanish either, as we will show in Chapter 6.
In fact, we will prove that R = 0 ⇔ R = 0. Clearly, R cannot be the curvature of the
induced linear connection on the associated vector bundle TM → M, which is flat too. This
discrepancy will be clarified in Chapters 5 and 6.
According to Lemma 2.2, if we fix p ∈ M and some coordinates around p, then Rijk (p, x)
transforms only in the index i under a coordinate change (x) → ( y), whereas Rijk (x, p)
transforms only in the indices j, k. This makes R a little awkward to work with. We will
now tame R in two ways for later use.
We define
←
− def
R ijk (x, y) = Riab (x, y)εja (y, x)εkb (y, x) (2.12)
and
−
→i def
R jk (x, y) = εai (y, x) Rajk (x, y) (2.13)
Clearly,
−
→ ←
−
R = 0 ⇐⇒ R = 0 ⇐⇒ R = 0 (2.14)
on U0 = M × M. We recall that the chain rule together with the definition of the 1-arrow
gives the transformation rules
∂ya
εki (x, z) = εai (y, z) (2.15)
∂xk
∂xi a
εki (z, x) = ε (z, y) (2.16)
∂ya k
the nonlinear curvature | 21
under a coordinate change (x) → ( y). Note that (2.16) is obtained from (2.15) by inver-
sion, and vice versa. Now, if we fix p ∈ M and a coordinate system around p once and for all,
←−
then R ijk (p, x) transforms as a tensor in all indices i, j, k under a coordinate change (x) → ( y)
−
→
as follows easily from (2.12), (2.15), and Lemma 2.2. Similarly, R ijk (x, p) transforms as a
tensor in all indices upon a coordinate change (x) → ( y). However, the tensors defined in
this way obviously depend also on our choice of the point p, a subtle dependence to be
clarified in Chapter 6.
3
• • • • • • •
Local Lie Groups
In [O1], Olver defines a local Lie group and makes the crucial observation that a local Lie
group does not necessarily “globalize” to a Lie group. In his words, “The theory of local Lie
groups is not a simple consequence of the global theory but has its own set of interesting and
delicate geometric structures.” The definition of a local Lie group that we will give below is
different than the one in [O1]. Nevertheless, it is fair to say that Part I of this book will
amply justify the deep insight of [O1]. In fact, we will show in Chapter 4 that a Lie group is
a special (globalizable) local Lie group. This fact reinstates the paradigm of local to global
to its historical record, as remarked by Olver [O4].
As in Chapter 2, we assume that M is a smooth manifold with a fixed splitting ε and denote
this data by (M, ε).
Definition 3.1 (M, ε) is called a local Lie group (LLG) if R = 0 on M × M.
A word of caution with regard to the terminology of Definition 3.1: In the theory of Lie
groups, a local Lie group refers to a neighborhood of the identity in some Lie group and
is therefore a local concept. We observe that a LLG according to Definition 3.1 is a global
concept.
In this chapter, (M, ε) is a LLG. According to Corollary 2.6, ε admits unique local
solutions with arbitrary initial conditions. Let G denote the set of all local solutions of ε.
We will use the notation (M, ε, G ) for the LLG (M, ε).
At this stage, it is useful to recall the definition of a pseudogroup on M. A pseudogroup
on M is a collection S of local diffeomorphisms of M satisfying the following:
1. If f , g ∈ S and Dom( f ) = Ran( g), then f ◦ g ∈ S .

2. If f ∈ S , then f −1 ∈ S .
3. If f ∈ S and U ⊂ Dom( f ), then f|U ∈ S .
4. If U = ∪ Uα and f is local diffeomorphism defined on U such that f|Uα ∈ S for all
α, then f ∈ S .
5. The identity diffeomorphism I belongs to S .
If for any p, q ∈ M there exists some f ∈ S with f (p) = q, then S is called transitive.
Proposition 3.2 Let (M, ε, G ) be an LLG. Then the set G of local solutions of ε is a transitive
pseudogroup on M.
Proof Properties 3–5 and transitivity are immediate from Definition 2.1 and (2.4), so it
suffices to check properties 1 and 2 locally in coordinates. For property 1, let y = y(x)
and z = z( y) be two solutions of (2.4). We then have
∂yi ∂zi
= εji x, y , = εji y, z (3.1)
∂x j ∂yj
Therefore,
∂zi ∂zi ∂ya

= a j = εai y, z εja (x, y) = εji (x, z) (3.2)
∂xj ∂y ∂x
in view of (1.26), and therefore z = z(x) belongs to G . Inverting both sides

of ∂yi /∂xj = εji (x, y) and using (1.28), we get ∂xi /∂yj = εji (y, x), and property
2 holds.
Corollary 2.7 shows that G is a very special pseudogroup: Its local diffeomorphisms
are determined on their connected domains by any of their values or 0-arrows. This fact
allows us to “analytically continue” elements of G along paths as follows.
Suppose (f , U) ∈ G , where U = Dom( f ), p ∈ U, and let C : x(t), 0 ≤ t ≤ 1, be a
(continuous) path from p = x(0) to some q = x(1). Let {Uα } , 1 ≤ α ≤ k, be an open
covering of C satisfying
(i) U1 = U
(ii) (fα , Uα ) ∈ G , f1 = f
(iii) fi = fj on Ui ∩ Uj
We call (fα , Uα ) a continuation of (f1 , U1 ) along C. Let (gβ , Vβ ), 1 ≤ β ≤ m, be a continu-

ation of (g1 , V1 ) along C such f1 = g1 on U1 ∩ V1 , and therefore f1 (p) = g1 (p). Suppose
q ∈ Ur ∩ Vs for some 1 ≤ r ≤ k and 1 ≤ s ≤ m. Using Corollary 2.7, we easily show that
fr = gs on Ur ∩ Vs and therefore fr (q) = gs (q). In short, two continuations of a local solution
around p along a path from p to q define the same values on the path. Even more succinctly,
we can state the following proposition:
Proposition 3.3 A continuation of a local solution around p along a path from p to q is unique
if it exists.
Proposition 3.3 suggests the following important definition:
Definition 3.4 (M, ε, G ) is complete if the elements of G can be continued indefinitely along all
paths.
Observe that we define completeness with the assumption R = 0; that is, that (M, ε) is
an LLG. It is an interesting question whether completeness can be defined without this
assumption.
local lie groups | 25
Some intuitive observations: (M, ε, G ) may fail to be complete because while continuing
some local solution f along a path from p to q, there may be a point r on the path between p
and q such that we may not be able to “continue through r” even though continuation from p
up to and excluding r is possible. If the continuation up to and including r is possible, then we
can continue through r. We define the continuation of f ∈ G along a path x(t), 0 ≤ t 1,
as above, except that the covering (Uα ) need not be finite. With these intuitions, we easily
deduce the following proposition:
Proposition 3.5 If M is compact, then (M, ε, G ) is complete.
To prove Proposition 3.5, we need the following technical lemma:
Lemma 3.6 Let (M, ε, G ) be a LLG and (p, q) ∈ M × M. There exists a sufficiently small
neighborhood (p, q) ∈ U × V of (p, q) with the following property: For any (x, y) ∈ U ×
V, the unique local solution g ∈ G with the initial condition g(x) = y can be defined
on U.
Lemma 3.6 states that G is always “locally complete” in a sense to be made clear below.
Using this lemma, we can now easily prove Proposition 3.5: Let (f , W) ∈ G , x(t), 0 ≤ t ≤ 1,
def
be a path with x(0) = p ∈ W and x(1) = q. We define t0 = sup{t ∈ [0, 1] | f has a con-
tinuation on [0, t)}. We choose a sequence tn → t0− . Since M is compact, f (tn ) has a
subsequence of the form f (tnk ) that converges to some r ∈ M. We solve (2.4) for some g
defined on V with the initial condition g(x(t0 )) = r, where V is as in Lemma 3.6. Since
(tnk , f (tnk )) ∈ V × g(V) eventually, we conclude that g(x(t)) = f (x(t)) for 0 ≤ t t0 , so g
gives a continuation of f on [0, t0 + ] for some 0 . This contradicts the definition of t0
unless t0 = 1.
To prove Lemma 3.6, we recall that Proposition 2.5 was proved by reducing the PDE
(2.4) to an ODE (see the Appendix of [L] for the technical details). We recall that a
vector field X on a smooth manifold N is complete if there exists an 0 such that the
1-parameter family of solutions h(t, x), 0 ≤ t, x ∈ N of X are defined on 0 ≤ t ≤ uniformly
for all x ∈ N. However, any p ∈ N has a neighborhood W such that this condition holds
on W. Therefore any vector field is locally complete but not necessarily globally complete.
Now the proof of Lemma 3.6 reduces to this local completeness of vector fields once we
reduce the PDE (2.4) to an ODE. We will leave the further technical details to the interested
reader.
Proposition 3.7 Let C1 and C2 be two paths from p to q and let f ∈ G be defined near p.
Suppose f can be continued indefinitely along all paths. If C1 is homotopic to C2 , then the
continuations along C1 and C2 define the same value at q.
The proof of Proposition 3.7 is the standard “monodromy” argument well known from
complex analysis. We recall here the main idea: We first observe that Proposition 3.7 holds
if C2 is a “small deformation” of C1 by the standard subdivision argument. For the general
case, since C1 is homotopic to C2 , we can start from C1 and reach C2 by a finite number of
small deformations. Since each small deformation does not change the value at q, the same
holds for the whole deformation.
Now we have the following important proposition:

Proposition 3.8 Suppose (M, ε, G ) is complete and M is simply connected. Then, for any
(f , U) ∈ G , there exists a unique diffeomorphism f ex : M → M such that f|U
ex = f . The set
def
G ex = {f ex | f ∈ G } is a group of transformations that acts simply transitively on M.
Proof Let (f , U) ∈ G with U connected. We fix some p ∈ U. For any x ∈ M, we choose a
path from p to x and define f ex (x) to be the terminal value of the continuation of f
along this path. Such a continuation is possible since (M, ε) is complete. Since M is
simply connected, any two paths from p to x are homotopic, and therefore f ex (x) is
independent of the chosen path and also independent of the choice of the point p ∈ U
by Proposition 3.7. Thus, we get a map f ex : M → M with f|U ex = f . By Corollary 2.7,
the extension f is unique. We claim that f : M → M is a diffeomorphism. It is onto:

ex ex
For any y ∈ M, we choose a path C from f (p) to y. By Proposition 3.2, (f −1 , f (U)) ∈ G .

Using completeness, we continue (f −1 , f (U)) along C and define x to be the terminal
value of this continuation at y. Now this continuation defines a continuation of (f , U)
along a path from p to x with the terminal value y, so f ex (x) = y and therefore f ex
is onto. A similar argument shows that f ex is also 1–1. Hence, we conclude that
f ex : M → M is a diffeomorphism. We claim that the set {f ex | f ∈ G } is closed under
composition and inversion. Indeed, f ex is a global solution of (2.4) on M, and therefore
(f ex , M) ∈ G . The conclusion now follows from Proposition 3.2. Transitivity is clear:
Any 0-arrow (p, q) integrates uniquely to a local solution f with f (p) = q that extends
uniquely to some f ex . Finally, Corollary 2.7 shows that the group G ex acts simply
transitively on M.
Suppose (g, U), (f , V) ∈ G , with g(U) = V. Note that the formulas (f ◦ g)ex = f ex ◦ g ex
and (f −1 )ex = (f ex )−1 are implicitly contained in the proof of Proposition 3.8.
Definition 3.9 Let (M, ε, G ) be an LLG. If f ∈ G extends (necessarily uniquely) to a diffeomor-
phism f ex : M → M, then f is globalizable. If all f ∈ G are globalizable, then (M, ε, G ) is
globalizable.
It follows that a complete and simply connected LLG is globalizable by Proposition 3.8.
Obviously, if (M, ε, G ) is globalizable, then it is complete.
For readers familiar with the definition of an abstract Lie group, the standard example of
a globalizable LLG arises from a Lie group G as follows: Let La be left translation by a and
def
define L(G) = {La | a ∈ G}. For any p, q ∈ G, there exists a unique La satisfying La (p) = q,
def
a = qp−1 . We define ε(p, q) = j1 (La )p . Clearly, ε is a splitting that admits La as a global
solution. We recall that if a transformation group G acts simply transitively on M, then the
pair (M, G) is a called a principal homogeneous space (PHS). It follows that (G, L(G)) is a
PHS. Similarly, (G, R(G)) is another PHS. If (M, ε, G ) is globalizable, then (M, G ex ) is yet
another PHS. We observe here the crucial fact that a Lie group G gives rise to two PHSs,
(G, L(G)) and (G, R(G)), whereas a globalizable LLG (M, ε, G ) gives rise to only one PHS,
(M, G ex ). Indeed, the transformation group G ex shuffles the points of M and there is no
such thing as shuffling from the right or left, and so this is only a matter of notation, not a
local lie groups | 27
fundamental concept. This means that we need to do more work to construct some abstract
Lie group G from the PHS (M, G ex ), and this will be the subject of the next chapter.
We have seen above that if (M, ε, G ) is complete, then the local solutions of (2.4) can be
continued uniquely along paths according to Proposition 3.3, but two continuations along
different paths from p to q may give different values at q if these paths are not homotopic. In
this case, we can lift G to the universal covering of M as follows.
Let π : M → M be a covering space of (M, ε) and take a, b ∈ M with π(a) = p, π(b) = q.
There exist open neighborhoods U, V of a, b such that π|U , π|V are diffeomorphisms. Now
using the 1-arrow ε(p, q) from p to q, we define the 1-arrow π −1 (ε)(a, b) from a to b by the
formula
−1 q,b
π −1 (ε)(a, b) = j1 (π|V
def
) ◦ ε(p, q) ◦ j1 (π|U )a,p (3.3)
Using (3.3) and (1.26), we easily check that π −1 (ε)(b, c) ◦ π −1 (ε)(a, b) = π −1 (ε)(a, c),
. We call (M
and therefore π −1 (ε) is a splitting on M , π −1 (ε)) the lift of (M, ε) to M
. If
(M, ε, G ) is an LLG and (f , W) is a local solution of ε that integrates the 1-arrow ε(p, q),
then, by restricting W, U, V so that π(U) = W and f (W) = V, it is clear from (3.3) that the
local diffeomorphism
−1
π −1 ( f ) = π|V
def
◦ f ◦ π|U (3.4)
is defined on U, maps a to b, and integrates the 1-arrow π −1 (ε)(a, b). Therefore,

R(ε) = 0 implies R(π −1 (ε)) = 0; that is, if (M, ε) is an LLG, then so is its lift
(M , π −1 (ε)). In this case, we denote the local solutions of π −1 (ε) by π −1 ( G ) and say
that (M , π −1 (ε), π −1 ( G )) lifts (M, ε, G ).
, π −1 (ε), π −1 ( G )).
Proposition 3.10 If (M, ε, G ) is complete, then so is (M
, a ∈ U, (h, U) be a solution of π −1 (ε) on U and let C be a path from a
Proof Let a, b ∈ M
to some b ∈ M . The path C projects to the path π(C) from π(a) = p to π(b) = q. By
restricting U if necessary, (h, U) is the lift of some local solution (f , W) of ε defined by
(3.4), where π(U) = W. Since (M, ε) is complete, (f , W) has some continuation along
π(C) from p to q. Since π is a local diffeomorphism, this continuation (or, rather, its
restriction to possibly smaller domains) lifts to a continuation of (h, U) from a to b,
showing that (M , π −1 (ε)) is complete.
Propositions 3.10 and 3.8 give the following corollary:

→ M is the universal covering, then
Corollary 3.11 If (M, ε, G ) is complete and π : M
, π −1 (ε), π −1 ( G )) is globalizable.
(M
Therefore, we have the PHS (M , π −1 ( G )ex ) if (M, ε, G ) is complete.
It is possible to carry the above analysis further. For instance, assuming the completeness
of (M, ε, G ), which f ∈ G are globalizable? We recall that f ∈ G with f (p) = p for some
p ∈ M must be the restriction of I to Dom( f ). Therefore, such f ∈ G are surely globalizable.
Let π : M → M be the universal covering space and D be the group of deck transformations
. We recall that D acts simply transitively on each fiber π −1 (p) and is isomorphic to
on M
the fundamental group π (M, p) of M. In particular, the elements of D commute with the
projection π . We will leave the proof of the following proposition to the interested reader:
Proposition 3.12 Suppose (M, ε, G ) is complete. Then the following are equivalent.:
(i) f ∈ G is globalizable;
(ii) π −1 ( f )ex commutes with D.
Corollary 3.13 Suppose (M, ε, G ) is complete. Then (M, ε, G ) is globalizable if and only if
D ⊂ π −1 ( G )ex is a normal (hence central) subgroup.
We will conclude this chapter with the following important question:
Q: Given some (M, ε), is there a splitting ε on M with R(ε) = 0? Equivalently, does a
parallelizable M admit an LLG structure?
As we will see in Chapter 13, for compact and simply connected 3-manifolds, Q is equivalent
to the Poincaré Conjecture (PC). Therefore, the answer is affirmative by the recent proof
of the PC due to Hamilton and Perelman. This gives an idea of the depth and the level of
difficulty of Q.
4
• • • • • • •
The Centralizer
In this chapter, (M, ε, G ) is an LLG as defined in Chapter 3. If (M, ε, G ) is globalizable, then

we obtain the PHS (M, G ex ) as we have seen in Chapter 3. For simplicity of notation, we will
henceforth denote (M, G ex ) by (M, G ). The purpose of this chapter is to define another
pseudogroup C( G ) and endow M with the structure of an abstract Lie group using G and
C( G ), with the assumption that (M, ε, G ) globalizes. The idea is simple: Let X be a finite set
and let P(X) be the set of all permutations of X, that is, the set of all 1–1 and onto maps of X.
Let G ⊂ P(X) be a subset that is closed under composition ◦ and inversion of permutations
and let C(G) ⊂ P(X) be the set of all permutations that commute with the elements G,
that is, the centralizer of G inside P(X). Clearly, C(G) is also closed under composition and
inversion. Now, if G acts simply transitively on X, then C(G) also acts simply transitively. To
see this, we fix p, q ∈ X arbitrarily. For any x ∈ X, there exists a unique f ∈ G with f (p) = x.
def
We define h : X → X by h(x) = f (q). We easily see that h is 1–1 and onto, h(p) = q, and
h commutes with G. Thus, C(G) acts transitively and we easily check that this action is also
simply transitive. Is there any conceptual reason why C(G) also acts simply transitively? To
answer this, we fix a base point e ∈ X and identify the elements of the abstract group (G, ◦)
with X in such a way that the action of G on X is identified with the, say, left action of G
on G X. Now the right action of G on G X gives another simply transitive action on
X, which can be identified with the above action of C(G). Note that there is no canonical
choice of left/right.
Now suppose that R = 0 and that (M, ε, G ) globalizes, so we have the PHS (M, G ),
where G acts simply transitively on M according to Proposition 3.8. We have the analogy
X M, G G , and P(X) Diff(M). We know that the elements of G are the unique
solutions of the nonlinear PDE (2.4). If we define C( G ) as the centralizer of G inside
Diff(M), what is the structure of C( G )? For instance, are its elements the solutions of
another nonlinear PDE?
Let f (x, y, z) denote the unique local solution f ∈ G of (2.4) in the variable z ∈ U
satisfying the initial condition f (x) = y. Using Lemma 3.6, we choose U such that f (x, y, z)
is defined for all x, y, z ∈ U. If (M, ε, G ) is globalizable, then clearly f (x, y, z) is defined for
all x, y, z ∈ M. Now we have
∂f i
(p, q, x) = εji (x, f (p, q, x)), f ∈ G, p, q, x ∈ U (4.1)
∂xj
Clearly,
f (x, y, x) = y and f (x, x, y) = y, x, y ∈ U (4.2)
Now we fix p, q ∈ U and define the map g on U by
def
g( y) = f (p, y, q), y∈U (4.3)
so g(p) = q by (4.2). Now, g is a 1–1 map by Corollary 2.7. Further, it is a diffeomorphism,

since the solutions of (2.4) depend smoothly on the initial conditions. Note that g(U) need
not be contained in U. We define the 1-arrowε(p, q) from p to q by the formula
def

ε(p, q) = j1 ( g)p,q (4.4)
In coordinates, (4.4) is given by

def ∂f i (p, y, q)
(p , q
i i
εji (p, q)) =
, i
p ,q , i
(4.5)
∂yj y=p
which is equivalent to

∂g i ( y)
=
εji (p, g(p)) (4.6)
∂yj y=p
We claim
∂g i ( y)
=
εji (y, g( y)) y∈U (4.7)
∂yj
def
that is, (4.6) holds not only at y = p but for all y ∈ U. So, let p, x ∈ U, q = f (p, p, q), and
def
y = f (p, x, q) = f (p, x, q). Therefore,
f (p, x, q) = f (p, x, f (p, p, q)) (4.8)
We differentiate (4.8) at x = p using (4.6). This gives

∂g i (x) ∂f i (p, x, q)
=
∂xj x=p ∂xj x=p
i
∂f (p, x, f (p, p, q))
=
∂xj x=p
=
εji (p, f (p, p, q))
=
εji (p, g(p)) (4.9)
which proves the claim, since p is arbitrary.

the centralizer | 31
Note that (4.7) is (2.4) with ε replaced by ε. However, there are two major differences:
First, we need the assumption R = 0 to solve (2.4) locally, whereas (4.7) always has the
unique local solution g defined by (4.3) with this assumption. Second, ε is defined on M×M,
whereas ε is defined on U × U, but if (M, ε, G ) is globalizable, then ε is defined also on
M × M.
Proposition 4.1

ε(p, r) =
ε(q, r) ◦
ε(p, q), p, q, r ∈ U (4.10)
ε is a local splitting defined on U × U.

that is,
Proof Let f ∈ G be defined on U, where U is as above. Suppose f (p) = r and f (q) = m,
p, q, r, m ∈ U. In our notation, this means that f (p, r, x) = f (x) = f (q, m, x) for all
x ∈ U. Therefore,
f (p, r, x) = f (q, f (p, r, q), x) p, q, r, x, f (p, r, q) ∈ U (4.11)
which generalizes (4.2).

We regard r = y as variable, fix x = r and differentiate (4.11) with respect to y at
y = p by the chain rule using (4.6). What we get is
i a
∂f i ∂f ∂f
(p, y, r) = (q, z, r) (p, y, q) (4.12)
∂yj y=p ∂za z=q ∂y
j
y=p
which is (4.10), in view of (4.6).
To summarize, any p ∈ M has a neighborhood U such that we have (U, ε) in the same
way as (M, ε). Let C( G , U) denote the set of local solutions of ε on U. As we did for
(M, ε, G ) in Chapter 3, we easily show using (4.7) that C( G , U) is closed under inversion
and composition whenever they are defined. Therefore, C( G , U) is a transitive pseudogroup
on U and we have the LLG (U, ε, C( G , U)) in the same way as (M, ε, G ). By our choice
of U, any g ∈ C( G , U) is defined on U, but it is not necessarily the case that g(U) ⊂ U,
as remarked above and as indicated by the condition f (p, r, q) ∈ U in (4.11). If (M, ε, G )
is globalizable, we can choose U = M and denote C( G , U) by C( G ), which is a global
transformation group that acts simply transitively on M. From the definition of C( G ), it
follows that C( G ) commutes with G . Indeed, let g ∈ C( G ), f ∈ G and p ∈ M. Suppose
g(p) = q and f (p) = r. Now g(f (p)) = g(r) = h(p, r, q), where h(p, r, q) is the value at q of
the unique solution of (2.4) that maps p to r. Therefore, h(p, r, q) = f (q) = f (g(p)). Since p
is arbitrary, we conclude that g ◦ f = f ◦ g.
To recapitulate, we have the following proposition:
Proposition 4.2 Let (M, ε, G ) be an LLG. Any p ∈ M has a neighborhood p ∈ U such that
(U,ε, C( G , U)) is a transitive LLG. If (M, ε, G ) is globalizable, then we have the PHSs
(M, C( G )) and (M, G ), where the transformations of C( G ) and G commute.
ε on U × U to a
Note that if M is simply connected, then we can extend the local splitting
global splitting on M×M as follows: Let p, q ∈ M and C be any path from p to q. We partition
C into sufficiently small curves by choosing the points x1 (= p), x2 , . . . , xn−1 , xn (= q) such
def
thatε(xk , xk+1 ) is defined for all k. We define
ε(p, q) =
ε(xn−1 , q) ◦ · · · ◦
ε(x1 , x2 ). Passing
to a common refinement shows that the definition of ε(p, q) is independent of the parti-
tion, and the standard monodromy argument shows that it is also independent of the path.
Henceforth in this chapter, we always assume that (M, ε, G ) is globalizable so that we
have the two commuting PHSs (M, G ) and (M, C( G )). Fixing some base point e ∈ M,
for any x ∈ M, there exists a unique f ∈ G (= G ex ) with f (e) = x. This gives a 1–1
correspondence between G and M. The same conclusion holds also for C( G ). Note that
f ∈ G and g ∈ C( G ) satisfying f (e) = g(e) need not be equal on M. There is an obvious
bijective map
e : G −→C( G ) (4.13)
that maps f ∈ G to the unique g satisfying g(e) = f (e). Clearly, e (g ◦ f ) = e ( f ) ◦ e ( g)

according to (4.3) and (4.6).
Now, let Diff(M) denote the group of global diffeomorphisms of M and let CDiff(M) ( G )
denote the centralizer of G inside Diff(M). Clearly, we have C( G ) ⊂ CDiff(M) ( G ).
Proposition 4.3 C( G ) = CDiff(M) ( G ) = {g ∈ Diff(M) | g ◦ f = f ◦ g, f ∈ G }.
Proof We claim that CDiff(M) ( G ) ⊂ C( G ). Let g ∈ CDiff(M) ( G ), g(p) = q, and f ∈ G with
f (p) = x. Now
(g ◦ f )(p) = (f ◦ g)(p)
⇐⇒ g(x) = f (p, x, q)
i i
∂g ∂f (p, x, q)
⇒ =
∂xj x=p ∂xj x=p
⇐⇒ j1 ( g)p =
ε(p, g(p)) (4.14)
using (4.3) and (4.4) in the last step. Since p is arbitrary, it follows that g is a solution
of (4.6) on M × M, proving the claim.
If G is a Lie group, we recall that (G, L(G)) and (G, R(G)) are two PHSs and also
globalizable LLGs. Clearly, the elements of L(G) and R(G) commute. Now, given the
two PHSs (M, G ) and (M, C( G )), our purpose is to turn M into a Lie group by defining
some multiplication and inversion on it in such a way that we will have L(M) = G and
R(M) = C( G ). Obviously, this Lie group structure cannot be canonical, because it is not
possible to distinguish between G and C( G ) since there is a complete symmetry between
them. In other words, we can also define this Lie group structure such that R(M) = G and
L(M) = C( G ). We warn the reader that this seemingly naive point turns out to be subtle
and will have intriguing consequences later, as already mentioned in Chapter 2.
We now resume our construction of the Lie group structure on M. First, we fix a base
point e ∈ M. We define an operation
Another random document with
no related content on Scribd:
personal identity. If we allow a man wit, it is part of the bargain that
he wants judgment: if style he wants matter. Rich, but a fool or miser
—a beauty, but vain; so runs the bond. ‘But’ is the favourite
monosyllable of envy and self-love. Raphael could draw and Titian
could colour—we shall never get beyond this point while the world
stands; the human understanding is not cast in a mould to receive
double proofs of entire superiority to itself. It is folly to expect it. If a
farther claim be set up, we call in question the solidity of the first,
incline to retract it, and suspect that the whole is a juggle and a piece
of impudence, as we threaten a common beggar with the stocks for
following us to ask a second alms. This is, in fact, one source of the
prevalence and deep root which envy has in the human mind: we are
incredulous as to the truth and justice of the demands which are so
often made upon our pity or our admiration; but let the distress or
the merit be established beyond all controversy, and we open our
hearts and purses on the spot, and sometimes run into the contrary
extreme when charity or admiration becomes the fashion. No one
envies the Author of Waverley, because all admire him, and are
sensible that admire him how they will, they can never admire him
enough. We do not envy the sun for shining, when we feel the benefit
and see the light. When some persons start an injudicious parallel
between him and Shakspeare, we then may grow jealous and uneasy,
because this interferes with our older and more firmly rooted
conviction of genius, and one which has stood a severer and surer
test. Envy has, then, some connexion with a sense of justice—is a
defence against imposture and quackery. Though we do not willingly
give up the secret and silent consciousness of our own worth to
vapouring and false pretences, we do homage to the true candidate
for fame when he appears, and even exult and take a pride in our
capacity to appreciate the highest desert. This is one reason why we
do not envy the dead—less because they are removed out of our way,
than because all doubt and diversity of opinion is dismissed from the
question of their title to veneration and respect. Our tongue, having a
license, grows wanton in their praise. We do not envy or stint our
admiration of Rubens, because the mists of uncertainty or prejudice
are withdrawn by the hand of time from the splendour of his works.
Fame is to genius—
‘Like to a gate of steel fronting the sun,
That renders back its figure and its heat.’
We give full and unbounded scope to our impressions when they are
confirmed by successive generations; as we form our opinions coldly
and slowly while we are afraid our judgment may be reversed by
posterity. We trust the testimony of ages, for it is true; we are no
longer in pain lest we should be deceived by varnish and tinsel; and
feel assured that the praise and the work are both sterling. In
contemporary reputation, the greater and more transcendant the
merit, the less is the envy attending it; which shows that this passion
is not, after all, a mere barefaced hatred and detraction from
acknowledged excellence. Mrs. Siddons was not an object of envy;
her unrivalled powers defied competitors or gainsayers. If Kean had
a party against him, it was composed of those who could not or
would not see his merits through his defects; and in like manner,
John Kemble’s elevation to the tragic throne was not carried by loud
and tumultuous acclamation, because the stately height which he
attained was the gradual result of labour and study, and his style of
acting did not flash with the inspiration of the God. We are backward
to bestow a heaped measure of praise, whenever there is any
inaptitude or incongruity that acts to damp or throw a stumbling-
block in the way of our enthusiasm. Hence the jealousy and dislike
shown towards upstart wealth, as we cannot in our imaginations
reconcile the former poverty of the possessors with their present
magnificence—we despise fortune-hunters in ambition as well as in
love—and hence, no doubt, one strong ground of hereditary right.
We acquiesce more readily in an assumption of superiority that in
the first place implies no merit (which is a great relief to the baser
sort), and in the second, that baffles opposition by seeming a thing
inevitable, taken for granted, and transmitted in the common course
of nature. In contested elections, where the precedence is understood
to be awarded to rank and title, there is observed to be less acrimony
and obstinacy than when it is supposed to depend on individual
merit and fitness for the office; no one willingly allows another more
ability or honesty than himself, but he cannot deny that another may
be better born. Learning again is more freely admitted than genius,
because it is of a more positive quality, and is felt to be less
essentially a part of a man’s self; and with regard to the grosser and
more invidious distinction of wealth, it may be difficult to substitute
any finer test of respectability for it, since it is hard to fathom the
depth of a man’s understanding, but the length of his purse is soon
known; and besides, there is a little collusion in the case:—
‘The learned pate ducks to the golden fool.’
We bow to a patron who gives us a good dinner and his countenance

for our pains, and interest bribes and lulls envy asleep. The most
painful kind of envy is the envy towards inferiors; for we cannot bear
to think that a person (in other respects utterly insignificant) should
have or seem to have an advantage over us in any thing we have set
our hearts upon, and it strikes at the very root of our self-love to be
foiled by those we despise. There is some dignity in a contest with
power and acknowledged reputation: but a triumph over the sordid
and the mean is itself a mortification, while a defeat is intolerable.
ON PREJUDICE
The Atlas.]
[April 11, 1830.
Prejudice, in its ordinary and literal sense, is prejudging any

question without having sufficiently examined it, and adhering to our
opinion upon it through ignorance, malice, or perversity, in spite of
every evidence to the contrary. The little that we know has a strong
alloy of misgiving and uncertainty in it: the mass of things of which
we have no means of judging, but of which we form a blind and
confident opinion as if we were thoroughly acquainted with them, is
monstrous. Prejudice is the child of ignorance; for as our actual
knowledge falls short of our desire to know, or curiosity and interest
in the world about us, so must we be tempted to decide upon a
greater number of things at a venture; and having no check from
reason or inquiry, we shall grow more obstinate and bigoted in our
conclusions, according as they have been rash and presumptuous.
The absence of proof, instead of suspending our judgments, only
gives us an opportunity to make things out according to our wishes
and fancies; mere ignorance is a blank canvas on which we lay what
colours we please, and paint objects black or white, as angels or
devils, magnify or diminish them at our option; and in the vacuum
either of facts or arguments, the weight of prejudice and passion falls
with double force, and bears down everything before it. If we enlarge
the circle of our previous knowledge ever so little, we may meet with
something to create doubt and difficulty; but as long as we remain
confined to the cell of our native ignorance, while we know nothing
beyond the routine of sense and custom, we shall refer everything to
that standard, or make it out as we would have it to be, like spoiled
children who have never been from home, and expect to find nothing
in the world that does not accord with their wishes and notions. It is
evident, that the fewer things we know, the more ready we shall be to
pronounce upon and condemn, what is new and strange to us; that
is, the less capable we shall be of varying our conceptions, and the
more prone to mistake a part for the whole. What we do not
understand the meaning of must necessarily appear to us ridiculous
and contemptible; and we do not stop to inquire, till we have been
taught by repeated experiments and warnings of our own fallibility,
whether the absurdity is in ourselves or in the object of our dislike
and scorn. The most ignorant people are rude and insolent, as the
most barbarous are cruel and ferocious. All our knowledge at first
lying in a narrow compass (bounded by local and physical causes)
whatever does not conform to this shocks us as out of reason and
nature. The less we look abroad, the more our ideas are introverted;
and our habitual impressions, from being made up of a few
particulars always repeated, grow together into a kind of concrete
substance, which will not bear taking to pieces, and where the
smallest deviation destroys the whole feeling. Thus the difference of
colour in a black man was thought to forfeit his title to belong to the
species, till books of voyages and travels, and old Fuller’s quaint
expression of ‘God’s image carved in ebony,’ have brought the two
ideas into a forced union, and Mr. Murray no longer libels men of
colour with impunity. The word republic has a harsh and
incongruous sound to ears bred under a constitutional monarchy;
and we strove hard for many years to overturn the French republic,
merely because we could not reconcile it to ourselves that such a
thing should exist at all, notwithstanding the examples of Holland,
Switzerland, and many others. This term has hardly yet performed
quarantine: to the loyal and patriotic it has an ugly taint in it, and is
scarcely fit to be mentioned in good company. If, however, we are
weaned by degrees from our prejudices against certain words that
shock opinion, this is not the case with all; for those that offend good
manners grow more offensive with the progress of refinement and
civilization, so that no writer now dare venture upon expressions that
unwittingly disfigure the pages of our elder writers, and in this
respect, instead of becoming callous or indifferent, we appear to
become more fastidious every day. There is then a real grossness
which does not depend on familiarity or custom. This account of the
concrete nature of prejudice, or of the manner in which our ideas by
habit and the dearth of general information coalesce together into
one indissoluble form, will show (what otherwise seems
unaccountable) how such violent antipathies and animosities have
been occasioned by the most ridiculous or trifling differences of
opinion, or outward symbols of it; for, by constant custom, and the
want of reflection, the most insignificant of these was as inseparably
bound up with the main principle as the most important, and to give
up any part was to give up the whole essence and vital interests of
religion, morals, and government. Hence we see all sects and parties
mutually insist on their own technical distinctions as the essentials
and fundamentals of religion, and politics, and, for the slightest
variation in any of these, unceremoniously attack their opponents as
atheists and blasphemers, traitors and incendiaries. In fact, these
minor points are laid hold of in preference, as being more obvious
and tangible, and as leaving more room for the exercise of prejudice
and passion. Another thing that makes our prejudices rancorous and
inveterate, is, that as they are taken up without reason, they seem to
be self-evident; and we thence conclude, that they not only are so to
ourselves, but must be so to others, so that their differing from us is
wilful, hypocritical, and malicious. The Inquisition never pretended
to punish its victims for being heretics or infidels, but for avowing
opinions which with their eyes open they knew to be false. That is,
the whole of the Catholic faith, ‘that one entire and perfect
chrysolite,’ appeared to them so completely without flaw and
blameless, that they could not conceive how any one else could
imagine it to be otherwise, except from stubbornness and
contumacy, and would rather admit (to avoid so improbable a
suggestion) that men went to a stake for an opinion, not which they
held, but counterfeited, and were content to be burnt for the pleasure
of playing the hypocrite. Nor is it wonderful that there should be so
much repugnance to admit the existence of a serious doubt in
matters of such vital and eternal interest, and on which the whole
fabric of the church hinged, since the first doubt that was expressed
on any single point drew all the rest after it; and the first person who
started a conscientious scruple, and claimed the trial by reason,
threw down, as if by a magic spell, the strongholds of bigotry and
superstition, and transferred the determination of the issue from the
blind tribunal of prejudice and implicit faith to a totally different
ground, the fair and open field of argument and inquiry. On this
ground a single champion is a match for thousands. The decision of
the majority is not here enough: unanimity is absolutely necessary to
infallibility; for the only secure plea on which such a preposterous
pretension could be set up is, by taking it for granted that there can
be no possible doubt entertained upon the subject, and by diverting
men’s minds from ever asking themselves the question of the truth of
certain dogmas and mysteries, any more than whether two and two
make four. Prejudice in short is egotism: we see a part, and
substitute it for the whole; a thing strikes us casually and by halves,
and we would have the universe stand proxy for our decision, in
order to rivet it more firmly in our own belief; however insufficient
or sinister the grounds of our opinions, we would persuade ourselves
that they arise out of the strongest conviction, and are entitled to
unqualified approbation; slaves of our own prejudices, caprice,
ignorance, we would be lords of the understandings and reason of
others; and (strange infatuation!) taking up an opinion solely from
our own narrow and partial point of view, without consulting the
feelings of others, or the reason of things, we are still uneasy if all the
world do not come into our way of thinking.
THE SAME SUBJECT CONTINUED
The Atlas.]
[April 18, 1830.
The most dangerous enemies to established opinions are those

who, by always defending them, call attention to their weak sides.
The priests and politicians, in former times, were therefore wise in
preventing the first approaches of innovation and inquiry; in
preserving inviolate the smallest link in the adamantine chain with
which they had bound the bodies and the souls of men; in closing up
every avenue or pore through which a doubt could creep in; for they
knew that through the slightest crevice floods of irreligion and heresy
would break in like a tide. Hence the constant alarm at free
discussion and inquiry; hence the clamour against innovation and
reform; hence our dread and detestation of those who differ with us
in opinion, for this at once puts us on the necessity of defending
ourselves, or of owning ourselves weak or in the wrong, if we cannot,
and converts that which was before a bed of roses, while we slept
undisturbed upon it, into a cushion of thorns; and hence our natural
tenaciousness of those points which are most vulnerable, and of
which we have no proof to offer; for, as reason fails us, we are more
annoyed by the objections and require to be soothed and supported
by the concurrence of others. Bigotry and intolerance, which pass as
synonymous, are, if rightly considered, a contradiction in terms; for,
if in drawing up the articles of our creed, we are blindly bigoted to
our impressions and views, utterly disregarding all others, why
should we afterwards be so haunted and disturbed by the last, as to
wish to exterminate every difference of sentiment with fire and
sword? The difficulty is only solved by considering that unequal
compound, the human mind, alternately swayed by individual
biasses and abstract pretensions, and where reason so often panders
to, or is made the puppet of, the will. To show at once the danger and
extent of prejudice, it may be sufficient to observe that all our
convictions, however arrived at, and whether founded on strict
demonstration or the merest delusion, are crusted over with the
same varnish of confidence and conceit, and afford the same firm
footing both to our theories and practice; or if there be any
difference, we are in general ‘most ignorant of what we are most
assured,’ the strength of will and impatience of contradiction making
up for the want of evidence. Mr. Burke says, that we ought to ‘cherish
our prejudices because they are prejudices;’ but this view of the case
will satisfy the demands of neither party, for prejudice is never easy
unless it can pass itself off for reason, or abstract undeniable truth:
and again, in the eye of reason, if all prejudices are to be equally
respected as such, then the prejudices of others are right, and ours
must in their turn be wrong. The great stumbling-block to candour
and liberality is the difficulty of being fully possessed of the
excellence of any opinion or pursuits of our own, without
proportionably condemning whatever is opposed to it; nor can we
admit the possibility that when our side of the shield is black, the
other should be white. The largest part of our judgments is prompted
by habit and passion; but because habit is like a second nature, and
we necessarily approve what passion suggests, we will have it that
they are founded entirely on reason and nature, and that all the
world must be of the same opinion, unless they wilfully shut their
eyes to the truth. Animals are free from prejudice, because they have
no notion or care about anything beyond themselves, and have no
wish to generalise or talk big on what does not concern them: man
alone falls into absurdity and error by setting up a claim to superior
wisdom and virtue, and to be a dictator and lawgiver to all around
him, and on all things that he has the remotest conception of. If mere
prejudice were dumb, as well as deaf and blind, it would not so much
signify; but as it is, each sect, age, country, profession, individual, is
ready to prove that they are exclusively in the right, and to go
together by the ears for it. ‘Rings the earth with the vain stir?’ It is
the trick for each party to raise an outcry against prejudice; as by this
they flatter themselves, and would have it supposed by others, that
they are perfectly free from it, and have all the reason on their own
side. It is easy indeed, to call names, or to separate the word
prejudice from the word reason; but not so easy to separate the two
things. Reason seems a very positive and palpable thing to those who
have no notion of it but as expressing their own views and feelings;
as prejudice is evidently a very gross and shocking absurdity (that no
one can fall into who wishes to avoid it), as long as we continue to
apply this term to the prejudices of other people. To suppose that we
cannot make a mistake is the very way to run headlong into it; for, if
the distinction were so broad and glaring as our self-conceit and
dogmatism lead us to imagine it is, we could never, but by design,
mistake truth for falsehood. Those, however, who think they can
make a clear stage of it, and frame a set of opinions on all subjects
by an appeal to reason alone, and without the smallest intermixture
of custom, imagination, or passion, know just as little of themselves
as they do of human nature. The best way to prevent our running
into the wildest excesses of prejudice and the most dangerous
aberrations from reason, is, not to represent the two things as having
a great gulph between them, which it is impossible to pass without a
violent effort, but to show that we are constantly (even when we
think ourselves most secure) treading on the brink of a precipice;
that custom, passion, imagination, insinuate themselves into and
influence almost every judgment we pass or sentiment we indulge,
and are a necessary help (as well as hindrance) to the human
understanding; and that, to attempt to refer every question to
abstract truth and precise definition, without allowing for the frailty
of prejudice, which is the unavoidable consequence of the frailty and
imperfection of reason, would be to unravel the whole web and
texture of human understanding and society. Such daring anatomists
of morals and philosophy think that the whole beauty of the mind
consists in the skeleton; cut away, without remorse, all sentiment,
fancy, taste, as superfluous excrescences; and, in their own eager,
unfeeling pursuit of scientific truth and elementary principles, they
‘murder to dissect.’ But of this I may say something in another paper.
THE SAME SUBJECT CONTINUED
It is a mistake, however, to suppose that all prejudices are false,

though it is not an easy matter to distinguish between true and false
prejudice. Prejudice is properly an opinion or feeling, not for which
there is no reason, but of which we cannot render a satisfactory
account on the spot. It is not always possible to assign a ‘reason for
the faith that is in us,’ not even if we take time and summon up all
our strength; but it does not therefore follow that our faith is hollow
and unfounded. A false impression may be defined to be an effect
without a cause, or without any adequate one; but the effect may
remain and be true, though the cause is concealed or forgotten. The
grounds of our opinions and tastes may be deep, and be scattered
over a large surface; they may be various, remote and complicated,
but the result will be sound and true, if they have existed at all,
though we may not be able to analyse them into classes, or to recall
the particular time, place, and circumstances of each individual case
or branch of the evidence. The materials of thought and feeling, the
body of facts and experience, are infinite, are constantly going on
around us, and acting to produce an impression of good or evil, of
assent or dissent to certain inferences; but to require that we should
be prepared to retain the whole of this mass of experience in our
memory, to resolve it into its component parts, and be able to quote
chapter and verse for every conclusion we unavoidably draw from it,
or else to discard the whole together as unworthy the attention of a
rational being, is to betray an utter ignorance both of the limits and
the several uses of the human capacity. The feeling of the truth of
anything, or the soundness of the judgment formed upon it from
repeated, actual impressions, is one thing: the power of vindicating
and enforcing it, by distinctly appealing to or explaining those
impressions, is another. The most fluent talkers or most plausible
reasoners are not always the justest thinkers.
To deny that we can, in a certain sense, know and be justified in
believing anything of which we cannot give the complete
demonstration, or the exact why and how, would only be to deny
that the clown, the mechanic (and not even the greatest
philosopher), can know the commonest thing; for in this new and
dogmatical process of reasoning, the greatest philosopher can trace
nothing above, nor proceed a single step without taking something
for granted;[60] and it is well if he does not take more things for
granted than the most vulgar and illiterate, and what he knows a
great deal less about. A common mechanic can tell how to work an
engine better than the mathematician who invented it. A peasant is
able to foretell rain from the appearance of the clouds, because (time
out of mind) he has seen that appearance followed by that
consequence; and shall a pedant catechise him out of a conviction
which he has found true in innumerable instances, because he does
not understand the composition of the elements, or cannot put his
notions into a logical shape? There may also be some collateral
circumstance (as the time of day), as well as the appearance of the
clouds, which he may forget to state in accounting for his prediction;
though, as it has been a part of his familiar experience, it has
naturally guided him in forming it, whether he was aware of it or not.
This comes under the head of the well-known principle of the
association of ideas; by which certain impressions, from frequent
recurrence, coalesce and act in unison truly and mechanically—that
is, without our being conscious of anything but the general and
settled result. On this principle it has been well said, that ‘there is
nothing so true as habit;’ but it is also blind: we feel and can produce
a given effect from numberless repetitions of the same cause; but we
neither inquire into the cause, nor advert to the mode. In learning
any art or exercise, we are obliged to take lessons, to watch others, to
proceed step by step, to attend to the details and means employed;
but when we are masters of it, we take all this for granted, and do it
without labour and without thought, by a kind of habitual instinct—
that is, by the trains of our ideas and volitions having been directed
uniformly, and at last flowing of themselves into the proper channel.
We never do anything well till we cease to think about the manner
of doing it. This is the reason why it is so difficult for any but natives
to speak a language correctly or idiomatically. They do not succeed in
this from knowledge or reflection, but from inveterate custom, which
is a cord that cannot be loosed. In fact, in all that we do, feel, or
think, there is a leaven of prejudice (more or less extensive), viz.
something implied, of which we do not know or have forgotten the
grounds.
If I am required to prove the possibility, or demonstrate the mode
of whatever I do before I attempt it, I can neither speak, walk, nor
see; nor have the use of my hands, senses, or common
understanding. I do not know what muscles I use in walking, nor
what organs I employ in speech: those who do, cannot speak or walk
better on that account; nor can they tell how these organs and
muscles themselves act. Can I not discover that one object is near,
and another at a distance, from the eye alone, or from continual
impressions of sense and custom concurring to make the distinction,
without going through a course of perspective and optics?—or am I
not to be allowed an opinion on the subject, or to act upon it, without
being accused of being a very prejudiced and obstinate person? An
artist knows that, to imitate an object in the horizon, he must use less
colour; and the naturalist knows that this effect is produced by the
intervention of a greater quantity of air: but a country fellow, who
knows nothing of either circumstance, must not only be ignorant but
a blockhead, if he could be persuaded that a hill ten miles off was
close before him, only because he could not state the grounds of his
opinion scientifically. Not only must we (if restricted to reason and
philosophy) distrust the notices of sense, but we must also dismiss
all that mass of knowledge and perception which falls under the head
of common sense and natural feeling, which is made up of the strong
and urgent, but undefined impressions of things upon us, and lies
between the two extremes of absolute proof and the grossest
ignorance. Many of these pass for instinctive principles and innate
ideas; but there is nothing in them ‘more than natural.’
Without the aid of prejudice and custom, I should not be able to
find my way across the room; nor know how to conduct myself in any
circumstances, nor what to feel in any relation of life. Reason may
play the critic, and correct certain errors afterwards; but if we were
to wait for its formal and absolute decisions in the shifting and
multifarious combinations of human affairs, the world would stand
still. Even men of science, after they have gone over the proofs a
number of times, abridge the process, and jump at a conclusion: is it
therefore false, because they have always found it to be true? Science
after a certain time becomes presumption; and learning reposes in
ignorance. It has been observed, that women have more tact and
insight into character than men, that they find out a pedant, a
pretender, a blockhead, sooner. The explanation is, that they trust
more to the first impressions and natural indications of things,
without troubling themselves with a learned theory of them; whereas
men, affecting greater gravity, and thinking themselves bound to
justify their opinions, are afraid to form any judgment at all, without
the formality of proofs and definitions, and blunt the edge of their
understandings, lest they should commit some mistake. They stay for
facts, till it is too late to pronounce on the characters. Women are
naturally physiognomists, and men phrenologists. The first judge by
sensations; the last by rules. Prejudice is so far then an involuntary
and stubborn association of ideas, of which we cannot assign the
distinct grounds and origin; and the answer to the question, ‘How do
we know whether the prejudice is true or false?’ depends chiefly on
that other, whether the first connection between our ideas has been
real or imaginary. This again resolves into the inquiry—Whether the
subject in dispute falls under the province of our own experience,
feeling, and observation, or is referable to the head of authority,
tradition, and fanciful conjecture? Our practical conclusions are in
this respect generally right; our speculative opinions are just as likely
to be wrong. What we derive from our personal acquaintance with
things (however narrow in its scope or imperfectly digested), is, for
the most part, built on a solid foundation—that of Nature; it is in
trusting to others (who give themselves out for guides and doctors)
that we are all abroad, and at the mercy of quackery, impudence,
and imposture. Any impression, however absurd, or however we may
have imbibed it, by being repeated and indulged in, becomes an
article of implicit and incorrigible belief. The point to consider is,
how we have first taken it up, whether from ourselves or the
arbitrary dictation of others. ‘Thus shall we try the doctrines,
whether they be of nature or of man.’
So far then from the charge lying against vulgar and illiterate
prejudice as the bane of truth and common sense, the argument
turns the other way; for the greatest, the most solemn, and
mischievous absurdities that mankind have been the dupes of, they
have imbibed from the dogmatism and vanity or hypocrisy of the
self-styled wise and learned, who have imposed profitable fictions
upon them for self-evident truths, and contrived to enlarge their
power with their pretensions to knowledge. Every boor sees that the
sun shines above his head; that ‘the moon is made of green cheese,’ is
a fable that has been taught him. Defoe says, that there were a
hundred thousand stout country-fellows in his time ready to fight to
the death against popery, without knowing whether popery was a
man or a horse. This, then, was a prejudice that they did not fill up of
their own heads. All the great points that men have founded a claim
to superiority, wisdom, and illumination upon, that they have
embroiled the world with, and made matters of the last importance,
are what one age and country differ diametrically with each other
about, have been successively and justly exploded, and have been the
levers of opinion and the grounds of contention, precisely because, as
their expounders and believers are equally in the dark about them,
they rest wholly on the fluctuations of will and passion, and as they
can neither be proved nor disproved, admit of the fiercest opposition
or the most bigoted faith. In what ‘comes home to the business and
bosoms of men,’ there is less of this uncertainty and presumption;
and there, in the little world of our own knowledge and experience,
we can hardly do better than attend to the ‘still, small voice’ of our
own hearts and feelings, instead of being browbeat by the effrontery,
or puzzled by the sneers and cavils of pedants and sophists, of
whatever school or description.
If I take a prejudice against a person from his face, I shall very
probably be in the right; if I take a prejudice against a person from
hearsay, I shall quite as probably be in the wrong. We have a
prejudice in favour of certain books, but it is hardly without
knowledge, if we have read them with delight over and over again.
Fame itself is a prejudice, though a fine one. Natural affection is a
prejudice: for though we have cause to love our nearest connections
better than others, we have no reason to think them better than
others. The error here is, when that which is properly a dictate of the
heart passes out of its sphere, and becomes an overweening decision
of the understanding. So in like manner of the love of country; and
there is a prejudice in favour of virtue, genius, liberty, which (though
it were possible) it would be a pity to destroy. The passions, such as
avarice, ambition, love, &c., are prejudices, that is amply exaggerated
views of certain objects, made up of habit and imagination beyond
their real value; but if we ask what is the real value of any object,
independently of its connection with the power of habit, or its
affording natural scope for the imagination, we shall perhaps be
puzzled for an answer. To reduce things to the scale of abstract
reason would be to annihilate our interest in them, instead of raising
our affections to a higher standard; and by striving to make man
rational, we should leave him merely brutish.
Animals are without prejudice: they are not led away by authority
or custom, but it is because they are gross, and incapable of being
taught. It is, however, a mistake to imagine that only the vulgar and
ignorant, who can give no account of their opinions, are the slaves of
bigotry and prejudice; the noisiest declaimers, the most subtle
casuists, and most irrefragable doctors, are as far removed from the
character of true philosophers, while they strain and pervert all their
powers to prove some unintelligible dogma, instilled into their minds
by early education, interest, or self-importance; and if we say the
peasant or artisan is a Mahometan because he is born in Turkey, or a
papist because he is born in Italy, the mufti at Constantinople or the
cardinal at Rome is so, for no better reason, in the midst of all his
pride and learning. Mr. Hobbes used to say, that if he had read as
much as others, he should have been as ignorant as they.
After all, most of our opinions are a mixture of reason and
prejudice, experience and authority. We can only judge for ourselves
in what concerns ourselves, and in things about us: and even there
we must trust continually to established opinion and current report;
in higher and more abstruse points we must pin our faith still more
on others. If we believe only what we know at first hand, without
trusting to authority at all, we shall disbelieve a great many things
that really exist; and the suspicious coxcomb is as void of judgment
as the credulous fool. My habitual conviction of the existence of such
a place as Rome is not strengthened by my having seen it; it might be
almost said to be obscured and weakened, as the reality falls short of
the imagination. I walk along the streets without fearing that the
houses will fall on my head, though I have not examined their
foundation; and I believe firmly in the Newtonian system, though I
have never read the Principia. In the former case, I argue that if the
houses were inclined to fall they would not wait for me; and in the
latter I acquiesce in what all who have studied the subject, and are
capable of understanding it, agree in, having no reason to suspect the
contrary. That the earth turns round is agreeable to my
understanding, though it shocks my sense, which is however too
weak to grapple with so vast a question.
ON PARTY SPIRIT
The Atlas.]
[April 25, 1830.
Party spirit is one of the profoundnesses of Satan, or in more

modern language, one of the dexterous equivoques and contrivances
of our self-love, to prove that we, and those who agree with us,
combine all that is excellent and praiseworthy in our own persons (as
in a ring-fence) and that all the vices and deformity of human nature
take refuge with those who differ from us. It is extending and
fortifying the principle of the amour-propre, by calling to its aid the
esprit de corps and screening and surrounding our favourite
propensities and obstinate caprices in the hollow squares or dense
phalanxes of sects and parties. This is a happy mode of pampering
our self-complacency, and persuading ourselves that we and those
that side with us, are ‘the salt of the earth;’ of giving vent to the
morbid humours of our pride, envy, and all uncharitableness, those
natural secretions of the human heart, under the pretext of self-
defence, the public safety, or a voice from Heaven, as it may happen;
and of heaping every excellence into one scale, and throwing all the
obloquy and contempt into the other, in virtue of a nickname, a
watch-word of party, a badge, the colour of a ribbon, the cut of a
dress. We thus desolate the globe, or tear a country in pieces, to show
that we are the only people fit to live in it; and fancy ourselves angels,
while we are playing the devil. In this manner, the Huron devours
the Iroquois, because he is an Iroquois, and the Iroquois the Huron
for a similar reason; neither suspects that he does it, because he
himself is a savage and no better than a wild beast; and is convinced
in his own breast that the difference of name and tribe makes a total
difference in the case. The Papist persecutes the Protestant, the
Protestant persecutes the Papist in his turn; and each fancies that he
has a plenary right to do so, while he keeps in view only the offensive
epithet which ‘cuts the common link of brotherhood between them.’
The church of England ill-treated the Dissenters, and the Dissenters,
when they had the opportunity, did not spare the church of England.
The Whig calls the Tory a knave, the Tory compliments the Whig
with the same title, and each thinks the abuse sticks to the party-
name, and has nothing to do with himself or the generic name of
man. On the contrary, it cuts both ways; but while the Whig says
‘The Tory is a knave, because he is a Tory,’ this is as much as to say, ‘I
cannot be a knave, because I am a Whig;’ and by exaggerating the
profligacy of his opponent, he imagines he is laying the sure
foundation, and raising the lofty superstructure of his own praises.
But if he says, which is the truth, ‘The Tory is not a rascal because he
is a Tory, but because human nature in power, and with the
temptation, is a rascal,’ then this would imply that the seeds of
depravity are sown in his own bosom, and might shoot out into full
growth and luxuriance if he got into place, which he does not wish to
appear till he does get into place.
We may be intolerant even in advocating the cause of Toleration,
and so bent on making proselytes to Free-thinking as to allow no one
to think freely but ourselves. The most boundless liberality in
appearance may amount in reality to the most monstrous ostracism
of opinion—not in condemning this or that tenet, or standing up for
this or that sect or party, but in assuming a supercilious superiority
to all sects and parties alike, and proscribing in the lump and in one
sweeping clause all arts, sciences, opinions, and pursuits but our
own. Till the time of Locke and Toland a general toleration was never
dreamt of: it was thought right on all hands to punish and
discountenance heretics and schismatics, but each party alternately
claimed to be true Christians and orthodox believers. Daniel Defoe,
who spent his whole life, and wasted his strength in asserting the
right of the Dissenters to a toleration (and got no thanks for it but the
pillory), was scandalized at the proposal of the general principle, and
was equally strenuous in excluding Quakers, Anabaptists, Socinians,
Sceptics, and all who did not agree in the essentials of Christianity,
that is, who did not agree with him, from the benefit of such an
indulgence to tender consciences. We wonder at the cruelties
formerly practised upon the Jews: is there anything wonderful in it?
They were at the time the only people to make a butt and a bugbear
of, to set up as a mark of indignity and as a foil to our self-love, for
the feræ naturæ principle that is within us and always craving its
prey to hunt down, to worry and make sport of at discretion, and
without mercy—the unvarying uniformity and implicit faith of the
Catholic church had imposed silence, and put a curb on our jarring
dissensions, heart-burnings, and ill-blood, so that we had no
pretence for quarrelling among ourselves for the glory of God or the
salvation of men:—a Jordanus Bruno, an Atheist or sorcerer, once
in a way, would hardly suffice to stay the stomach of our theological
rancour, we therefore fell with might and main upon the Jews as a
forlorn hope in this dearth of objects of spite or zeal; or, as the whole
of Europe was reconciled in the bosom of holy mother church, went
to the holy land in search of a difference of opinion and a ground of
mortal offence; but no sooner was there a division of the Christian
world than Papist fell upon Protestant, Protestants upon schismatics,
and schismatics upon one another, with the same loving fury as they
had before fallen upon Turks and Jews. The disposition is always
there, like a muzzled mastiff—the pretext only is wanting; and this is
furnished by a name, which, as soon as it is affixed to different sects
or parties, gives us a license, we think, to let loose upon them all our
malevolence, domineering humour, love of power and wanton
mischief, as if they were of different species. The sentiment of the
pious English bishop was good, who, on seeing a criminal led to
execution, exclaimed, ‘There goes my wicked self!’
If we look at common patriotism, it will furnish an illustration of
party-spirit. One would think by an Englishman’s hatred of the
French, and his readiness to die fighting with and for his
countrymen, that all the nation were united as one man in heart and
hand—and so they are in war-time—and as an exercise of their
loyalty and courage; but let the crisis be over, and they cool
wonderfully, begin to feel the distinctions of English, Irish, and
Scotch, fall out among themselves upon some minor distinction; the
same hand that was eager to shed the blood of a Frenchman will not
give a crust of bread or a cup of cold water to a fellow-countryman in
distress; and the heroes who defended the wooden walls of Old
England are left to expose their wounds and crippled limbs to gain a
pittance from the passenger, or to perish of hunger, cold, and neglect
in our highways. Such is the effect of our boasted nationality: it is

An Alternative Approach To Lie Groups and Geometric Structures Ercument H Ortacgil Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

An Alternative Approach To Lie Groups and Geometric Structures Ercument H Ortacgil Full Chapter

Uploaded by

Copyright:

Available Formats

An Alternative Approach to Lie Groups

and Geometric Structures Ercüment H.

PART I Fundamental Concepts

PART II Some Consequences

PART III How to Generalize?

19 The Definition of a Prehomogeneous Geometry (PHG) . . . . . . . . . . . 171

Appendix: Torsion-Free Connections 201

Everything should be made as simple as possible

(c, b) ◦ (b, a) = (c, a) (1.2)

(b, a) ◦ (a, b) = (a, a) (1.3)

(a, b)−1 = (b, a)

(yi , zi ) ◦ (xi , yi ) = (xi , zi ) (1.5)

(xi , yi )−1 = (yi , xi ) (1.6)

∂wi ∂wi ∂ya ∂xb

which projects a 1-arrow from a to b to the 0-arrow (a, b) from a to b.

U1b,c × U1a,b −→ U1a,c (1.14)

U1a,b −→ U1b,a (1.17)

The local formulas for (1.16) and (1.18) now become

and the second equality holds if and only if we have

Proposition 1.2 The following are equivalent:

The Nonlinear Curvature

j1 ( f )p, f (p) = ε(p, f (p)) (2.1)

j1 ( f )x,f (x) = ε(x, f (x)), x ∈ Dom( f ) (2.2)

j1 ( f )x, f (x) = ε(x, f (x)), x∈U (2.3)

∂ 2f i ∂εji (x, f (x)) ∂εji (x, f (x)) ∂f a

where x ∈ U, and y ∈ V = f (U). Therefore, Rikj (x, f (x)) = 0, x ∈ U, is a necessary condi-

∂xc ∂xb a ∂wi

Definition 2.4 R is the nonlinear curvature of (M, ε).

An important property of R is given by the following proposition:

∂εai (x, y) a ∂εja (y, x)

Setting y = x in (2.7) and again using (1.27), we get

Local Lie Groups

1. If f , g ∈ S and Dom( f ) = Ran( g), then f ◦ g ∈ S .

∂zi ∂zi ∂ya

in view of (1.26), and therefore z = z(x) belongs to G . Inverting both sides

We call (fα , Uα ) a continuation of (f1 , U1 ) along C. Let (gβ , Vβ ), 1 ≤ β ≤ m, be a continu-

Proposition 3.5 If M is compact, then (M, ε, G ) is complete.

To prove Proposition 3.5, we need the following technical lemma:

Now we have the following important proposition:

the extension f is unique. We claim that f : M → M is a diffeomorphism. It is onto:

For any y ∈ M, we choose a path C from f (p) to y. By Proposition 3.2, (f −1 , f (U)) ∈ G .

is defined on U, maps a to b, and integrates the 1-arrow π −1 (ε)(a, b). Therefore,

Propositions 3.10 and 3.8 give the following corollary:

In this chapter, (M, ε, G ) is an LLG as defined in Chapter 3. If (M, ε, G ) is globalizable, then

Now we fix p, q ∈ U and define the map g on U by

so g(p) = q by (4.2). Now, g is a 1–1 map by Corollary 2.7. Further, it is a diffeomorphism,

In coordinates, (4.4) is given by

f (p, x, q) = f (p, x, f (p, p, q)) (4.8)

We differentiate (4.8) at x = p using (4.6). This gives

which proves the claim, since p is arbitrary.

ε is a local splitting defined on U × U.

f (p, r, x) = f (q, f (p, r, q), x) p, q, r, x, f (p, r, q) ∈ U (4.11)

which generalizes (4.2).

which is (4.10), in view of (4.6).

that maps f ∈ G to the unique g satisfying g(e) = f (e). Clearly, e (g ◦ f ) = e ( f ) ◦ e ( g)

We bow to a patron who gives us a good dinner and his countenance

Prejudice, in its ordinary and literal sense, is prejudging any

The most dangerous enemies to established opinions are those

It is a mistake, however, to suppose that all prejudices are false,

Party spirit is one of the profoundnesses of Satan, or in more

You might also like