0% found this document useful (0 votes)
326 views219 pages

(University Texts in The Mathematical Sciences) Manjusha Majumdar, Arindam Bhattacharyya - An Introduction To Smooth Manifolds-Springer (2023)

Uploaded by

Chahma youssaf
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
326 views219 pages

(University Texts in The Mathematical Sciences) Manjusha Majumdar, Arindam Bhattacharyya - An Introduction To Smooth Manifolds-Springer (2023)

Uploaded by

Chahma youssaf
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 219

Manjusha Majumdar

Arindam Bhattacharyya

An Introduction
to Smooth
Manifolds
An Introduction to Smooth Manifolds
Manjusha Majumdar · Arindam Bhattacharyya

An Introduction to Smooth
Manifolds
Manjusha Majumdar Arindam Bhattacharyya
Department of Pure Mathematics Department of Mathematics
University of Calcutta Jadavpur University
Kolkata, West Bengal, India Kolkata, West Bengal, India

ISBN 978-981-99-0564-5 ISBN 978-981-99-0565-2 (eBook)


https://doi.org/10.1007/978-981-99-0565-2

Mathematics Subject Classification: 58-01, 58AXX, 58A05, 58A10, 58A15, 58C15, 58C25

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Dedicated to
Gahana Majumdar and Bipasa
Bhattacharyya
Preface

This book is an outcome of lectures delivered by one of the authors to the post-
graduate students at the Department of Pure Mathematics, University of Calcutta,
India, for several years. The need of the students has motivated the authors to write a
textbook. The only prerequisites are good working knowledge of point-set topology
and linear algebra.
It is said that mathematics can be learnt by solving problems, not only by reading
it. To serve this purpose, this book contains a sufficient number of examples and
exercises after each section of every chapter. Some exercises are routine ones for the
general understanding of each section. We have given hints about difficult exercises.
Answers to all exercises are given at the end of each section. We hope that this
approach will help the readers for getting this beautiful subject accessible.
We do not believe that there can be any complete book on the topic of manifold.
We are sure our book is far from completion as such. However, we are equally sure
that our book has some exceptional merits, and students will be benefitted if they go
through the whole book with all exercises.
Chapter 1 is the study of calculus on Rn . We have started the first section on
smooth functions. The concept of the diffeomorphic function is as important as
diffeomorphic manifold. We have given a few exercises on that. Tangent vector is
one of the powerful concepts of studying geometry. It has been defined with respect
to a curve in Rn in the second section.
The germ of a function has been defined in the third section. The last two sections
are on inverse function theorem and implicit function theorem with examples and
exercises.
Chapter 2 is the study of manifold. We have defined topological manifold and
then smooth manifold in Sect. 2.3. Many exercises have been given for a better
understanding of the concept of atlas of smooth manifold. Germs on topological
manifold have been explained in Sect. 2.2. Stereographic projections and orientable
surfaces are two attractive concepts, which have been explained separately in Sect. 2.4
and 2.5, respectively. Product manifold has been explained separately in Sect. 2.6.
Smooth functions on smooth manifold have been explained with solved problems in
Sect. 2.7. In this section, we have included diffeomorphic smooth manifold. Tangent

vii
viii Preface

vector has been introduced with respect to a differentiable curve in Sect. 2.8. The next
section is the study of inverse function theorem for smooth manifold. Section 2.10
deals with vector field and its geometrical interpretation has been explained in the
next section. Section 2.12 deals with push-forward vector fields which lead to the
concept of submanifold and hence to critical points and regular points of the manifold.
We have discussed Submanifolds separately in the next section. Push-forward vector
fields give rise to another kind of vector field which has been explained with solved
problems in Sect. 2.14. Finally, the last section of this chapter gives the algebraic
interpretation of the vector field. It is also termed as flow while studying dynamical
system which is an interesting topic of mathematics.
Chapter 3 is the study of differential forms. Differential forms have wide appli-
cations in Lie group, differential topology, differentials and their multiple integrals
over a differentiable manifold. However, in this book, we have mainly considered
the first role played by differential forms on manifolds.
The first section of this chapter is on 1-form, which is also called cotangent space. It
can be thought of as dual vector space of tangent space of the manifold. Thus, tangent
space and cotangent space can be thought of as “siamese twins” at every point of
the manifold. Members of cotangent space are also called co-vectors. We have given
the formal definition of r-form (r > 1) in the next section. Differential r-forms are
tensor fields of type (0, r) which are skew-symmetric. They have wide applications
in thermodynamics. We have also studied exterior product or wedge product in this
section, which is nothing but the generalization of the concept of cross-product
between two vectors in 3-dimension. This beautiful concept was introduced by R.
G. Grassmann, which nowadays also called “Grassmann algebra” and the reason for
this name “algebra” has been explained in this section for those students who are of
inquisitive nature. Exterior derivative to a manifold is the same as that of “curl” to
R3. All the classical concepts, namely gradient and divergence, can be expressed in
terms of this concept. We have given the proof of existence and uniqueness of such
operations. Finally, pull-back differential form has been studied in the last section.
This pull-back operation and exterior differentiation commute each other which has
been explained by a theorem, followed by many exercises.
The last chapter is on Lie group. The Lie group structure allows us to discuss
continuity and differentiability in a group structure. It was introduced by Norwegian
mathematician S. Lie in the late nineteenth century. Lie groups play an important role
in modern geometry. They are the fundamental building blocks for Gauge theories.
The first section is the study of Lie group and the two C ∞ transformations on it.
The behaviour of a Lie group is determined by its behaviour in the neighbourhood
of its identity element, and hence a famous theorem has been studied in the next
section. Due to the two translations, two types of invariant vector fields occur in this
group. Naturally, two types of invariant differential forms are also there. Well-known
theorems and results have been studied in the next two sections.
For the unique structure of a Lie group, one should have a natural quest for studying
group homomorphism and algebra homomorphism on it. The unique feature is the
study of one-parameter group of transformations induced by the invariant vector field
of a Lie group. Section 4.6 is the study of the action of a Lie group on a manifold.
Preface ix

We have gone through many books and articles during the preparation of this
book, which have been listed in Bibliography.
We wish to express our sincere acknowledgement to our respected teachers,
colleagues and friends for their valued suggestions.
One of the authors of this book would like to express her deep gratitude to her
deceased teacher, Prof. M. C. Chaki, University of Calcutta. His wonderful lectures
on this beautiful subject always generated great excitement and enthusiasm for the
students. Moreover, his affection, guidance and blessings towards her are memorable
forever.
We invite suggestions and comments from our readers for incorporation, if any,
in future editions.
We wish to thank Dr. S. Kundu, Assistant Professor, Loreto College, Kolkata,
India, for his excellent work on typing the entire manuscript. He has offered some
valuable suggestions also.
We also wish to thank the budding mathematician Mr. Sumanjit Sarkar, Research
Scholar, under Prof. A. Bhattacharyya, for his tireless effort in re-checking the whole
manuscript.
Finally, we acknowledge Mr. Shamim Ahmed, Springer, New Delhi, who took all
necessary care and perseverance for the preparation of this book.
Both authors also wish to convey their regards to the respected reviewers for the
enrichment of this book.

Kolkata, India Manjusha Majumdar


2022 Arindam Bhattacharyya
Contents

1 Calculus on Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Smooth Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Tangent Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Germ of a Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Inverse Function Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Implicit Function Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Manifold Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.1 Topological Manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.2 Smooth Germs on a Topological Manifold . . . . . . . . . . . . . . . . . . . . 40
2.3 Smooth Manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.4 Stereographic Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5 Orientable Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.6 Product Manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.7 Smooth Function on Smooth Manifold . . . . . . . . . . . . . . . . . . . . . . . 64
2.8 Differential Curve and Tangent Vector . . . . . . . . . . . . . . . . . . . . . . . . 69
2.9 Inverse Function Theorem for Smooth Manifold . . . . . . . . . . . . . . . 74
2.10 Vector Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.11 Integral Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.12 Differential of a Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.13 Submanifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
2.14 f -Related Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
2.15 One Parameter Group of Transformations on a Manifold . . . . . . . . 114
3 Differential Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.1 Cotangent Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.2 r -form, Exterior Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.3 Exterior Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
3.4 Pull-Back Differential Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

xi
xii Contents

4 Lie Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175


4.1 Lie Group, Left and Right Translation . . . . . . . . . . . . . . . . . . . . . . . . 175
4.2 Invariant Vector Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
4.3 Invariant Differential Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
4.4 Automorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
4.5 One-Parameter Subgroup of a Lie Group . . . . . . . . . . . . . . . . . . . . . . 200
4.6 Lie Transformation Group (Action of a Lie Group
on a Manifold) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
About the Authors

Manjusha Majumdar, Ph.D., is former Professor of Pure Mathematics at the


University of Calcutta, India. Her main interest lies in smooth manifolds. She has
published a number of research papers in several international journals of repute.
Under her supervision, 10 students have already been awarded their Ph.D. degrees.
She is a member of many national and international mathematical societies and serves
on the editorial board of several journals of repute. She has visited several institutions
in India and abroad on invitation. She is the Principal Investigator of the prestigious
e-PG Pathshala in Mathematics and MOOC in Mathematics, recommended by the
Ministry of Human Resource Development, Government of India. She co-authored
a book entitled Differential Geometry.

Arindam Bhattacharyya, Ph.D., is Professor, Department of Mathematics,


Jadavpur University, Kolkata, India. His main interest lies in smooth manifolds,
geometric flows and computational geometry. He has published a number of research
papers in several international journals of repute. Under his supervision, 19 research
scholars have already been awarded their Ph.D. degrees and 8 are pursuing their
research. He is a member of many national and international mathematical societies
and serves on the editorial board of several journals of repute. He has visited several
institutions in India and abroad on invitation. He has organized several international
conferences and research schools in collaboration with internationally reputed math-
ematical organizations. He is the Course Coordinator of Differential Geometry of
the prestigious e-PG Pathshala in Mathematics and MOOC in Mathematics, recom-
mended by the Ministry of Human Resource Development, Government of India, and
of Netaji Subhas Open University, India. He co-authored a book entitled Differential
Geometry with Prof. Manjusha Majumdar.

xiii
List of Figures

Fig. 1.1 Tangent plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10


Fig. 1.2 The inverse function theorem of a single variable . . . . . . . . . . . . 16
Fig. 1.3 The inverse function theorem of double variables . . . . . . . . . . . . 17
Fig. 1.4 Inverse function theorem on Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Fig. 1.5 Implicit Function Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Fig. 2.1 Locally Euclidean space of dimension n . . . . . . . . . . . . . . . . . . . . 36
Fig. 2.2 Overlap region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Fig. 2.3 Double cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Fig. 2.4 Cross(i) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Fig. 2.5 Cross(ii) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Fig. 2.6 Germ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Fig. 2.7 ϕ ◦ φ −1 is C ∞ on φ(U ∩ V ∩ W ) . . . . . . . . . . . . . . . . . . . . . . . . . 50
Fig. 2.8 Not atlas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Fig. 2.9 Atlas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Fig. 2.10 Stereographic projection on S 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Fig. 2.11 Stereographic projection on S 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Fig. 2.12 Möbius band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Fig. 2.13 Product manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Fig. 2.14 Smooth function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Fig. 2.15 Differentiable mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Fig. 2.16 Differentiable map between punctured sphere and cylinder . . . . 68
Fig. 2.17 Differential curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Fig. 2.18 Integral curves of the vector field X = −y ∂∂x + x ∂∂y . . . . . . . . . . 85
Fig. 2.19 Push-forward mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Fig. 2.20 Geometrical interpretation of f ∗ (X p ) . . . . . . . . . . . . . . . . . . . . . . 93
Fig. 2.21 Local one-parameter group of transformations . . . . . . . . . . . . . . . 120
Fig. 2.22 Existence theorem of local one-parameter group
of transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Fig. 3.1 Pull-Back Differential Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Fig. 4.1 g∼= Te (G) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Fig. 4.2 Lie Algebra Homomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

xv
Chapter 1
Calculus on Rn

1.1 Smooth Functions

Let R denote the set of real numbers. For an integer n > 0, let Rn denote the set
of all ordered n-tuples (x 1 , x 2 , . . . , x n ) of real numbers. Individual n-tuple will be
denoted at times by a single letter, e.g. x = (x 1 , x 2 , . . . , x n ), y = (y 1 , y 2 , . . . , y n )
and so on.
A real-valued function f : U ⊂ Rn → R, U being an open set, is said to be of
class C k if the following conditions hold:

(i) all its partial derivatives of the order less than or equal to k exist, and
(ii) are continuous functions on U .

By class C ∞ , we mean that all orders of partial derivatives of f exist and are contin-
uous at every point of U . A function of class C ∞ is also called a smooth function.
Actually, “Smoothness” is a synonym for C ∞ . By class C 0 , we mean that f is merely
continuous from U to R. By class C ω on U , we mean that f is real analytic on U .
A C ω function is C ∞ function but the converse is not true.
1
Example 1.1 Let f : R → R be defined by f (x) = x 3 . Then
1
x − 3 , x = 0;
2

f (x) = 3
0, x = 0.

Hence, f is C 0 but not C 1 .

Example 1.2 The polynomial, sine, cosine and exponential functions on the real
line are all C ∞ , which are also analytic, i.e.C ω .

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 1
M. Majumdar and A. Bhattacharyya, An Introduction to Smooth Manifolds,
https://doi.org/10.1007/978-981-99-0565-2_1
2 1 Calculus on Rn

Problem 1.1 Let f : R → R be defined by



e− x 2 , x = 0;
1

f (x) =
0, x = 0.

Show that f is a function of class C ∞ .

Solution: Note that

e− h 2
1
 f (0 + h) − f (0)
f (0) = lim = lim .
h→0 h h→0 h

Put h = u1 , then

∞
f (0) = lim ue−u
2
.
u→∞ ∞
Using L’Hospital’s rule, we find

e−u
2
 1
f (0) = lim u 2 = lim = 0.
u→∞ 2ue u→∞ 2u

Again
 2 − 12
f (x) = e x .
x3
Therefore
2 − h2
1
2e−u
2
 h3
e 1
f (0) = lim = lim 1
, taking u = .
h→0 h u→∞
u4
h

Using L’Hospital’s Rule successively, we will find

 24u
f (0) = 2 lim .
u→∞ 4ueu 2 + 8ueu 2 + 8u 3 eu 2

Finally, we will find



f (0) = 0.

Proceeding in the same manner, we will find

f n (0) = 0, n = 1, 2, 3, . . . . . . .

Hence, we claim that f is a function of class C ∞ .


Example 1.3 The function defined in Example 1.1 above, does not have power series
expansion at x = 0. Hence it is not a C ω function.
1.1 Smooth Functions 3

Problem 1.2 Consider the functions



0, x ≤ 1;
f 1 (x) = − (x−1)
1
e 2
, x > 1.

f 2 (x) = 0, for − ∞ < x < ∞.

Prove that f 1 , f 2 are differentiable on R.

Solution: For x ≤ 1, f 1 (x) = 0, so f 1 (1) = 0. For x > 1, f 1 (x) = 2(x − 1)e− x 2 .


1

Now

e− h 2
1

 f (1 + h) − f (1)
R f 1 (1) = lim = lim
h→0 h h→0 h
1
u2 1
= lim u , u = 2
u→∞ e h
= 0.
e− h 2
1
f (1 − h) − f (1)
L f 1 (1) = lim = lim = 0.
h→0 h h→0 h

Thus, f 1 (1) exists and continuous. So f 1 is differentiable on R.


As f 2 (x) is a constant function, it has finite derivative everywhere and so f 2 is
differentiable on R.
4
Problem 1.3 Let f : R → R be defined by f (x) = x 3 . Show that the function g
defined by  x  x
4 3 7
g(x) = f (t)dt = t 3 dt = x 3 ,
0 0 7

is C 2 but not C 3 .

Solution: Now g  (x) = f (x) = x 3 . Thus, g(x) is C 1 . Again g (x) = 43 x 3 , so g(x)
4 1

is C 2 but not C 3 at x = 0.
Exercises

Exercise 1.1 Let f : R → R be defined by



e− x , x > 0;
1

f (x) =
0, x ≤ 0.

Show that f is a function of class C ∞ .

Exercise 1.2 Define f : R → R by f (x) = x 3 . Is f −1 of class C ∞ ?


4 1 Calculus on Rn

Exercise 1.3 Let f : R → R be defined by



x + x 2 cos x1 , x = 0;
f (x) =
0, x = 0.

Show that

(i) f is continuous.
(ii) f is differentiable at all points.
(iii) the derivative is discontinuous at x = 0.

Exercise 1.4 Let f : R → R be defined by



x 2 sin x1 , x = 0;
f (x) =
0, x = 0.

Show that f is differentiable at x = 0 but f  is not continuous at x = 0.

Answers

1.2 No.

For i = 1, 2, . . . , n, let u i : Rn → R be the natural co-ordinate functions i.e.

u i ( p) = pi , where p = ( p 1 , p 2 , . . . , p n ) ∈ Rn . (1.1)

Such u i ’s are continuous functions from Rn to R and we call this n-tuple of functions
u 1 , u 2 , . . . , u n ; the standard co-ordinate system of Rn . If f : U ⊂ Rn → Rn is a
mapping, then f is determined by its co-ordinate functions f 1 , f 2 , . . . , f n where

f i = u i ◦ f, i = 1, 2, 3, . . . , n (1.2)

and each f i is a real-valued function. Thus

f ( p) = ( f 1 ( p), f 2 ( p), . . . , f n ( p)) ∀ p ∈ Rn . (1.3)

The map f is of class C ∞ if each component function f i , i = 1, 2, 3, . . . , n is also


so.
If f : U ⊂ Rn → Rm is a mapping such that
 
f (x 1 , x 2 , . . . , x n ) = f 1 (x 1 , x 2 , . . . , x n ), f 2 (x 1 , x 2 , . . . , x n ), . . . , f m (x 1 , x 2 , . . . , x n ) ,

we define the Jacobian matrix of f at (x 1 , x 2 , . . . , x n ), denoted by J , as


1.1 Smooth Functions 5
⎛ ⎞
∂f 1 ∂f 1
· · · ∂∂ xf n
1

∂x1 ∂x2
⎜ ∂f 2 ∂f 2 2 ⎟

J =⎜ · · · ∂∂ xf n ⎟ ,
∂x1 ∂x2 ⎟
⎝ ··· ··· ··· ··· ⎠
∂f ∂f
· · · ∂∂ fx n
m m m

∂x1 ∂x2

provided each of its first-order


 partial derivatives
 exist on U .
∂f ∂f ∂f
If m = 1, the matrix ··· is called the gradient of f , denoted
∂x1 ∂x2 ∂xn
by ‘grad f ’ or ∇ f .
The function f is said to be continuously differentiable on U , if each f i has
first-order continuous partial derivatives on U . If f is continuously differentiable on
U and the Jacobian is non-null, then f is one-to-one in U .
A mapping f : U ⊂ Rn → V ⊂ Rn , U, V being open sets in Rn , is said to be
homeomorphism if

(i) f is bijective and


(ii) f, f −1 are continuous.

Problem 1.4 Let f : R → R be such that f (x) = 5x + 3. Show that f is a home-


omorphism on R.

Solution: Here

f (x) − f (y) = 5(x − y)


⇒ f (x) = f (y) if and only if x = y.

Thus, f is one-to-one. Further to examine whether f is onto, we are to examine if


there exists a pre-image x of y under f such that f (x) = y holds. So y has a pre-
image y−35
in the domain set R. Since y is arbitrary, each element in the domain set
R has a pre-image under f . Thus, f is onto. Hence, f is bijective so f −1 : R → R
exists, which is defined by
x −3
f −1 (x) = .
5

Here f −1 is continuous. Also, f is continuous. Consequently, f is a homeomorphism.

Problem 1.5 Let f : R → R be such that f (x) = x 3 . Test

(i) whether f is of class C ∞ or not.


(ii) whether f is a homeomorphism or not.
6 1 Calculus on Rn

Solution:

(i) Here

f (0) = lim h 2 = 0
h→0

f (x) = 3x 2 .

Thus f has a finite derivative and f is of class C ∞ .


(ii) Here

f (x) − f (y) = x 3 − y 3
⇒ f (x) = f (y) if and only if x = y.

Thus, f is one-to-one. Further to examine whether f is onto, we are to examine


if there exists a pre-image x of y under f such that f (x) = y holds. So y has
1
a pre-image y 3 in the domain set R. Since y is arbitrary, each element in the
domain set R has a pre-image under f . Thus, f is onto. Hence, f is bijective so
f −1 : R → R exists, which is defined by

f −1 (x) = x 3 .
1

Here f −1 is continuous. Also, f is continuous. Consequently, f is a homeomor-


phism.

Exercises

Exercise 1.5 Let f : R → R be such that f (x) = x 2 . Is f a homeomorphism on


R?
x
Exercise 1.6 The function f : (−1, 1) → R is defined by f (x) = . Is f a
1 − x2
homeomorphism on (−1, 1)?
 π π
Exercise 1.7 Let f : − , → R be such that f (x) = tan x. Show that f is a
2 2
homeomorphism.

Exercise 1.8 Let f : R2 → R2 be such that φ(u, v) = (veu , u). Is φ a homeomor-


phism on R2 ?

Exercise 1.9 Consider the mapping φ : R3 → R3 given by

φ(x 1 , x 2 , x 3 ) = (x 1 + 1, x 2 + 2, x 3 + 3).

Show that φ is a homeomorphism.


1.1 Smooth Functions 7

Answers
1.5 No. 1.6 Yes. 1.8 Yes.
Remark 1.1 A homeomorphism between open subsets of Rn and Rm , n = m is not
possible. For details, refer to any book on topology.
A mapping f : U ⊂ Rn → V ⊂ Rn , U, V being open sets in Rn , is said to be dif-
feomorphism if
(i) f is a homeomorphism of U onto V and
(ii) f, f −1 are of class C ∞ .
Problem 1.6 Let f : R → R be such that f (x) = x 3 . Test whether f is a diffeo-
morphism or not.
Solution: We have shown in Example 1.5 that f is a homeomorphism and f is of
class C ∞ . Now
f −1 (x) = x 3 .
1

Such f −1 is not C 1 , as ( f −1 (x)) = 13 x − 3 , which is not defined at x = 0. Thus f −1


2

is not of class C ∞ . Consequently, f is not a diffeomorphism.


Problem 1.7 Let f : R2 → R2 be such that f (u, v) = (uev + v, uev − v). Show
that f is a diffeomorphism.
Solution: Note that
 1   
 ∂f   v v
  e ue + 1 
∂f 1

|J | =  ∂∂uf 2 ∂v = v v = −2ev = 0, ∀ v.
 ∂f 2  e ue − 1 
∂u ∂v

Hence, f is invertible. If
2 2
(ξ, η) = f (x 1 , x 2 ), then ξ = x 1 e x + x 2 , η = x 1 e x − x 2 .
2
Thus ξ + η = 2x 1 e x and ξ − η = 2x 2 . Consequently,

1 1
(ξ − η) and x 1 = (ξ + η)e 2 (η−ξ ) .
1
x2 =
2 2
We now define  
−1 1 η−ξ 1
f (ξ, η) = (ξ + η)e , (ξ − η) .
2
2 2

Also, f, f −1 is continuous, hence homeomorphism. Also, both f, f −1 are of class



C . So f is diffeomorphism.
Problem 1.8 Let f : R3 → R3 be such that f (u, v, ω) = (ueω + veω , ueω − veω ,
ω). Test whether f is a diffeomorphism or not at (1, 1, 0).
8 1 Calculus on Rn

Solution: Note that


 ω ω 
e e veω + ueω 
 ω
|J | =  e −eω −veω + ueω  = −e2ω = 0 at (1, 1, 0).
0 0 1 

Thus, f is invertible at (1, 1, 0). Again, if (ξ, η, θ ) = f (u, v, ω), then

1
ξ + η = 2ueω ⇒ u = (ξ + η)e−ω
2
1
ξ − η = 2veω ⇒ v = (ξ − η)e−ω
2
θ = ω.

We now define  
ξ + η −θ ξ − η −θ
f −1 (ξ, η, θ ) = e , e ,θ .
2 2

Also f, f −1 is continuous, hence homeomorphism at (1, 1, 0). Also, both are of class
C ∞ . So f is diffeomorphism at (1, 1, 0).
Exercises
 
Exercise 1.10 Let f : − π2 , π2 → R be defined by f (x) = tan x. Show that f is a
diffeomorphism.
Exercise 1.11 Let φ : R2 → R2 be defined by φ(u, v) = (veu , u). Determine
whether φ is a diffeomorphism or not.
Exercise 1.12 Consider the map f : R2 → R2 /{0, 0} defined by f (x, y) = (e x
sin y, e x cos y).
(i) Prove that Jacobian determinant of f does not vanish at any point of R2 .
(ii) Is f a diffeomorphism?
Exercise 1.13 Let φ : R3 → R3 be the map defined by
2 3
x 1 = e2x + e2x
1 3
x 2 = e2x − e2x
x 3 = x 1 − x 2.

Show that φ is a diffeomorphism.


Exercise 1.14 Consider the C ∞ function φ : R3 → R3 defined by

φ(x 1 , x 2 , x 3 ) = (x 1 cos x 3 − x 2 sin x 3 , x 1 sin x 3 + x 2 cos x 3 , x 3 ).

Prove that φ is a diffeomorphism from the unit sphere S 2 onto itself.
S2
1.2 Tangent Vector 9

Exercise 1.15 Prove that the mapping φ : R3 → R3 given by f (u, v, w) = (u sin ω


+ v cos ω, u cos ω − v sin ω, ω) is a diffeomorphism.

Answers
1.11 diffeomorphism. 1.12 (ii) No, f is not one-to-one.

1.2 Tangent Vector

It is known that in R3 , any line, say any curve γ (t), through a point p ∈ R3 , parallel
to a non-zero vector v has equation of the form

γ (t) = p + tv, t being the parameter.

Thus, we can write

γ (t) = ( p 1 + tv 1 , p 2 + tv 2 , p 3 + tv 3 ), p = ( p 1 , p 2 , p 3 ), v = (v 1 , v 2 , v 3 ).

Hence, any point, on this curve, has the co-ordinate, say (x 1 , x 2 , x 3 ), where

x i ≡ γ i (t) = ( pi + tv i ), i = 1, 2, 3.

Thus, in Rn , the curve through p = ( p 1 , p 2 , . . . , p n ), parallel to the direction of the


non-zero vector v = (v 1 , v 2 , . . . , v n ) is of the form

γ (t) = ( p 1 + tv 1 , p 2 + tv 2 , . . . , p n + tv n ), γ (0) = ( p 1 , p 2 , . . . , p n ) ≡ p ∈ Rn
x i ≡ γ i (t) = ( pi + tv i ), i = 1, 2, 3, . . . , n, t being the parameter.
(1.4)
Let f be a C ∞ function in a neighbourhood of p of Rn . Then the tangent vector
at p, in the direction of v, is defined to be the directional derivative, denoted by Dv f ,
as follows: 
f (γ (t)) − f (γ (0)) d 
Dv f = lim = f (γ (t))
t→0 t dt t=0

 ∂ f (γ (t))  dγ i (t) 
=  t = 0
∂γ i t=0 dt
 ∂ f ( p)
= v i , by (1.4).
∂xi
 ∂ f ( p)
Dv f = vi . (1.5)
∂xi
We also write
 ∂
Dv = vi ( p). (1.6)
∂xi
10 1 Calculus on Rn

Fig. 1.1 Tangent plane

Thus, the directional derivative acts as an operator on functions (Fig. 1.1).


Remark 1.2 Note that in R3 , a vector at p, is a tangent vector, defined by v say, to
a surface S in R3 , if it lies on the tangent plane at p.
Thus, the tangent space at p of Rn , denoted by Tp (Rn ), is the collection of all
tangent vectors v at p. Such a space is a vector space and hence from the Fundamen-
tal Theorem of finite-dimensional Vector Space, any v ∈ Tp (Rn ) can be expressed
 n
uniquely as v = v i ei , ∀ vi ∈ R, where {e1 , e2 , . . . , en } is a basis of Tp (Rn ).
i=1

Problem 1.9 Let v = (2, 3, 0) denote a vector in R3 . Find Dv f , for a fixed point
p = (−2, π, 1) where f : R3 → R is defined by f = x 1 x 3 cos x 2 .
Solution: In this case, p + tv = (−2 + 2t, π + 3t, 1). Therefore, f ( p + tv) =
2 cos 3t − 2t cos 3t. Hence,

d
Dv f = f ( p + tv) = −2.
dt t=0

Alternative Method
 ∂f
From (1.5), we obtain Dv f = vi ( p). In this case
i
∂xi

Dv f = 2(x 3 cos x 2 )(−2, π, 1) + 3(−x 1 x 3 sin x 2 )(−2, π, 1)


= −2.
  
sin 2t cos 2t x
Problem 1.10 Let γ (t) = be a curve in R3 with initial
cos 2t − sin 2t y
point p ∈ R3 be such that γ (0) = p. Find the velocity vector γ  (0) at p. Hence
compute Dv f , where f : R2 → R is defined by f = 2x + y 3 .
Solution: Note that
d d
γ (t) = (x sin 2t + y cos 2t, x cos 2t − y sin 2t) = (2x, −2y).
dt t=0 dt t=0
1.2 Tangent Vector 11

Thus, the velocity vector at p in R2 is given by

∂ ∂
2x − 2y .
∂x ∂y

Hence
 ∂f
Dv f = vi ( p), where v = (2x, −2y)
i
∂xi

is given by
Dv f = 4x − 6y 3 .

Alternative Method
In this case
 
p + tv = (x, y) + t (2x, −2y) = (1 + 2t)x, (1 − 2t)y .

Thus
f ( p + tv) = 2(1 + 2t)x + (1 − 2t)3 y 3 .

Consequently
d
Dv f = f ( p + tv)|t=0 = 4x − 6y 3 .
dt

Problem 1.11 Let p = (1, 1, 0) be a point in R3 and let

γ p (t) = (et , cos t, t), t ∈ R

be a curve with initial point p ∈ R3 . Find the velocity vector v in R3 and hence
compute Dv f , where f : R3 → R is defined by f = x z cos y.
Solution: Here
d  
γ p (t)t=0 = (et , − sin t, 1)t=0 = (1, 0, 1) = v.
dt

Thus the velocity vector at p in R3 is given by

∂ ∂
+ .
∂x ∂z

Hence

n
∂f
Dv f = vi ( p) = cos 1. (1.7)
i=1
∂xi
12 1 Calculus on Rn

Alternative
Here
p + tv = (1, 1, 0) + t (1, 0, 1) = (1 + t, 1, t).

Therefore,

d 
Dv f = f ( p + tv)|t=0 = (cos 1 + 2t cos 1)t=0 = cos 1.
dt

Exercises

Exercise 1.16 Let v = (2, −3, 4) denote a vector in R3 . For a fixed point p =
(2, 5, 7), compute Dv f where

(i) f : R3 → R is defined by f = x 3 y.
(ii) f : R3 → R is defined by f = z 7 .
(iii) f : R3 → R is defined by f = e x cos z.

Exercise 1.17 Let p = (x, y) be a point in R2 and let


  
sin t cos t x
γ p (t) = , t ∈R
cos t sin t y

be a curve with initial point p in R2 . Find the velocity vector v in R2 . Hence compute
Dv f , where
(i) f : R2 → R is defined by f = x 2 y.
(ii) f : R2 → R is defined by f = e x cos y.
(iii) f : R2 → R is defined by f = x cos y.

Exercise 1.18 Let p = (1, 0, 0) be a point in R3 and let

γ p (t) = (et , sin 2t, t), t ∈ R

be a curve with initial point p in R3 . Find the velocity vector v in R2 . Hence compute
Dv f , where
(i) f : R2 → R is defined by f = x 3 y.
(ii) f : R2 → R is defined by f = xe z .
(iii) f : R2 → R is defined by f = ye x cos z.

Answers
1.16 (i) 96 (ii) 4 · 77 (iii) 2e2 (cos 7 − 2 sin 7)
2
1.17 (i) x y (ii) e (x cos y + y sin y)
x
(iii) x cos y + x y sin y
1.18 (i) 2 (ii) 2 (iii) 2e
1.3 Germ of a Function 13

1.3 Germ of a Function

Let us consider the set of all pairs ( f, U ) where f : U ⊂ Rn → R is a C ∞ function


and U is the neighbourhood of a point p of Rn . If (g, V ) is another such pair, then
we define an equivalence relation in this way:
( f, U ) is equivalent to (g, U ), symbolically ( f, U ) ∼ (g, V ) if there exists an
open set W such that f = g , where p ∈ W ⊂ U ∩ V . It can be shown that ‘∼’
W
is an equivalence relation. The equivalence class ( f, U ) is called the germ of f at
p ∈ U ⊂ Rn . We write it as F( p). It can be shown that such F( p) of Rn is

(i) an algebra over R.


(ii) a module over R,
where the defining relations are

⎨ ( f + g)( p) = f ( p) + g( p),
( f g)( p) = f ( p)g( p), (1.8)

(λ f )( p) = λ f ( p), ∀ f, g ∈ F( p), λ ∈ R.

The proof, of (i) and (ii) stated above, is beyond the scope of this book and hence it
is left to the reader.
From (1.5), we now observe that

Dv ( f g)( p) = (Dv f )g( p) + f ( p)(Dv g). (1.9)

Such an equation is also known as Leibnitz Product Rule.


Problem 1.12 Let v = (x 3 , −yz, 0) denote a vector in R3 . For a fixed point p =
(x, y, z), compute Dv ( f g) where f : R3 → R and g : R3 → R are defined respec-
tively by f = x z and g = y 2 .
Solution: We know from (1.9) that

Dv ( f g)( p) = (Dv f )g( p) + f ( p)Dv g. (1.10)

Now
p + tv = (x, y, z) + t (x 3 , −yz, 0) = (x + t x 3 , y − t yz, z).

Thus
f ( p + tv) = (x + t x 3 )z = x z + t x 3 z.

Therefore
14 1 Calculus on Rn

d 
Dv f = f ( p + tv)t=0 = x 3 z,
dt
g( p) = y 2 , f ( p) = x z,
g( p + tv) = (y − t yz)2 .

Therefore
d 
Dv g = g( p + tv)t=0 = −2y 2 z.
dt
Thus, from (1.10), we have

Dv ( f g)( p) = x 3 y 2 z − 2x y 2 z 2 .

Alternative
Here
 ∂f
Dv f = vi ( p) = x 3 z.
∂xi
Also
 ∂g
Dv g = vi ( p) = −2y 2 z.
∂xi
Using (1.10), we get the desired result.
Problem 1.13 Compute Dv ( f g) where f : R3 → R and g : R3 → R are defined
respectively as f = x y 2 − yz 2 , g = xe y and v = (−1, 2, 1) denotes a vector in R3 ,
for a fixed p = (2, −2, 1).
Solution: Here

p + tv = (2 − t, −2 + 2t, 1 + t)
f ( p) = 10
f ( p + tv) = (2 − t)(2t − 2)2 − 2(t − 1)(t + 1)2 .

Now
d 
f ( p + tv)t=0 = Dv f = −18.
dt
Again
g( p) = 2e−2 , g( p + tv) = (2 − t)e2t−2 .

So
d 
g( p + tv)t=0 = Dv g = 3e−2 .
dt
Thus
Dv ( f g)( p) = −6e−2 .
1.4 Inverse Function Theorem 15

Exercises

Exercise 1.19 For any constant function C, prove that Dv C = 0.

Exercise 1.20 Let v = (x, −y) denote a vector in R2 . For a fixed point p = (x, y),
compute Dv ( f g) where f : R2 → R and g : R2 → R are defined respectively by:
(i) f = x 2 y and g = e x cos y
(ii) f = e x sin y and g = x cos y
(iii) f = xe y and g = ye x .

Exercise 1.21 Let v = (1, −2, 1) denote a vector in R3 . For a fixed point p =
(2, −2, 1), compute Dv ( f g), where f : R3 → R and g : R3 → R are defined respec-
tively by:
(i) f = x 2 yz and g = e x cos y
(ii) f = e x sin y and g = x z cos y

Answers
(1.19) Use (1.9) and R-linearity property.
(1.20) (i) (x 2 y cos y + x 3 y cos y + x 2 y 2 sin y)e x . (ii) e x (x 2 + x) sin y cos y −
e x x y cos 2y.
(iii) (x 2 y − x y 2 )e x+y .
(1.21) (i) 16e2 (sin 2 − 2 cos 2). (ii) −e2 (5 cos 2 sin 2 + 4 cos 4).

1.4 Inverse Function Theorem

Suppose U be some open subset of the Euclidean space Rn and the non-linear map-
ping F : U → Rn is continuously differentiable. Let x̃ ∈ U . Suppose that, at the

point the differential F (x̃) : Rn → Rn is one-to-one and onto. This implies that the
non-linear map F inherits local invertibility in the vicinity of the point x̃. Precisely,
we can say that ∃ an open subset V of Rn such that x̃ ∈ V and an open subset W of Rn
satisfying F : V → W is one-to-one and onto. Also, the inverse F −1 is continuously
differentiable. This originates the notion of Inverse Function theorem.
Theorem 1.1 (Inverse Function Theorem of a single variable) Let U ⊆ R be open
and suppose that the function F : U → R is a continuously differentiable function.
Let a ∈ U such that f  (a) = 0. Then there exists an open interval I containing the
point a and an open interval J containing its image f (a) such that the function f :
I → J is one-to-one and onto. Moreover, the inverse function theorem f −1 : J → I
is also continuously differentiable, and for a point y in J , if x is a point in I at which
f (x) = y, then (Fig. 1.2)
 −1  1
f (y) =  .
f (x)
16 1 Calculus on Rn

Fig. 1.2 The inverse


function theorem of a single
variable

Proof Suppose f  (a) > 0. Since a is an interior point of U and the function f  :
U → R is continuous, therefore ∃ a real quantity s > 0 such that the closed interval
[a − s, a + s] ⊂ U and f  (x) > 0 for all points x ∈ [a − s, a + s]. By virtue of
the Mean value theorem, we can say that the function f : [a − s, a + s] → R is
strictly increasing. In particular, f : [a − s, a + s] → R is one-to-one. Furthermore,
taking into consideration the Intermediate Value Theorem, if the point y lies between
f (a − s) and f (a) + s, ∃ x ∈ (a − s, a + s) with f (x) = y. Let us define I = (a −
s, a + s) and J = ( f (a) − s, f (a) + s) = (b − s, b + s) where b = f (a). Then the
function f : I → J is one-to-one and onto.
For the concluding part of the theorem, it follows from the Intermediate Value
theorem that J is a neighbourhood of b. For y ∈ J , with y = b define x ≡ f −1 so
that
f −1 (y) − f −1 (b) 1
= f (x)− f (a) .
y−b x−a

Since the inverse function f −1 : J → R is continuous, therefore

lim x ≡ lim f −1 (y) = f −1 (b) ≡ a.


y→b y→b

By the composition property for limits, the quotient property of limits, and the defi-
nition of the differentiability of f : I → J at a, it follows that

f −1 (y) − f −1 (b) 1 1
lim = lim f (x)− f (a)
= .
y→b y−b y→b
x−a
f  (a)

 
Thus f −1 is differentiable at b, and its derivative at b is given by f −1 (y) = 1
f  (x)
.

Theorem 1.2 (Inverse Function Theorem in the plane) Let U (⊆ R2 ) open and sup-
pose that the mapping F : U → R2 is continuously differentiable. Let (a, b) ∈ U
1.4 Inverse Function Theorem 17

Fig. 1.3 The inverse function theorem of double variables


such that the derivative matrix F (a, b) be invertible. Then ∃ a neighbourhood V of
(a, b) and a neighbourhood W of its image F(a, b) such that F : V → W is one-
to-one and onto. Moreover, the inverse mapping F −1 : W → V is also continuously
differentiable, and for a point (u, v) ∈ W , if (x, y) ∈ V such that F(x, y) = (u, v),
then the derivative matrix of the inverse mapping at the point (u, v) is given by the
formula (Fig. 1.3)
 
(F −1 ) (u, v) = F (x, y)−1 .

Observe that in the proof of the last theorem, we used the Intermediate Value
Theorem, a result that does not easily generalize to mappings whose image lies in
the plane R2 . An n × n matrix is invertible if and only if its determinant is non-zero,
and when the matrix is invertible, there is a formula called Cramer’s Rule for the
inverse matrix. For 2 ×2 matrices,
 Cramer’s Rule is clear by inspection.
 Indeed,

a11 a12 −1 a −a
for a 2 × 2 matrix A = , if det A = 0 then A = det A 1 22 12
. In
a21 a22 −a21 a11
particular, for the mapping F : U → R in the statement of the Inverse Function
2
 
Theorem in the Plane F (a, b) holds if and only if det F (a, b) = 0. If the mapping
F : U → R is represented in terms of component function as
2

F(x, y) = (F1 (x, y), F2 (x, y)), (x, y) ∈ U,

then ⎛ ∂F ∂ F1

∂x
1
(x, y) ∂y
(x, y)
F (x, y) = ⎝ ∂ F ⎠.


∂ F2
∂x
2
(x, y) ∂y
(x, y)


So the assumption det F (a, b) = 0 is equivalent to

∂ F1 ∂F ∂ F1 ∂F
(a, b) 2 (a, b) − (a, b) 2 (a, b) = 0.
∂x ∂y ∂y ∂x
18 1 Calculus on Rn

The above explicit formula for the inverse of a 2 × 2 matrix permits us to use for-

mula (F −1 ) (u, v) = F  (x, y)−1 to compute the partial derivatives of the component
functions of the inverse mapping F −1 : W → V . Indeed, write the inverse mapping
in component functions as

F −1 (u, v) = (g(u, v), h(u, v)), (u, v) ∈ W,

such that  ∂g 
 (u, v) ∂g (u, v)
(F −1 ) (u, v) = ∂u
∂h
∂v
∂h .
∂u
(u, v) ∂v
(u, v)

For a point (u, v) ∈ W , let (x, y) ∈ V at which u = F1 (x, y), v = F2 (x, y). For

notation, set J (x, y) ≡ F (x, y). Then, using the above computation of the inverse

of a 2 × 2 matrix, it follows that formula F −1 (u, v) = [F (x, y)]−1 is equivalent to

∂g 1 ∂ F2
(u, v) = · (x, y)
∂u J (x, y) ∂ y
∂g 1 ∂ F1
(u, v) = − · (x, y)
∂v J (x, y) ∂ y
∂h 1 ∂ F2
(u, v) = − · (x, y)
∂u J (x, y) ∂ x
∂h 1 ∂ F1
(u, v) = · (x, y).
∂v J (x, y) ∂ x

Example 1.4 For a point (x, y) ∈ R2 , let us define

F(x, y) = (exp(x − y) + x 2 y + x(y − 1)5 , 1 + x 2 + x 4 + (x y)5 ).

Since each of its component functions is continuously differentiable, therefore, the


mapping F : R2 → R2 is continuously differentiable. At the point (a, b) = (1, 1),
we have  
 3 0
F (1, 1) = .
11 5

The determinant of F (1, 1) is non-zero. In view of the Inverse Function Theorem, ∃
neighbourhoods V of the (1, 1) and W of its image (2, 4) such that the mapping F :
U → V is one-to-one and onto and that the inverse mapping F −1 : V → U is also
continuously differentiable. Moreover, if the inverse is represented in components
as F −1 (u, v) = (g(u, v), h(u, v)), then it follows that

∂g 1 ∂g ∂h 11 ∂h 1
(2, 4) = , (2, 4) = 0, (2, 4) = − , (2, 4) = .
∂u 3 ∂v ∂u 15 ∂v 5
1.4 Inverse Function Theorem 19

Example 1.5 Let us define the function f : R2 → R2 by

f (x, y) = (cos(x 2 + y), sin(x 2 + y)), (x, y) ∈ R2 .

Then f is continuously differentiable as each component function is also so. But 


is any point at which the conclusion of the Inverse Function Theorem holds true. To
see this, observe that if (u, v) ∈ f (R2 ), then u 2 + v 2 = 1, a circle of radius 1 with
centre at the origin. The image does not contain an open subset of the plane, so 
any open sets U and V in R2 such that f : U → V is one-to-one and onto.

Example 1.6 Define the function f : R2 → R2 by

f (x, y) = (x 2 − y 2 , 2x y), (x, y) ∈ R2 .

Since each of its component function is obviously continuously differentiable, there-


fore, f is also so. Consider (a, b)(= (0, 0)) ∈ R2 . We have
 
 2a −2b
f (a, b) = ,
2b 2a

so | f (a, b)| = 4(a 2 + b2 ) = (0, 0). Applying Inverse Function Theorem, it follows
that there exist neighbourhoods U of (a, b) and V of f (a, b) such that the mapping
f : U → V is one-to-one and onto and has an inverse f −1 : V → U that also is
continuously differentiable. Suppose f −1 (ã, b̃) = (g(ã, b̃), h(ã, b̃)). Then, if we set
(ã◦ , b̃◦ ) = f (a, b), it follows that

∂g a ∂g b
(ã , b̃ ) = , (ã◦ , b̃◦ ) =
∂ ã ◦ ◦ 2(a 2 + b2 ) ∂ b̃ 2(a 2 + b2 )

∂h −b ∂h a
(ã , b̃ ) = , (ã◦ , b̃◦ ) = .
∂ ã ◦ ◦ 2(a 2 + b2 ) ∂ b̃ 2(a 2 + b2 )

But the assumptions of the Inverse Function Theorem fails to be true at the point
(0, 0), since  
 00
f (0, 0) = .
00

Moreover, Inverse Function Theorem also fails at this point because if f (x, y) =
f (−x, −y) holds for every (x, y) in the plane,  neighbourhood of (0, 0) on which
the mapping f is one-to-one.

Problem 1.14 The point (1, e) lies on the graph of y = xe x . Find an open set con-
taining y = e, such that ∃ is a continuous function x = g(y) defined on it, for which
x = g(y) ⇒ y = xe x and g(e) = 1.
20 1 Calculus on Rn

dy
Solution: Since = (1 + x)e x > 0 on (−1, ∞), the given function is injective
dx
when restricted to this interval and has range (−e−1 , ∞), which is an open subset U
of R containing e. Therefore, there is a continuous inverse g with domain U .

An Interpretation of the Inverse Function Theorem

Given two functions F1 : R2 → R and F2 : R2 → R and two numbers a1 and a2 ,


consider the system of equations

F1 (x, y) = a1 ; F2 (x, y) = a2 .

A natural question arises whether there exist any solutions to this system of equations
 is unique. If we2define the mapping F : R → R
2 2
and, if there beany, then the solution
by F(x, y) = F1 (x, y), F2 (x, y) for (x, y) ∈ R , these two questions about the
existence and uniqueness of the solutions of the system can be rephrased as questions
about the image of the mapping F : R2 → R2 and whether it has the property of
being one-to-one. The following example shows how the Inverse Function Theorem
provides information about systems of equations. Consider the system of equations

e x−y + x 2 y + x(y − 1)5 = 2; 1 + x 2 + x 4 + (x y)5 = 4.

Observe that the point (x, y) = (1, 1) is a solution of this system. The mapping
F : R2 → R2 defined by
 
F(x, y) = e x−y + x 2 y + x(y − 1)5 , 1 + x 2 + x 4 + (x y)5 .

for (x, y) in R2 is precisely the mapping considered in the Example 1.4. Referring
to Example 1.4, we can say that ∃ a δ > 0 and a neighbourhood U of the point (1, 1)
such that for any numbers a1 and a2 with (a1 − 2)2 + (a2 − 4)2 < δ 2 , the system of
equations

e x−y + x 2 y + x(y − 1)5 = a1 ; 1 + x 2 + x 4 + (x y)5 = a2

has exactly one solution.


Now we are going to state General Inverse Function Theorem.
Theorem 1.3 (Inverse Function Theorem on Rn ) Let U (⊆ Rn ) be open and suppose
that the mapping F : U → Rn is continuously differentiable. Let x∗ ∈ U at which

the derivative matrix F (x∗ ) is invertible. Then ∃ an open neighbourhood V of the
point x∗ and an open neighbourhood W of its image F(x∗ ) such that the mapping
F : V → W is one-to-one and onto. Moreover, the inverse mapping F −1 : W → V
1.4 Inverse Function Theorem 21

is also continuously differentiable, and for a point y ∈ W , if x is the point in V such


that F(x) = y, then
 
(F −1 ) (y) = F (F −1 (y))−1 .

Before we proceed with the proof of the Inverse Function Theorem for n-variables,
let us prove the following lemma:
Lemma 1.1 Suppose f : O(⊂ Rn ) → Rm is differentiable in a convex open set O

and there exists M ∈ R such that | f (x)| ≤ M, ∀ x ∈ O. Then

|| f (b) − f (a)|| ≤ M||b − a||, ∀ a, b ∈ O.

Proof Let us fix a, b in O and define γ : R → Rn , ∀ t ∈ R. Since O is convex,


γ (t) ∈ E provided t ∈ [0, 1]. Let us set g(t) = ( f ◦ γ )(t). Then
   
g (t) = f (γ (t))γ (t) = f (γ (t))(b − a),

which implies
  
|g (t)| ≤  f (γ (t))||b − a|| ≤ M||b − a||, ∀ t ∈ [0, 1].


This completes the proof of the lemma.

Proof of the Main Theorem:



Since F is continuous at x∗ , therefore, for a preassigned > 0 there exists an open
neighbourhood V ⊂ U of x∗ such that
 
x ∈ V ⇒ |F (x) − F (x∗ )| < . (1.11)

1 1
Let us choose = . Then the preceding equation becomes
2 |F  (x∗ )−1 |

  1    1
|F (x) − F (x∗ )| < ⇒ |F (x∗ )−1 | · |F (x) − F (x∗ )| < . (1.12)
|F  (x∗ )−1 | 2

Then we see that for x ∈ V , F (x) is invertible. Now, for any y ∈ Rn , let us define a
function ψ y : V → Rn by
      
ψ y (x) = x + F (x∗ )−1 y − F(x) = F (x∗ )−1 F (x∗ ) + y − F(x) . (1.13)

Then
x is a fixed point of ψ y ⇔ y = F(x). (1.14)
22 1 Calculus on Rn

Fig. 1.4 Inverse function theorem on Rn

Now as a consequence of the chain rule the composition function ψ y has


     
ψ y (x) = F (x∗ )−1 F (x∗ ) − F (x) .

Combining (1.12) with the last equation, we obtain

 1
|ψ y (x)| < , x ∈ V. (1.15)
2
Hence, by virtue of Lemma 1.1, it follows that

||ψ y (x1 ) − ψ y (x2 ))|| ≤ ||x1 − x2 ||, ∀ x1 , x2 ∈ V. (1.16)

Thus, ψ y has at most one fixed point in V so that by (1.15), F(x) = y holds for at
most one x ∈ V . This proves F is injective on V .
Next, consider F(V ) = W . Then F : V → W is an injective map. Therefore,
there exists an inverse map F −1 : W → V as illustrated in the figure below (Fig. 1.4):
In order to complete the proof of the first part of the theorem, it only remains to
show W is open. Choose w◦ ∈ W be arbitrary. Then for some x◦ ∈ V , F(x◦ ) = w◦ .
Let > 0 be sufficiently small enough such that ||x − x◦ || ≤ ⇒ x◦ ∈ V , so that

B (x◦ ) = {x ∈ Rn : ||x − x◦ || ≤ } ⊂ V.

To show W is open, it suffices to show that

y ∈ Rn , ||y − w◦ || < ⇒ y ∈ W.
2|F  (x∗ )−1 |

In order to prove this, suppose ||y − w◦ || < 2|F  (x∗ )−1 |


. Then for any x ∈ B (x◦ ), we
find
1.4 Inverse Function Theorem 23

||ψ y (x) − x◦ || ≤ ||ψ y (x) − ψ y (x◦ )|| + ||ψ y (x◦ ) − x◦ ||


1 
≤ ||x − x◦ || + ||F (x∗ )−1 (y − F(x◦ ))||, by (1.13), (1.16)
2

≤ + |F (x∗ )−1 | ||y − w◦ ||
2

≤ + |F (x∗ )−1 |
2 2|F  (x∗ )−1 |
≤ ,

which proves ψ y (x) ∈ B (x◦ ). Consequently, by virtue of (1.16), ψ y : B (x◦ ) →


B (x◦ ) is a contraction map, where B (x◦ ) is closed in Rn . We can invoke contraction
principle in Rn , to conclude that ψ y has a unique fixed point in B (x◦ ) ⊂ V . Thus,
by (1.14), we have y ∈ F(B (x◦ )) ⊂ F(V ) = W . This completes the argument that
W is open.
Furthermore, for any > 0

||y − w◦ || < , y ∈ W ⇒ y ∈ F(B (x◦ ))


2|F  (x∗ )−1 |
⇒ y = F(x) for some x ∈ B (x◦ )
⇒ F −1 (y) = B (x◦ )
⇒ ||F −1 (y) − x◦ || ≤ ,

which proves the continuity of the inverse map F −1 .


To prove the second part of the theorem, let us proceed as follows:
Here for sufficiently small h, we have
  
F F −1 (y) + h − F(F −1 (y)) = F (F −1 (y))(h) + ||h||R(h), (1.17)

where R(h) being the remainder term and R(h) → 0 as ||h|| → 0. Note that,

F −1 (y) ∈ V ⇒ F (F −1 (y)) is invertible. Since F −1 is continuous, therefore, for
sufficiently small k, suppose h = F −1 (y + k) − F −1 (y). Then for every such k, we
find   
F F −1 (y + k) − F(F −1 (y)) = F (F −1 (y))(h) + ||h||R(h).

Applying F (F −1 (y))−1 to both sides of the last equality, we get
 
h = F (F −1 (y))−1 (k) − ||h||F (F −1 (y))−1 (R(h)), (1.18)

which is the same as


 
F −1 (y + k) − F −1 (y) = F (F −1 (y))−1 (k) − ||h||F (F −1 (y))−1 (R(h)).
24 1 Calculus on Rn

  
We claim the existence of (F −1 ) (y) and (F −1 ) (y) = F (F −1 (y)). For this, it suf-
fices to show that
||h||  −1
F (F (y))(R(h)) → 0 as ||k|| → 0.
||k||

By continuity of F −1 , we have ||k|| → 0 ⇒ ||h|| → 0. So it is sufficient to prove


||h||
||k||
remains bounded. Now (1.18) yields

 
||h|| ≤ |F (F −1 (y))| ||k|| + ||h|| |F (F −1 (y))| ||R(h)||,

which further implies

||h||    
1 − |F (F −1 (y))| ||R(h)|| ≤ |F (F −1 (y))|.
||k||

||h||
Since ||h|| → 0 ⇒ R(h) → 0, it follows that remains bounded. Thus, for every
||k||
  
y ∈ W , we prove the existence of (F −1 ) (y) and (F −1 ) (y) = F (F −1 (y)). Further-
 
more, the equality shows that (F −1 ) is the composition of F −1 , F and the inversion

maps. All these are continuous, and therefore, (F −1 ) is continuous. This establishes
the last part of the theorem.

Remark 1.3 1. In summary, F is locally invertible at x∗ with a continuously dif-


ferentiable local inverse or F has a continuously differentiable local inverse
at x∗ . The term local inverse here refers to the function F −1 .
2. In the proof of the above theorem, we have used the notion norm of a linear
mapping and the Contraction Principle. For details, the reader may refer to
any standard book on Linear Algebra and Multivariable Analysis.
3. The benefit behind using the contraction principle in the proof of Inverse Function
Theorem is that, it can be extended to the condition when Rn is replaced by an
infinite-dimensional space. Alternatively, it is possible to prove the theorem
using the compactness of a closed ball in Rn , where the above benefit fails to be
true.

Problem 1.15 Let U and V be open subsets of Rn and let F : U → V be contin-


uously differentiable and bijective, so that the inverse map F −1 : V → U exists.
 
Suppose F (x) is invertible for every x ∈ U . Show that (F −1 ) exists on the entire
given set V .

Solution: Let q ∈ V . Then q = F( p) for some p ∈ U . Since F ( p) is invertible,
in view of the inverse function theorem ∃ open sets U1 ⊂ U and V1 ⊂ Rn such that
p ∈ U1 and F(U1 ) = V1 and F has a differentiable local inverse on V1 . But then
q = F( p) ∈ V1 and the local inverse is, therefore, differentiable at q. However, F −1
has to agree with the local inverse, and therefore, differentiable at q.
1.4 Inverse Function Theorem 25

Problem 1.16 If F : E → Rn is a continuously differentiable mapping of an open



set E ⊂ Rn and if f (x) is invertible for every x ∈ E, then prove that f is an open
mapping of E into R n .

Solution: Let U ⊂ E be open and b ∈ F(U ). Then q = F( p) for some p ∈ U .



Since F ( p) is invertible, the inverse function theorem yields open sets U1 ⊂ U
and V ⊂ Rn such that p ∈ U1 and f (U1 ) = V . But then F( p) ∈ V ⊂ f (U ). Since
q = F( p) and V is open, implies f (U ) is open.
Remark 1.4 Here f : E → Rn is an open mapping means that f maps every open
subset of E into an open subset of R n .
Problem 1.17 Let U be an open subset of Rn and let f : U → Rn be a continuously

differentiable map such that f (x) is invertible for every x ∈ U . Suppose V is an
open subset of U such that its closure V̄ is bounded and contained in U , and f is
injective on the closure. Show that the image f (V̄ ) is the closure of an open set.

Solution: It is trivial to show that f (V̄ ) ⊂ f (V ). To prove the reverse inclusion,


consider any y ∈ f (V ). Then ∃ a sequence xn in V such that f (xn ) → y. Since
V̄ is bounded (hence compact), xn → x(∈ V̄ ) when xn is replaced by a suitable
subsequence. Since f is continuous, it follows that f (xn ) → f (x), so that y = f (x).
Since x ∈ V̄ , we have y ∈ f (V̄ ). So f (V ) ⊂ f (V̄ ), and hence f (V ) = f (V̄ ). By
virtue of the last problem, note that the set f (V ), of which f (V ) is the closure, is an
open set.
Exercises
Exercise 1.22 Define the function F : P → R2 , where P = {(x1 , x2 ) ∈ R2 : x1 =
0}, by

y1 = F1 (x1 , x2 ) = x1 cos x2 , y2 = F2 (x1 , x2 ) = x1 sin x2 , x1 = 0.

At what points (x1 , x2 ) ∈ P does the Inverse Function Theorem apply?


x2 y2
Exercise 1.23 Consider the equation + = 1, (x, y) ∈ R2 .
8 8
(i) Explicitly define the function g : I → R that has the property that in the neigh-
bourhood of the solution (2, 3), all the solutions are of the form (x, g(x)) for
x ∈ I and check that
(ii) Explicitly define the function h : J → R that has the property that in a neigh-
bourhood of the solution (2, 3), all the solutions are of the form (h(y), y) for y
in J .

Exercise 1.24 Let f : R → R such that f (x, y) = (x 2 + y, 2x + y 2 ). Find f and
determine the values of (x, y) for which f is NOT invertible. Given that f is invertible

at (0, 0), let g be its inverse. Find g (0, 0).
26 1 Calculus on Rn

1.5 Implicit Function Theorem

Let U be an open subset of the plane R2 and f : U → R is continuously differentiable


function. In general, the solution set of the equation is not the graph of a function
expressing y as a function of x. Hence, the solution set is a very complicated subset of
∂f
the plane. However, if the point (a, b) is a solution of this equation and (a, b) = 0,
∂y
then ∃ a neighbourhood of the point (a, b) with the property that the solutions of the
above equation that are in this neighbourhood make up the graph of a continuously
differentiable function g : I → R, where I is an open interval about a. Moreover,
the derivative of the implicitly defined function g : I → R can be computed in terms
of the partial derivatives of the function f : U → R. This concept is called Dini’s
Theorem. It has an extension, called the General Implicit Function Theorem, that
provides a similar local description of the set of solutions of an equation of the
form F(u) = 0, u ∈ U , where U is an open subset of Euclidean space Rn+k and the
mapping F : U → Rn+k is continuously differentiable.
Theorem 1.4 (Dini’s Theorem) Let U ⊆ R2 be open. Suppose that the function
f : U → R is continuously differentiable. Let (a, b) ∈ U such that f (a, b) = 0 and
∂f
(a, b) = 0. Then ∃ a positive real quantity r and a continuously differentiable
∂y
function g : I → R, where I ≡ (a − r, a + r ) such that

f (x, g(x)) = 0, ∀ x ∈ I (1.19)

and

whenever |x − a| < r, |y − b| < r and f (a, b) = 0, then y = g(x). (1.20)

Moreover,
∂f ∂f
(x, g(x)) + (x, g(x))g  (x) = 0, ∀x ∈ I. (1.21)
∂x ∂y

∂f ∂f
Proof Suppose (a, b) > 0. Since U is open and the function : U → R is
∂y ∂y
continuous and positive at the point (a, b), ∃ positive numbers m and n such that
the rectangular region R = [a − m, a + m] × [b − m, b + m] ⊂ U and

∂f
(x, y) ≥ n ∀ (x, y) ∈ R. (1.22)
∂y

With the aid of the Mean Value Theorem for real-valued functions, if |x − a| ≤
m & b − m ≤ y1 < y2 ≤ b + m holds, then

f (x, y1 ) < f (x, y2 ). (1.23)


1.5 Implicit Function Theorem 27

In particular, f (a, b) = 0 ⇒ f (a, b − m) < 0 < f (a, b + m). Since f : U → R is


continuously differentiable, therefore, f is continuous. Then ∃ a positive real quantity
r (< m) such that

f (x, b − m) < 0 < f (x, b + m), ∀ x ∈ I ≡ (a − r, a + r ).

Let x ∈ I . Since f (x, b − m) < 0 and f (x, b + m) > 0, by virtue of the Intermedi-
ate Value Theorem, ∃ y ∈ (b − m, b + m) at which f (x, y) = 0, and (1.23) implies
that ∃ only one such point, say g(x). This defines a function g : I → R having
properties (1.19) and (1.20).
Our claim: g : I → R is continuously differentiable and that the differentiation
formula (1.21) holds at the point a. Indeed, let a + h ∈ I . Then by definition, f (a +
h, g(a + h)) = 0 and f (a, g(a)) = 0. Hence, f (a + h, g(a + h)) − f (a, g(a)) =
0. Considering the Mean Value Theorem for scalar functions of two real variables,
∃ some points on the segment between the points (a, g(a)) and (a + h, g(a + h)),
which we label q(h), at which

∂f ∂f
f (a + h, g(a + h)) − f (a, g(a)) = (q(h))h + (q(h))[g(a + h) − g(a)].
∂x ∂y

This implies,
∂f ∂f
(q(h))h + (q(h))[g(a + h) − g(a)] = 0.
∂x ∂y

Thus
∂f
(q(h))
g(a + h) − g(a) = − ∂∂ xf h. (1.24)
∂y
(q(h))

∂f
Since the function : I → R is continuous and the closed square R is a sequentially
∂x
compact subset of the plane R2 , the Extreme Value Theorem guarantees the existence
of a positive number M, such that, for every (x, y) ∈ R

∂f
(x, y) ≥ M.
∂x
Combining the inequality (1.22) with the foregoing one, it follows from (1.24) that

M
|g(a + h) − g(a)| ≤ |h|, a + h ∈ I.
n

Hence, the function g : I → R is continuous at the point a. Since the point q(h) lies
on the segment between the points (a, g(a)) and (a + h, g(a + h)), we conclude
that
lim q(h) = (a, b).
h→0
28 1 Calculus on Rn

Fig. 1.5 Implicit Function Theorem

Dividing (1.24) by h and using the continuity of the first-order partial derivatives of
f : U → R at the point (a, b), it follows that
∂f
g(a + h) − g(a) (a, b)
lim = − ∂∂ xf ,
h→0 h ∂y
(a, b)

which means that g is differentiable at a and formula (1.21) holds at a. But any other
point x ∈ I satisfies the same assumptions as does the point a, and hence (1.21)
holds at all points in I (Fig. 1.5).

x2 y2
Example 1.7 Let us consider the equation + = 1, (x, y) ∈ R2 . The set of
16 25
solutions of the given equation consists of points in R2 lying on an ellipse with (0, 0)
as its centre. Let us begin with the solution (0, 5). Then for any r lying between 0
 4, define an interval I = (−r, r ) and define the function G : I → R by G(x) =
and
x2
5 1 − 2 , x ∈ I . Then there exists a neighbourhood of (0, 5) having the property
4
that the set of solutions of the given equation in this neighbourhood consists of points
of the form (x, G(x)), ∀ x ∈ I .
Next, let us consider the second component of (0, 5). Here, it is not possible to
find a neighbourhood J of 5, a function H : J → R, and a neighbourhood of (0, 5)
in which the set of solutions of the given equation consists of the points of the form
(H (y), y), ∀ y ∈ J . At every other vertices of the ellipse, it is possible to find a
neighbourhood of the vertex in which the set of solutions of the given equation has
1.5 Implicit Function Theorem 29

a similar description. On the other hand, at (a, b) that is not a vertex of the ellipse, is
a neighbourhood of (a, b) the set of solutions of the given equation determines both
x as a function y and vice-versa.

Example 1.8 Consider the equation

cos(x + y) + exp(y + x 2 ) + 3x + (x y)5 − 2 = 0, (x, y) ∈ R2 . (1.25)

Let us define

f (x, y) = cos(x + y) + exp(y + x 2 ) + 3x + (x y)5 − 2.

Then (x, y) is a solution of (1.25) if and only if f (x, y) = 0. Note that (0, 0) is a
solution of (1.25) and that

∂f ∂f
(0, 0) = 3, (0, 0) = 1.
∂x ∂y

Here f is a continuously differentiable function. So taking advantage of Dini’s


Theorem, we find a positive number r and a continuously differentiable function
g : I → R, where I is the open interval (−r, r ), such that
2
cos(x + g(x)) + expg(x)+x +3x + x 5 (g(x))5 − 2 = 0, ∀ x ∈ I.

Moreover, if (x, y) is a solution of (1.25) with |x| < r and |y| < r , then y = g(x).
Finally, g  (0) is determined by the formula

∂f ∂f
(0, 0) + (0, 0)g  (0) = 0 ⇒ g  (0) = −3.
∂x ∂y

Example 1.9 Let us consider the equation x 2 − y 2 = 0, (x, y) ∈ R2 . The set of


solutions of the given equation consists of points in R2 lie on the line x = y or
x = −y. At each solution (a, b) = (0, 0) of the given equation, there exists a neigh-
bourhood of (a, b) in which the set of solutions of the given equation determines
both x as a function of y and vice-versa. The origin (0, 0) is a solution of the given
equation, but  any neighbourhood of the origin in which the set of solutions coin-
cides with the graph of a function expressing one of the components of (x, y) as a
function of the other one.
30 1 Calculus on Rn

Problem 1.18 Show that the system of equations

3x1 + x2 − x3 − u 3 = 0
x1 − x2 + 2x3 + u = 0
2x1 + 2x2 − 3x3 + 2u = 0,

cannot be solved for x1 , x2 , x3 in terms of u.

Solution: Let

f (x1 , x2 , x3 , u) = 3x1 + x2 − x3 − u 3
g(x1 , x2 , x3 , u) = x1 − x2 + 2x3 + u
h(x1 , x2 , x3 , u) = 2x1 + 2x2 − 3x3 + 2u.

Then,  ∂f 
 ∂f ∂f   
 ∂ x1 ∂ x2 ∂u   3 1 −3u 2 
   
 ∂g ∂g ∂g   
 =  ∂x ∂ x2 ∂u  =  1 −1 1  = −12 − 12u 2 ,
 1   
 ∂h ∂h ∂h  2 2 2 
 ∂x ∂ x2 ∂u 
1

which can never be 0. If there were to exist a solution for x1 , x2 , x3 valid on some
interval (in which u varies), then the fact that (2x1 + 2x2 − 3x3 ) = (3x1 + x2 − x3 ) −
(x1 − x2 + 2x3 ) would imply that −2u = u 3 + u on that interval, which is impossi-
ble.

Implicit Function Theorem on Rn

Let m and n be positive integers. Suppose U be an open subset of Rm+n , and and that
the mapping F : U → Rm+n is continuously differentiable. Consider the equation

F(u) = 0, u ∈ U.

In the case where m = 1 and n = 1, we already considered this equation in Dini’s


Theorem. The object of this theorem is to state the General Implicit Function
Theorem, an extension of Dini’s Theorem to more general equations of the form
F(u) = 0, u ∈ U . In order to emphasize the analogy between the general case and
the case where m = 1 and n = 1, it is useful to introduce the following notation: For a
point u ∈ Rm+n , we separate the first m components of u from the last n components
and label them as follows:

u = (x, y) = (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ), (x, y) ∈ U.


1.5 Implicit Function Theorem 31

If the mapping F : U → Rn is written in terms of its component functions, F =


(F1 , F2 , . . . , Fn ), this equation, in turn, can be expressed as the following system of
n non-linear scalar equations in m + n scalar unknowns:

F1 (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ) = 0
..
.
Fi (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ) = 0
..
.
Fn (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ) = 0.

where (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ), (x, y) ∈ U .


Now we are in a position to state the General Implicit Function Theorem, a direct
generalization of Dini’s Theorem.
Statement 1.1 Implicit Function Theorem on Rn : Let m and n be positive inte-
gers. Suppose U be an open subset of Rn+m (≡ Rn × Rm ) and that the mapping
F : U → Rn is continuously differentiable. At the point (x0 , y0 ) ∈ U , suppose that
F(x0 , y0 ) = 0. Let T1 : Rn → Rn and T2 : Rm → Rn be two linear maps defined by
 
T1 (h 1 ) = F (x0 , y0 )(h 1 , 0), T2 (h 2 ) = F (x0 , y0 )(0, h 2 );

so that F (x0 , y0 )(h 1 , h 2 ) = T1 (h 1 ) + T2 (h 2 ) for all h 1 ∈ Rn , h 2 ∈ Rm . Suppose T1
is invertible. Then

(a) there exists open sets V1 ⊂ U, V2 ⊂ Rm with (x0 , y0 ) ∈ V1 , y0 ∈ V2 and a


unique map G : V2 → Rn such that

(G(y), y) ∈ V1 , F(G(y), y) = 0 ∀ y ∈ V2 ;

(b) for every (x, y) ∈ V1 such that F(x, y) = 0, we have y ∈ V2 and x = G(y);

(c) furthermore, G is continuously differentiable and G(y◦ ) = x◦ , G (x0 , y0 ) =
−1
−T1 T2 .

Proof (a) Let us define a map f : U → Rn+m by f (x, y) = (F(x, y), y). Then f
is differentiable on U and
 
f (x, y)(h 1 , h 2 ) = (F (x, y)(h 1 , h 2 ), ∀ (h 1 , h 2 ) ∈ Rn+m .

This implies f is continuously differentiable and it follows that



f (x0 , y0 )(h 1 , h 2 ) = (T1 (h 1 ) + T2 (h 2 ), h 2 ), (h 1 , h 2 ) ∈ Rn+m .
32 1 Calculus on Rn

Since T1 is invertible, therefore



f (x0 , y0 )(h 1 , h 2 ) = 0 ⇒ (T1 (h 1 ) + T2 (h 2 ), h 2 ) = 0 ⇒ T1 (h 1 ) + T2 (h 2 ) = 0, h 2 = 0
⇒ T1 (h 1 ) = 0, h 2 = 0
⇒ h 1 = 0, h 2 = 0.


This shows f (x0 , y0 ) is injective and so is also surjective, hence invertible. So
by the virtue of Inverse Function Theorem, there exists open sets V1 ⊂ U, V3 ⊂
Rn+m such that (x0 , y0 ) ∈ V1 , f : V1 → V3 is injective and the inverse map f −1 :
V3 → V1 is continuously differentiable. Moreover,

f (x0 , y0 ) = (F(x0 , y0 ), y◦ ) = (0, y◦ ).

So (0, y◦ ) ∈ V3 . Suppose V2 = {y ∈ Rm : (0, y) ∈ V3 }. Then y◦ ∈ V2 and V2 is


open. For any y ∈ V2 , we have (0, y) ∈ V3 and hence there exists (x, z) ∈ V1 such
that f (x, z) = (0, y). But f (x, z) = (F(x, z), z). Therefore, (F(x, z), z) =
(0, y), so that y = z and F(x, y) = 0. If F(x̃, y) = 0 with (x̃, y) ∈ V1 , then
(F(x̃, y), y) = (0, y), i.e.,

F(x̃, y) = (0, y) = f (x, y).

But f is injective on V1 . So x̃ = x. Hence, there exists a unique x for which


F(x, y) = 0 and (x, y) ∈ V1 . Taking x as G(y), (a) is established.
(b) Let (x, y) ∈ V1 and F(x, y) = 0. Then f (x, y) ∈ V3 . But f (x, y)=(F(x, y), y)
and F(x, y) = 0. This means (0, y) ∈ V3 , so that y ∈ V2 . By definition of G
above, G(y) is the unique η such that F(η, y) = 0 and (η, y) ∈ V1 . This implies
x = η = G(y)
(c) Since F(x0 , y0 ) = 0, (x0 , y0 ) ∈ V1 and y◦ ∈ V2 , therefore, G(y0 ) = x0 . Now, for
any y ∈ V2 , we have F(G(y), y) = 0 by (a), so that f (G(y), y) = (0, y), from
which it follows that (G(y), y) = f −1 (0, y). Thus, G is the composition of the
maps
y → (0, y), (x, y) → f −1 (x, y), (x, y) → x,

where the first and third are linear maps, while the second is continuously dif-
ferentiable. It follows that G is continuously differentiable. Since F(G(y), y) =
0 ∀ y ∈ V2 , the mapping y → F(G(y), y) must have derivative 0 everywhere.
On the other hand, we find that the derivative of the mapping y → F(G(y), y)
at y◦ maps h 2 ∈ Rm into
   
F (G(y◦ ), y◦ )(G (y◦ ), y◦ ) = F (x0 , y0 )(G (y◦ )h 2 , h 2 ), because G(y◦ = x◦ )

= T1 G (y◦ )h 2 + T2 h 2 , by hypothesis.
 
Since F (G(y◦ ), y◦ ) = 0, then G (y◦ )h 2 = −T1−1 T2 h 2 for all h 2 ∈ Rm . This com-
pletes the proof of (c).
1.5 Implicit Function Theorem 33

Remark 1.5 The implicit function theorem above provides a sufficient condition in
order that a continuous solution G of F(x, y) = 0 for x in terms of y satisfying the
requirement that G(y◦ ) = x◦ should exist and be unique. However, our theorem does
not explicitly mention the word solution.

Example 1.10 Consider the system of equations

ln(7 + x22 + x32 ) + x1 x3 + e x1 +x4 + 7 = 0,


x13 exp{cos(x22 + x42 )} + x1 + 2x4 + (x2 + x1 + x4 )4 = 0,

where (x1 , x2 , x3 , x4 ) ∈ R4 . Note that the point (0, 0, 0, 0) is a solution of this system
of equations. For a point (x1 , x2 , x3 , x4 ) ∈ R4 , let us define

F(x1 , x2 , x3 , x4 ) = ln(7 + x22 + x32 ) + x1 x3 + e x1 +x4 + 7, x13 exp{cos(x22 + x42 )}

+x1 + 2x4 + (x2 + x1 + x4 )4 .

Here F(x1 , x2 , x3 , x4 ) = (F1 (x1 , x2 , x3 , x4 ), F2 (x1 , x2 , x3 , x4 )), where

F1 (x1 , x2 , x3 , x4 ) ≡ ln(7 + x22 + x32 ) + x1 x3 + e x1 +x4 + 7 = 0,


F2 (x1 , x2 , x3 , x4 ) ≡ x13 exp{cos(x22 + x42 )} + x1 + 2x4 + (x2 + x1 + x4 )4 ) = 0.

Then the mapping F : R4 → R2 is continuously differentiable, as F1 and F2 is also


so, and that its derivative matrix at the point 0 = (0, 0, 0, 0) is
 
 1001
F (0) = .
1002

Thus, the 2 × 2 matrix


 ∂F ∂ F1   
1
∂ x1
(0, 0, 0, 0) ∂ x4
(0, 0, 0, 0) 11
∂ F2 ∂ F2 = ,
∂ x1
(0, 0, 0, 0) ∂ x4
(0, 0, 0, 0) 12

is invertible. We apply the Implicit Function Theorem to choose a positive number


r and continuously differentiable functions g : B → R and h : B → R, where B =
Br (0, 0), such that if x22 + x32 < r 2 , then (g(x2 , x3 ), x2 , x3 , h(x2 , x3 )) is a solution
of the given system of equations. Moreover, if the point (x1 , x2 , x3 , x4 ) ∈ R4 is a
solution of the given system of equations and if x12 + x42 < r 2 and x22 + x32 < r 2 ,
then x1 = g(x2 , x3 ) and x4 = h(x2 , x3 ).
34 1 Calculus on Rn

Exercises

Exercise 1.25 Consider the equation

exp{x − 2 + (y − 1)2 − 1} = 0.

Show that Dini’s Theorem applies at the solution (2, 1). Explicitly define the function
g : I → R that has the property that in a neighbourhood of the solution (2, 1), all
the solutions are of the form (x, g(x) for x ∈ I and check that formula (1.21) holds
for the derivative g  : I → R.

Exercise 1.26 Consider the given system of equations:

(x 2 + y 2 + z 2 )3 − x + z = 0, cos(x 2 + y 4 ) + e z − 2 = 0.

Use the Implicit Function Theorem to analyze the solutions of the given systems of
equations near the solution 0.
2
Exercise 1.27 For e x + y 2 + z − 4x y 3 − 1 = 0, use the Implicit Function Theo-
rem to analyze the solutions of the given systems of equations near the solution 0.
Chapter 2
Manifold Theory

2.1 Topological Manifold

Curves and surfaces are the fundamental concepts of studying geometry in a 3-


dimensional space. The quest for studying these two concepts in a space of higher
dimension yields the concept of manifold theory.
A locally Euclidean space of dimension n is a topological space such that every
point of this space has a neighbourhood homeomorphic to an open subset of Rn .
A topological manifold M of dimension n, denoted by M n , is a Hausdorff, second
countable, locally Euclidean space of dimension n (Fig. 2.1).
Thus, for each p ∈ M, there exists a neighbourhood U of M and a homeomor-
phism φ of U onto an open subset φ(U ) of Rn . The pair (U, φ) is called a chart.
Each such chart (U, φ) on M induces a set of n-real-valued functions on U defined
by
x i = u i ◦ φ, i = 1, 2, 3, . . . , n (2.1)

where u i ’s are defined by (1.1). The functions (x 1 , x 2 , . . . , x n ) are called the coor-
dinate functions or a coordinate system on U and U is called the domain of the
coordinate system. The chart (U, φ) is sometimes called an n-coordinate chart.
From (2.1), one obtains
x i ( p) = φi ( p), by (1.1).

Thus one can write  


φ( p) = x 1 ( p), x 2 ( p), . . . , x n ( p) . (2.2)

Let (V, ψ) be another chart of p of M such that p ∈ U ∩ V .


Let (y 1 , y 2 , . . . , y n ) be a local coordinate system on V such that

y i = u i ◦ ψ, i = 1, 2, 3, . . . , n (2.3)
 
ψ( p) = y 1 ( p), y 2 ( p), . . . , y n ( p) . (2.4)

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 35
M. Majumdar and A. Bhattacharyya, An Introduction to Smooth Manifolds,
https://doi.org/10.1007/978-981-99-0565-2_2
36 2 Manifold Theory

Fig. 2.1 Locally Euclidean


space of dimension n

Example 2.1 Rn is a topological manifold covered by a single chart (Rn , IRn ) where
I is the identity map.

Example 2.2 Every open subset U of Rn is a topological manifold with chart


(U, IU ).

Example 2.3 Every discrete topological space M is a 0-dimensional topological


manifold, the charts being given by the pair ({ p}, φ p ) where p → 0, p ∈ M.

Example 2.4 Let f : R3 → R3 be defined by f (x, y, z)=(x, x 2 + y 2 + z 2 − 1, z),


|J | = 2y. By virtue of Inverse Function Theorem, f is a local diffeomorphism at
p = (x, y, z) if and only if y = 0. Thus, the function f can serve as a local coordinate
system at any point p not on the x-axis and z-axis.

Remark 2.1 (i) A 1-dimensional manifold is locally homeomorphic to open inter-


val.
(ii) A 2-dimensional manifold is locally homeomorphic to open disc.

Remark 2.2 A topological manifold is


(i) locally connected
(ii) locally compact
(iii) normal and metrizable.
For proof, refer to any standard textbooks on general topology.

Problem 2.1 Does the map f : R → R defined by f (x) = x 2 form a chart?

Solution: 1 Note that f : R → R defined by f (x) = x 2 is not a homeomorphism


on R (refer to Exercise 1.5). Thus, the given map does not form a chart.

Problem 2.2 Consider the open subsets U and V of the unit circle S 1 of R2 given
by

U = {(cos α, sin α) : α ∈ (0, 2π)}, V = {(cos α, sin α) : α ∈ (−π, π)}

and the maps φ : U → R defined by φ(cos α, sin α) = α, α ∈ (0, 2π) and ψ : V →


R defined by ψ(cos α, sin α) = α, α ∈ (−π, π). Prove that (U, φ) and (V, ψ) are
charts on R2 .
2.1 Topological Manifold 37

Fig. 2.2 Overlap region

Solution: 2 Note that the two maps φ and ψ are homeomorphisms onto the open
subsets (0, 2π) and (−π, π) of R, respectively. Consequently, (U, φ) and (V, ψ) are
charts on R2 .
2
Problem 2.3 Prove that the graph y = x 3 in R2 is a topological manifold.

Solution: 3 The graph is a subspace of R2 and hence it is Hausdorff and second


2
countable. Also (x, x 3 ) → x, so it is homeomorphic to R. Consequently, it is locally
2
Euclidean. Hence, the graph y = x 3 in R2 is a topological manifold.

Problem 2.4 Find the functional relation between the two local coordinate systems
defined in the overlap region of a topological manifold.

Solution: 4 Let (U, φ) and (V, ψ) be two charts of a point p ∈ U ∩ V of a topo-


logical manifold M (Fig. 2.2).
Let φ( p) = q ∈ Rn , q ∈ φ(U ∩ V ). Suppose g : φ(U ∩ V ) → ψ(U ∩ V ) is
defined by
g(q) = (ψ ◦ φ−1 )(q). (2.5)

Then

g(q) = g(φ( p)) = (ψ ◦ φ−1 )(φ( p)), by (2.5)


i.e. g(q) = ψ( p)
 
or u i g(φ( p)) = u i (ψ( p))
or g i (φ( p)) = y i ( p) by (1.1), (2.3)
or g i (x 1 ( p), x 2 ( p), . . . , x n ( p)) = y i ( p)
i.e. y i = g i (x 1 , x 2 , . . . , x n ).

Problem 2.5 Give an example of a non-Hausdorff locally Euclidean space.

Solution: 5 Let A ⊂ R2 be such that



A=U {(0, 1)},
38 2 Manifold Theory

where U = {(x, 0)|x ∈ R}. Let U1 = U \ {(0, 0)} ∪ {(0, 1)}. We define φ : U → R
by φ(x, 0) = x and ψ : U1 → R by

0, x = 0
ψ(x, 1) =
x, x = 0.

Both φ and ψ are well-defined


 on U and U1 , respectively. Also φ and ψ are injective
maps in R and U U1 = A. So, (U, φ) and (U1 , ψ) are charts and hence it is a
locally Euclidean space.
Let V1 be an open neighbourhood
 of (0,
0) and V2 be an open neighbourhood of
(0, 1) in A. Then both φ(U V 1)  1 V2 ) are open subsets of R containing
and ψ(U  0.
So ∃ a = 0 such that a ∈ φ(U V1 ) ψ(U1 V2 ), which implies (a, 0) ∈ V1 V2 .
Hence the topology of A is non-Hausdorff. Thus A fails to form a topological man-
ifold.

Problem 2.6 Consider the cone C = {(x, y, z) ∈ R3 : x 2 + y 2 = z 2 } with the sub-


space topology as induced by the usual one of R3 . Prove that C is not a topological
manifold.

Solution: 6 Our claim is that the space C is not a locally Euclidean space. It suffices
to show that the point (0, 0, 0) ∈ C does not have a neighbourhood homeomorphic
to an open subset of R2 (Fig. 2.3).
Let U be an open neighbourhood of (0, 0, 0) in C. Let φ : U → V be a home-
omorphism between U and an open subset V of R2 . Then for some sufficiently
small r > 0, ∃ an open disc Br (φ(0, 0, 0)) with φ(0, 0, 0) as its centre such that
Br (φ(0, 0, 0)) ⊂ V . Now the punctured disc Br (φ(0, 0, 0)) \ {φ(0, 0, 0)} is con-
nected. But U \ {(0, 0, 0)} is not connected. In fact,

U \ {(0, 0, 0)} = U1 U2 ,

where
U1 = {(x, y, z) ∈ U : z > 0}, U2 = {(x, y, z) ∈ U : z < 0}.

So, U1 U2 = φ and U1 , U2 are open in C. Hence, U can be expressed as a disjoint
union of two non-empty open subsets of C. Thus C is not a locally Euclidean space.

Problem 2.7 Show that the cross in R2 with the subspace topology cannot be a
topological manifold.

Solution: 7 Our claim is to prove that the cross is not locally Euclidean at the
intersection q (Fig. 2.4).
If possible, let us assume that the cross is locally Euclidean of dimension n at
the point q. Then ∃ a homeomorphism φ : V → Br (0, 0, 0, . . . , 0), where V is an
2.1 Topological Manifold 39

Fig. 2.3 Double cone

Fig. 2.4 Cross(i)

open neighbourhood of q and Br (0, 0, 0, . . . , 0) ⊂ Rn is an open ball with centre


(0, 0, 0, . . . , 0) and radius r (sufficiently small enough). Here we assume φ(q) =
(0, 0, 0, 0, . . . , 0). The homeomorphism ψ : V \ {q} → Br (0, 0, 0, . . . , 0) \ (0, 0,
0, . . . , 0) acts as a restriction map to a homeomorphism φ. Now, if n ≥ 2 then

Br (0, 0, 0, . . . , 0) \ (0, 0, 0, . . . , 0)

is connected, and if n = 1 then it has two connected components (Fig. 2.5).


Since V \ {q} has four connected components, there  any homeomorphism from
V \ {q} to Br (0, 0, 0, . . . , 0) \ {(0, 0, 0, . . . , 0)}. This contradiction proves that the
cross is not locally Euclidean at q.
40 2 Manifold Theory

Fig. 2.5 Cross(ii)

Exercises
Exercise 2.1 Does the map f : R → R defined by f (x) = x 2k , k ∈ N, form a chart?
Exercise 2.2 Find all points in R2 , in a neighbourhood of which the function f :
R2 → R2 defined by f (x1 , x2 ) = (x12 + x22 − 1, x2 ) can serve as a local coordinate
system.
Exercise 2.3 Consider Exercise 1.12 of Sect. 1.1. Can f be taken as a local coor-
dinate map?
Exercise 2.4 As defined by (2.5), h : ψ(U ∩ V ) → φ(U ∩ V ) is defined by

h(r ) = (φ ◦ ψ −1 )(r ), r ∈ ψ(U ∩ V ). (2.6)

Show that the functional relation between the two local coordinate systems defined
in the overlap region of a topological manifold is given by

x i = h i (y 1 , y 2 , . . . , y n ).

Exercise 2.5 Let X = S 1 S 2 where

S 1 = {(x1 , x2 ) ∈ R2 : (x1 − 1)2 + x22 = 1}, S 2 = {(x1 , x2 ) ∈ R2 : (x1 + 1)2 + x22 = 1}.

Suppose X inherits the topology from R2 . Is X a topological manifold?

Answers
2.1. Yes. 2.2. Not on the y-axis. 2.3. Yes. 2.5. No.

2.2 Smooth Germs on a Topological Manifold

Let us begin with a definition:


Let M be an n-dimensional topological manifold. Let f : M → R be any func-
tion. Let p ∈ M. If, for every admissible coordinate chart (U, φ) of M satisfying
p ∈ U, ( f ◦ φ−1 ) : φ(U ) → R is C ∞ at the point φ( p) ∈ Rn , then we say that f is
C ∞ at p in M.
2.2 Smooth Germs on a Topological Manifold 41

Remark 2.3 Note that, if f is C ∞ at p in M, then f is continuous at p.

Theorem 2.1 Let M be an n-dimensional topological manifold. Let f : M → R be


any function with p ∈ M. If there exists an admissible coordinate chart (U, φ) of M
at p ∈ U such that ( f ◦ φ−1 ) : φ(U ) → R is C ∞ at φ( p), then f is C ∞ at p.

Proof Let (V, ψ) be any admissible coordinate chart of M with p ∈ V . Our claim
is that ( f ◦ ψ −1 ) : φ(V ) → R is C ∞ at ψ( p). Here

f ◦ ψ −1 = f ◦ (φ−1 ◦ φ) ◦ ψ −1 = ( f ◦ φ−1 ) ◦ (φ ◦ ψ −1 )

is C ∞ at ψ( p) as ( f ◦ φ−1 ) and (φ ◦ ψ −1 ) are C ∞ at φ( p) and ψ( p), respectively.


This completes the proof.

Let M be an n-dimensional topological manifold. Let f : M → R be any function.


By f is C ∞ at p in M (or f is smooth on M), we mean that f is C ∞ at every point
p ∈ M. The set of all smooth functions f : M → R on M is denoted by C ∞ (M).
Remark 2.4 Note that, if f is C ∞ at M, then f is continuous.
Let M and N be respectively n- and m-dimensional topological manifolds. Let f :
M → N be any continuous function. Let p ∈ M. If, for every admissible coordinate
chart (U, φ) of M satisfying p ∈ U and (V, ψ) of N with F( p) ∈ V , the mapping

ψ ◦ ( f ◦ φ−1 ) : φ(U ∩ f −1 (V )) → ψ(V )

is C ∞ at φ( p), then f is C ∞ at p.
Exercise
Exercise 2.6 Let M and N be respectively n- and m-dimensional topological man-
ifolds. Let f : M → N be any continuous function. Let p ∈ M. If there exists an
admissible coordinate chart (U, φ) of M at p ∈ U and (V, ψ) of N with f ( p) ∈ V ,
the mapping
ψ ◦ ( f ◦ φ−1 ) : φ(U ∩ f −1 (V )) → ψ(V )

is C ∞ at φ( p), then f is C ∞ at p.
Let M be an n-dimensional topological manifold, and N be an m-dimensional
topological manifold. If there exists a function f : M → R such that f is a diffeo-
morphism from M onto N , then we say that the manifolds M and N are isomorphic
(or diffeomorphic).
42 2 Manifold Theory

Let M be an n-dimensional topological manifold. Let p ∈ M. By C p∞ (M) (or


simply C p∞ ), we mean the collection of all real-valued functions f whose Dom f is
an open neighbourhood of p ∈ M, and for every admissible coordinate chart U, φ
of M satisfying p ∈ U ,

( f ◦ φ−1 ) : φ(Dom f ∩ U ) → R

is C ∞ at the point φ( p) ∈ Rn . Observe that if f ∈ C p∞ (M), then f is continuous on


some open neighbourhood of p.
Let M be an n-dimensional topological manifold. For every f, g ∈ C ∞ (M), we
define ( f + g) : M → R as follows: for every x ∈ M,

( f + g)(x) = f (x) + g(x).

For every f ∈ C ∞ (M), and for every real t, we define t f : M → R as follows: for
every x ∈ M,
(t f )(x) = t f (x).

Remark 2.5 It is clear that the set C ∞ (M), together with vector addition, and scalar
multiplication defined as above, constitutes a real linear space.

For every f, g ∈ C ∞ (M), we define ( f · g) : M → R as follows: for every x ∈


M,
( f · g)(x) = f (x)g(x).

Remark 2.6 It is easy to see that C ∞ (M) is an algebra.

Let M be an n-dimensional topological manifold. Let p ∈ M. Let γ1 and γ2 be,


in  p (M), set of all parametrized curves in M through p. Then the relation ‘≺’on
 p (M) is defined as follows: by γ1 ≺ γ2 , we mean that for any admissible coordinate
chart (U, φ) of M at p ∈ U ,
 
(φ ◦ γ1 ) (0) = (φ ◦ γ2 ) (0).

Proposition 2.1 Let M be an n-dimensional topological manifold. Let p ∈ M. Let


γ1 and γ2 be in  p (M). If there exists an admissible coordinate chart (U, φ) of M at
 
p ∈ U such that (φ ◦ γ1 ) (0) = (φ ◦ γ2 ) (0), then γ1 ≺ γ2 .

Proof Let us take any admissible coordinate chart (ψ, V ) of M satisfying p ∈ V .


 
Our claim is (ψ ◦ γ1 ) (0) = (ψ ◦ γ2 ) (0). Now
2.2 Smooth Germs on a Topological Manifold 43

  
(ψ ◦ γ1 ) (0) = (ψ ◦ φ−1 ) ◦ (φ ◦ γ1 ) (0)
    
= (ψ ◦ φ−1 ) (φ ◦ γ1 )(0) (φ ◦ γ1 ) (0)
    
= (ψ ◦ φ−1 ) (φ ◦ γ1 )(0) (φ ◦ γ2 ) (0)
    
= (ψ ◦ φ−1 ) (φ( p) (φ ◦ γ2 ) (0) , where γ1 (0) = p
    
= (ψ ◦ φ−1 ) (φ ◦ γ2 )(0) (φ ◦ γ2 ) (0)
 
= (ψ ◦ φ−1 ◦) ◦ (φ ◦ γ2 )
 
= ψ ◦ (φ−1 ◦ φ) ◦ γ2 (0)

= (ψ ◦ γ2 ) (0).

Problem 2.8 Let M be an n-dimensional topological manifold. Let p ∈ M. Then,


the relation ‘≺’ on  p (M) is an equivalence relation.

Solution: 8 • Reflexive: Since M be an n-dimensional topological manifold and


p ∈ M, there exists an admissible coordinate chart (U, φ) of M such that p ∈ U .
 
Since (φ ◦ γ) (0) = (φ ◦ γ) (0), then γ ≺ γ.
• Symmetric: Let γ1 ≺ γ2 hold. Let us take an admissible coordinate chart (U, φ)
 
of M such that p ∈ U . Since γ1 ≺ γ2 , then (φ ◦ γ1 ) (0) = (φ ◦ γ2 ) (0) implies
γ2 ≺ γ1 .
• Transitive: Let γ1 ≺ γ2 and γ2 ≺ γ3 holds. We are to prove γ1 ≺ γ3 . Let us take an
admissible coordinate chart (U, φ) of M such that p ∈ U . Since γ1 ≺ γ2 , therefore
   
(φ ◦ γ1 ) (0) = (φ ◦ γ2 ) (0). Also for γ2 ≺ γ3 , we have (φ ◦ γ2 ) (0) = (φ ◦ γ3 ) (0).
 
Combining the last two relations, we obtain (φ ◦ γ1 ) (0) = (φ ◦ γ3 ) (0) implies
γ1 ≺ γ3 .

Remark 2.7 Let M be an n-dimensional topological manifold. Let p ∈ M. By the


last proposition, the quotient set  p (M)/ ≺ is the collection of all equivalence classes
[γ], where γ ∈  p (M). Thus,

 p (M)/ ≺= {[γ] : γ ∈  p (M)}

where
[γ] = {γi ∈  p (M) : γ ≺ γi }.

Let M be an n-dimensional topological manifold. Let p ∈ M. Let us define a function


φ∗ :  p (M)/ ≺→ Rn by

φ∗ ([γ]) = (φ ◦ γ) (0).

Here φ∗ is well-defined.
Problem 2.9 Let M be an n-dimensional topological manifold. Let p ∈ M. Suppose
(U, φ) is any admissible coordinate chart of M such that p ∈ U . Then φ∗ as defined
above is bijective.
44 2 Manifold Theory

Solution: 9 Injectivity: Let [γ1 ] ∈  p (M)/ ≺ and [γ2 ] ∈  p (M)/ ≺, where γ1 , γ2 ∈


 p (M). Suppose φ∗ ([γ1 ]) = φ∗ ([γ2 ]). We are to prove [γ1 ] = [γ2 ], i.e. γ1 ≺ γ2 . Now
 
φ∗ ([γ1 ]) = φ∗ ([γ2 ]) ⇒ (φ ◦ γ1 ) (0) = (φ ◦ γ2 ) (0) ⇒ γ1 ≺ γ2 , by Proposition 2.1.

Surjectivity: Let q ∈ Rn . We are to find γ ∈  p (M) such that

 (φ ◦ γ)(t) − (φ ◦ γ)(0) φ(γ(t)) − φ(γ(0)) φ(γ(t)) − φ( p)


φ∗ ([γ]) = (φ ◦ γ) (0) = lim = lim = lim = q.
t→0 t t→0 t t→0 t

Let us define a function γ1 : (−1, 1) → Rn by γ1 (t) = tq + φ( p). Set

γ = φ−1 ◦ γ1 .

Then γ(0) = p holds. Here γ1 is continuous on (−1, 1). Since (U, φ) is a coordinate
chart of M, φ−1 is 1 − 1, onto and continuous at φ(U ), which is an open subset of
Rn . Since p ∈ U, φ( p) ∈ φ(U ). Also φ−1 is continuous at γ1 (0) = φ( p) ∈ φ(U ).
Moreover, φ(U ) forms an open neighbourhood of γ1 (0). Since γ1 is continuous,
φ(U ) being an open neighbourhood of γ1 (0) ∃ δ > 0 with δ < 1 and for every
t ∈ (−δ, δ), we have γ1 (t) ∈ φ(U ). Hence, γ(t) = (φ−1 ◦ γ1 )(t) = φ−1 (γ1 (t)) ∈ U .
Since γ(t) ∈ U for every t ∈ (−δ, δ), it implies that γ is defined on (−δ, δ) for some
δ > 0. Now for every t ∈ (−δ, δ),

γ(t) = (φ−1 ◦ γ1 )(t) = φ−1 (tq + φ( p)) ∈ U,

this shows that γ maps from (−δ, δ) to U . Furthermore, γ1 is C ∞ at every t ∈


(−δ, δ), φ ◦ γ is C ∞ at every t and (U, φ) being an admissible coordinate chart of
M such that γ(t) ∈ U , by virtue of Exercise 2.6, γ is C ∞ at every t ∈ (−δ, δ). Thus
γ ∈  p (M).
Finally,

φ(γ(t)) − φ( p) φ(φ−1 ◦ γ1 (t)) − φ( p) φ(γ1 (t)) − φ( p) φ(tq + φ( p)) − φ( p)


lim = lim = lim = lim = q.
t→0 t t→0 t t→0 t t→0 t

This completes the proof.

Since φ∗ is 1 − 1 and onto, φ−1 ∗


exists and is also 1 − 1 and onto. Let us define a
binary composition ⊕ and external composition  on  p (M)/ ≺ as follows.
For every [γ1 ], [γ2 ] ∈  p (M)/ ≺, where γ1 , γ2 ∈  p (M)/ ≺,

[γ1 ] ⊕ [γ2 ] = φ−1



(φ∗ ([γ1 ]) + φ∗ ([γ2 ])),

and
t  [γ1 ] = φ−1

(t (φ∗ (γ1 ))), ∀ t ∈ R.
2.2 Smooth Germs on a Topological Manifold 45

Fig. 2.6 Germ

Remark 2.8 The quotient set ( p (M)/ ≺, ⊕, ) forms a real linear space.

Remark 2.9 Since  p (M)/ ≺ is a real linear space, φ∗ is 1 − 1 and onto, φ∗ is linear,
i.e.
φ∗ ([γ1 ] ⊕ (t  [γ2 ])) = φ∗ ([γ1 ]) + t φ∗ ([γ2 ]),

φ∗ is an isomorphism between  p (M)/ ≺ and Rn . Hence the dimension of ( p (M)/ ≺


is n. The set  p (M)/ ≺ is denoted by Tp (M) (Fig. 2.6).

Problem 2.10 Let M be an n-dimensional topological manifold and p ∈ M. Let 


be a relation on C ∞ ∞
p (M) defined as follows: for every f, g ∈ C p (M), by f  g we
mean that there exists an open neighbourhood V of p such that f (x) = g(x) for
every x ∈ V . Then  is an equivalence relation over C ∞
p (M).

Solution: 10 Reflexivity: Left to the reader.


Surjectivity: Left to the reader.
Transitivity: Let f  g and g  h hold where f, g, h ∈ C ∞
p (M). Since f  g,
from the definition of , there exists an open neighbourhood V of p such that
f (x) = g(x) for every x ∈ V . By similar reason, for g  h, there exists an open
neighbourhood W of p such that g(x) = h(x) for every x ∈ W . Since V, W are open
neighbourhoods of p ∈ M, therefore V ∩ W is an open neighbourhood of p ∈ M.
Also f (x) = h(x) for every x ∈ V ∩ W . Hence f  h holds.

Remark 2.10 The quotient set C ∞


p (M)/ , of all equivalence classes, is denoted
by F p (M). Thus
F p (M) = {[ f ] : f ∈ C ∞
p (M)},

where
[ f ] = {g : g ∈ C ∞
p (M), g  f }

is a C ∞ -germ at p on M. So, the members of F p (M) are called C ∞ -germs at p on


M.
46 2 Manifold Theory

Let M be an n-dimensional topological manifold and p ∈ M. For every f, g ∈


C∞
p (M), we define ( f + g) : Dom f ∩ Dom g → R as follows: for every x ∈
Dom f ∩ Dom g,
( f + g)(x) = f (x) + g(x).

For every f, g ∈ C ∞
p (M), we define f · g : Dom f ∩ Dom g → R as follows: for
every x ∈ Dom f ∩ Dom g,

( f · g)(x) = f (x)g(x).

Remark 2.11 Let M be an n-dimensional topological manifold and p ∈ M. Let


f, g ∈ C ∞ ∞
p (M). Then f + g, f · g : Dom f ∩ Dom g → R, is in C p (M). Also for
any t ∈ R,
t f : Dom f → R is in C ∞p (M).

Let M be an n-dimensional topological manifold and p ∈ M. For every f, g ∈


C∞
p (M) and t ∈ R, we define

[ f ] + [g] = [ f + g], t[ f ] = [t f ], [ f ][g] = [ f · g].

Remark 2.12 F p (M) forms a real linear space. Also, F p (M) forms an algebra.

Remark 2.13 Let M be an n-dimensional topological manifold and p ∈ M. If f, g ∈


C∞ ∞
p (M), then ( f + g) ∈ C p (M).

Remark 2.14 Let M be an n-dimensional topological manifold and p ∈ M. If f, g ∈


C∞ ∞
p (M), then ( f · g) ∈ C p (M).

Remark 2.15 Let M be an n-dimensional topological manifold and p ∈ M. If f ∈


C∞ ∞
p (M) and t ∈ R, then t f ∈ C p (M).

Problem 2.11 Let M be an n-dimensional topological manifold and p ∈ M. Let


γ ∈  p (M). Let [ f ] ∈ F p (M), where f ∈ C ∞
p (M). Then

( f ◦ γ)(0 + t) − ( f ◦ γ)(0)
lim
t→0 t

exists.

Solution: 11 Since γ ∈  p (M), it follows from the definition of  p (M) that ∃ δ >
0 such that γ : (−δ, δ) → M, γ(0) = p and γ is a smooth map on (−δ, δ), and
so is continuous on (−δ, δ). Here, the function f : Dom f → R is in ∈ C ∞ p (M),
where Dom f is an open neighbourhood of p ∈ M, so f is continuous on some
open neighbourhood U (⊆ Dom f ) of p. Since γ(0) = p and p ∈ Dom f , therefore
0 ∈ Dom ( f ◦ γ). Moreover, as γ is continuous with γ(0) = p and U being an open
neighbourhood of p, ∃  > 0 such that  < δ and for every t ∈ (−, ), we have γ(t) ∈
U (⊆ Dom f ), and hence ( f ◦ γ)(t) = f (γ(t)) ∈ R. This implies that (−, ) ⊆
2.2 Smooth Germs on a Topological Manifold 47

Dom ( f ◦ γ). Hence 0 is an interior point of f ◦ γ. Here, φ ◦ γ : (−δ, δ) → Rn is


C ∞ at the point 0 in R and ( f ◦ φ−1 ) : φ(Dom f ∩ U ) → R is C ∞ at the point
(φ ◦ γ)(0) in Rn . So the composite function f ◦ γ is C ∞ at the point 0 in R. Hence,
d( f ◦ γ)(t)  ( f ◦ γ)(0 + t) − ( f ◦ γ)(0)
, i.e. lim exists.
dt t=0 t→0 t
So the following definition is well defined:
Let M be an n-dimensional topological manifold and p ∈ M. Let γ ∈  p (M).
Let [ f ] ∈ F p (M), where f ∈ C ∞
p (M). Then

( f ◦ γ)(0 + t) − ( f ◦ γ)(0)
lim
t→0 t

exists, and is denoted by < γ, [ f ] >. Thus

( f ◦ γ)(0 + t) − ( f ◦ γ)(t)
< γ, [ f ] >≡ lim ,
t→0 t

i.e.
d( f ◦ γ)(t) 
< γ, [ f ] >≡ .
dt t=0

Theorem 2.2 Let M be an n-dimensional topological manifold and p ∈ M. Let


γ ∈  p (M). Suppose [ f ], [g] ∈ F p (M), where f, g ∈ C ∞
p (M). Then

1. < γ, [ f ] + [g] >=< γ, [ f ] > + < γ, [g] >


2. < γ, c[ f ] >= c < γ, [ f ] >, where c ∈ R.
In other words, <, > is linear in the second variable.
Proof 1. Here,

(( f + g) ◦ γ)(0 + t) − (( f + g) ◦ γ)(0)
< γ, [ f ] + [g] > = <γ, [ f + g] >= lim
t→0 t
( f + g)(γ)(t) − ( f + g)(γ)(0)
= lim
t→0 t
{( f (γ)(t)) − ( f (γ)(0))} + {(g(γ)(t)) − (g(γ)(0))}
= lim
t→0 t
{( f (γ)(t)) − ( f (γ)(0))} {(g(γ)(t)) − (g(γ)(0))}
= lim +
t→0 t t
{( f (γ)(0 + t)) − ( f (γ)(0))} {(g(γ)(0 + t)) − (g(γ)(0))}
= lim +
t→0 t t
= <γ, [ f ] > + < γ, [g] > .

2. Left to the reader.


48 2 Manifold Theory

Theorem 2.3 Let M be an n-dimensional topological manifold and p ∈ M. Let γ ∈


 p (M). Suppose [ f ], [g] ∈ F p (M), where f, g ∈ C ∞
p (M). Further, suppose [γ] ∈
Tp (M), where γ ∈  p (M). Then
1. [γ]([ f ] + [g]) = [γ]([ f ]) + [γ]([g]).
2. [γ]([t f ]) = t ([γ]([ f ]), where t ∈ R.
3. [γ]([ f ][g]) = ([γ][ f ])(g( p)) + f ( p)([γ][g]).

Proof 1. Left to the reader.


2. Left to the reader.
3. Here
d(( f · g) ◦ γ)(t)  d(( f · g)(γ(t))) 
[γ]([ f ][g]) = [γ]([ f · g]) = =
dt t=0 dt t=0

d( f (γ(t)) · g(γ(t))) 
=
dt t=0

d(( f ◦ γ)(t) · (g ◦ γ)(t)) 


=
dt t=0

d( f ◦ γ)(t)  d(g ◦ γ)(t) 


= (g ◦ γ)(0) + ( f ◦ γ)(0)
dt t=0 dt t=0

= ([γ][ f ])g(γ(0)) + f (γ(0))([γ][g])


= ([γ][ f ])(g( p)) + f ( p)([γ][g]).

2.3 Smooth Manifold

We are now going to introduce a differentiable structure on a topological manifold.


For this, let us at first introduce compatible charts or C ∞ -related charts.
Two charts (U, φ) and (V, ψ) on a topological manifold M are said to be C ∞ -
compatible or C ∞ -related if


⎪ either U ∩ V = φ, or;

U ∩ V = φ, and the
(2.7)

⎪ transition maps φ ◦ ψ −1 : ψ(U ∩ V ) → φ(U ∩ V ), and

ψ ◦ φ−1 : φ(U ∩ V ) → ψ(U ∩ V ) are of class C ∞ .

In short, we say compatible. These two maps are called the transition functions
between the charts. If U ∩ V = φ, then the two charts are obviously C ∞ -compatible.

Problem 2.12 Prove that the compatibility of charts is not an equivalence relation.

Solution: 12 Let (U, φ), (V, ψ) and (W, ϕ) be three charts on a topological mani-
fold.
2.3 Smooth Manifold 49

(i) Here, (U, φ) is C ∞ -compatible to itself as φ ◦ φ−1 = I is C ∞ . Hence compat-


ibility of charts is reflexive.
(ii) Let us assume that the chart (U, φ) is C ∞ -compatible to (V, ψ). Then by (2.7),
we have 
φ ◦ ψ −1 : ψ(U ∩ V ) → φ(U ∩ V ) and
(∗)
ψ ◦ φ−1 : φ(U ∩ V ) → ψ(U ∩ V )

are of class C ∞ , where p ∈ U ∩ V . To prove (V, ψ) is C ∞ -compatible to


(U, φ), we have to show

ψ ◦ φ−1 : φ(U ∩ V ) → ψ(U ∩ V ) and
φ ◦ ψ −1 : ψ(U ∩ V ) → φ(U ∩ V )

are of class C ∞ , which follows from (∗). Consequently, the compatibility of


charts is symmetric.
(iii) Let us assume that the chart (U, φ) is C ∞ -compatible to (V, ψ) and that the
chart (V, ψ) is C ∞ -compatible to (W, ϕ). To prove the compatibility of charts
is transitive, we need to show (U, φ) is C ∞ -compatible to (W, ϕ).
Since (V, ψ) is C ∞ -compatible to (W, ϕ), therefore

ψ ◦ ϕ−1 : ϕ(V ∩ W ) → φ(V ∩ W ) and
(∗∗)
ϕ ◦ ψ −1 : ψ(V ∩ W ) → ϕ(V ∩ W )

are of class C ∞ .
Note that

φ ◦ ϕ−1 = (φ ◦ ψ −1 ) ◦ (ψ ◦ ϕ−1 ) and ϕ ◦ φ−1 = (ϕ ◦ ψ −1 ) ◦ (ψ ◦ φ−1 ).

Now for any p ∈ U ∩ V ∩ W , we have φ ◦ ϕ−1 and ϕ ◦ φ−1 are C ∞ on ϕ(U ∩


V ∩ W ) and φ(U ∩ V ∩ W ), respectively. But, in particular, if we consider
any p ∈ (U ∩ W ) \ (U ∩ V ∩ W ) then φ ◦ ϕ−1 and ϕ ◦ φ−1 fails to be C ∞ on
ϕ(U ∩ W ) and φ(U ∩ W ), respectively. Hence compatibility of charts is not
transitive.
This proves C ∞ -compatibility of charts is not an equivalence relation (Fig. 2.7).

An atlas on a topological manifold M is a collection of pairwise C ∞ -compatible


charts {(Uα , φα ) : α ∈ I } such that
⎧ 
⎨ (i) Uα = M
α∈I

(ii) φα ◦ φ−1
β or φβ ◦ φ−1
α is C

on φβ (Uα ∩ Uβ ) or on φα (Uα ∩ Uβ ).
(2.8)
A differential structure on M is an atlas A which is maximal, i.e. if (U, φ)
is another chart such that φ ◦ φ−1
α and φα ◦ φ
−1
respectively on φα (U ∩ Uα ) and

φ(U ∩ Uα ) are of class C , then (U, φ) ∈ A.
50 2 Manifold Theory

Fig. 2.7 ϕ ◦ φ−1 is C ∞ on φ(U ∩ V ∩ W )

A smooth manifold or a differentiable manifold of class C ∞ of dimension n is


a pair (M, A) where M is a topological manifold of dimension n and A is an atlas.
Proposition 2.2 Let A be an atlas on a smooth manifold M. Then, there is a unique
differential structure  on M such that A ⊂ .
Proof Let  be the set of all charts on M which are C ∞ -related to every chart of A.
Then clearly, A ⊂ .
If  is an atlas, by definition, it is maximal and has a differential structure. We
are now going to prove the uniqueness of such .
If possible, let  be another differential structure on M with the desired property.
Then A ⊂  ⊂ , by definition of . But by the maximality of  ,  = .
Now, let (U, φ) and (V, ψ) be any two charts of . We wish to show that φ ◦ ψ −1 :
ψ(U ∩ V ) → φ(U ∩ V ) is of class C ∞ . For this, let x ∈ ψ(U ∩ V ) and (Ui , φi ) ∈ 
be such that ψ −1 (x) ∈ Ui . Now

φ ◦ φi−1 : φi (U ∩ Ui ) → Rn and
φi ◦ ψ −1 : ψ(Ui ∩ V ) → Rn

are of class C ∞ .
Let W be an open neighbourhood of x such that W ⊂ φ(U ∩ V ∩ Ui ). Now on W ,
φ ◦ ψ −1 = φ ◦ φi−1 ◦ φi ◦ ψ −1 is of class C ∞ , as a composition of two C ∞ -functions
is also so. This completes the proof.
Remark 2.16 Now, it must be clear that to introduce differential structure, one needs
to find a chart or a coordinate map. We proceed as follows.
Let f i : i = 1, 2, 3, . . . , n be n-real-valued C ∞ functions defined on M. Let the
set { f i } be non-vanishing Jacobian at p ∈ M. Then by Inverse Function  Theo-
rem, there exists a neighbourhood V of p and a neighbourhood U of f i ( p) : i =
2.3 Smooth Manifold 51

1, 2, 3, . . . , n such that f = ( f 1 , f 2 , . . . , f n ) mapping V into U in 1 − 1 manner
has an C ∞ -inverse. Let it be denoted by φ−1 . Then (V, φ) is the desired chart.
Consider f : R2 → R2 defined by f (x, y) = (x 2 + 3y 2 , x y). Thus

f 1 (x, y) = x 2 + 3y 2 , f 2 (x, y) = x y.

Therefore
 
 2x 6y  √

|J | =   = 2x 2 − 6y 2 = 0, if and only if x ± 3y = 0.
y x 

Consequently, for all other points of R2 , we can find a neighbourhood V of p = (x, y)


and the neighbourhood U of ( f 1 , f 2 ) which is mapped by φ−1 to V . Thus (V, φ) is
a chart.

Example 2.5 Rn is a smooth manifold with respect to the atlas {(U, φ)} where
U = Rn and φ is the identity map.

Example 2.6 Any open subset W of a smooth manifold M n is a smooth manifold  


of the same dimension. For, if {(Uα , φα )} is an atlas of M, then Uα ∩ W, φα Uα ∩W
is an atlas for W where 
φα Uα ∩W : Uα ∩ W → Rn

is of class C ∞ .

Remark 2.17 Any chart is said to be compatible with an atlas A, if it is compatible


with all the charts of A.

Problem 2.13 Let A be an atlas on a topological manifold. If two charts (V, ψ) and
(W, θ) are both compatible with A, then they are compatible with each other.

Solution: 13 Let p ∈ V ∩ W . By (2.7), let A = {(Uα , φα ) : α ∈ ∧} be the atlas.


Since (V, ψ) and (W, θ) are compatible with A, p ∈ Uα for some α, i.e. p ∈ V ∩
W ∩ Uα . Now
θ ◦ ψ −1 = (θ ◦ φ−1
α
) ◦ (φα ◦ ψ −1 ).

Again φα ◦ ψ −1 : ψ(V ∩ W ∩ Uα ) → φα (V ∩ W ∩ Uα ) is C ∞ , as (V, ψ) is compat-


ible with A. Finally,

φ−1
α
◦ (φα ◦ ψ −1 ) : ψ(V ∩ W ∩ Uα ) → V ∩ W ∩ Uα is C ∞ i.e.
(θ ◦ φ−1
α
) ◦ (φα ◦ ψ −1 ) : ψ(V ∩ W ∩ Uα ) → θ(V ∩ W ∩ Uα ) is C ∞

on ψ(V ∩ W ∩ Uα ) and hence on ψ( p). Consequently, θ ◦ ψ −1 is C ∞ . Similarly,


it can be shown that ψ ◦ θ−1 is C ∞ on θ(V ∩ W ∩ Uα ) and hence on θ( p). Thus,
(V, ψ) and (W, θ) are both compatible with each other.
52 2 Manifold Theory

Problem 2.14 Prove that A = {(U, φ), (V, ψ)}, where U, V, φ, ψ are defined in
Problem 2.2 is an atlas on S 1 .
Solution: 14 In Problem 2.2, it has been proved that (U, φ) and (V, ψ) are charts
on R2 , where
(i) S 1 = U ∪ V
(ii) Now

φ−1 : φ(U ∩ V ) → U ∩ V and ψ ◦ φ−1 : φ(U ∩ V ) → ψ(U ∩ V )

such that

α, if α ∈ (0, π)
(ψ ◦ φ−1 )(α) = ψ(cos α, sin α) =
α − 2π, if α ∈ (π, 2π)

and hence C ∞ . Also

φ ◦ ψ −1 : ψ(U ∩ V ) → φ(U ∩ V )

is such that φ(U ∩ V )(α) = φ(cos α, sin α) = α, α ∈ (0, π), and hence C ∞ .
Thus A = {(U, φ), (V, ψ)} is an atlas on R2 .
Problem 2.15 Prove that S 1 is a 1-dimensional manifold.

Solution: 15 Let S 1 = {(x, y) : (x, y) ∈ R2 , x 2 + y 2 = 1} be a unit circle in R2 .
We give S 1 , the topology of a subspace of R2 . Let

U1 = {(x, y) ∈ S 1 : y > 0}, U2 = {(x, y) ∈ S 1 : y < 0}


U3 = {(x, y) ∈ S 1 : x > 0}, U4 = {(x, y) ∈ S 1 : x < 0}.

Then each Ui is an open subset of S 1 and S 1 = Ui i = 1, 2, 3, 4. We define

φ1 : U1 → R be such that φ1 (x, y) = x


φ2 : U2 → R be such that φ2 (x, y) = x
φ3 : U3 → R be such that φ3 (x, y) = y
φ4 : U1 → R be such that φ4 (x, y) = y.

Then each φi is a homeomorphism on R and hence each (Ui , φi ) is a chart of S 1 . Now


U1 ∩ U2 = , so (Ui , φi ), i = 1, 2 are C ∞ -related. Further, U1 ∩ U3 = . Let p ∈
U1 ∩ U3 . Then (φ1 ◦ φ−13
)(y) = x, (φ3 ◦ φ−1
1
)(x) = y are of class C ∞ . Proceeding
in this manner, it can be shown that A = {(Ui , φi ) : i = 1, 2, 3, 4} is an atlas of S 1
and hence (S 1 , A) is a 1-dimensional manifold.
Problem 2.16 Prove that the topological space M(m × n, R) of all m × n order
matrices with real entries form a smooth manifold of dimension mn.
2.3 Smooth Manifold 53

Solution: 16 Let us define the map φ : M → Rmn by

φ(A) = (a11 , a12 , . . . , a1n ; . . . , . . . ; am1 , am2 , . . . , amn ),

where A = (ai j ), i = 1, 2, . . . , m; j = 1, 2, . . . , n. Here φ is one-to-one and onto.


Moreover with the topology induced by Rmn on M, φ is a homeomorphism. So
(M, φ) forms a chart on M(m × n, R), whose domain is whole of M. Let us denote
A = {(M, φ)} to be the collection of charts of M(m × n, R). It only remains to show
that all pairs of members in A are C ∞ -compatible. For this, let (M, φ) ∈ A and
(M, ψ) ∈ A. The transition functions

φ ◦ ψ −1 : Rmn → Rmn and ψ ◦ φ−1 : Rmn → Rmn

are both identity maps and hence C ∞ -compatible. So A forms a C ∞ -atlas of M(m ×
n, R). Thus, M(m × n, R) form a smooth manifold of dimension mn.
In particular, if m = n then M(n × n, R) forms a smooth manifold of dimen-
sion n 2 .

Problem 2.17 Prove that the topological space G L(n, R) forms a smooth manifold
of dimension n 2 .

Solution: 17 Here G L(n, R) = {A ∈ M(n × n, R) : |A| = 0}. It is clear that the


determinant map

D : M(n × n, R) → R, i.e. (ai j ) → |(ai j )|,

where i, j = 1, 2, 3, . . . , n is continuous (also smooth). Hence, the inverse image of


the open subset R \ {0} of R is open in M(n × n, R), i.e. D−1 (R \ {0})(= G L(n, R))
is open in M(n × n, R). And M(n × n, R) being n 2 -dimensional smooth manifold
(refer to Problem 2.16), we can say that G L(n, R) is also n 2 -dimensional smooth
manifold (refer to Example 2.6).

Problem 2.18 Give an example of a non-Hausdorff space having differentiable


structures.

Solution: 18 Let A ⊂ R2 be such that



A=U {(0, 1)},

where U = {(x, 0)|x ∈ R}. Let U1 = {(x, 1)|x ∈ R}. We define φ : U → R by


φ(x, 0) = x and ψ : U1 → R by

0, x = 0
ψ(x, 1) =
x, x = 0.
54 2 Manifold Theory

Fig. 2.8 Not atlas

Both φ and ψ are well-defined


 on U and U1 , respectively. Also φ and ψ are home-
omorphisms and U1 U2 = A. The transitions functions φ ◦ ψ −1 and ψ ◦ φ−1 are
the identity functions on R \ {0}. So {(U, φ), (U1 , ψ)} forms a C ∞ atlas on A.
Let V1 be an open neighbourhood
 of (0,
0) and V2 be an open neighbourhood of
(0, 1) in A. Then both φ(U V 1) and 1 V2 ) are open subsets of R containing
ψ(U  0.
So ∃ a = 0 such that a ∈ φ(U V1 ) ψ(U1 V2 ), which implies (a, 0) ∈ V1 V2 .
Hence the topology of A is non-Hausdorff.
Problem 2.19 Consider S = {(x, 0) ∈ R2 : x ∈ (−1, 1)} ∪ {(x, x) ∈ R2 : x ∈ (0, 1)}
where U = {(x, 0) : x ∈ (−1, 1)}, φ : U → R is such that φ(x, 0) = x and V =
{(x, 0) : x ∈ (−1, 0]} ∪ {(x, x) : x ∈ (0, 1)}, ψ : V → R be such that ψ(x, 0) =
x, ψ(x, x) = x. Is A = {(U, φ), (V, ψ)} an atlas on S?

Solution: 19 Endow S = U ∪ V with the subspace topology inherited from R2 . Let


S1 = {(x, 0) : x ∈ (−1, 0]} and S2 = {(x, x) : x ∈ (0, 1)}. Then V = S1 ∪ S2 . Here
φ and ψ are both homeomorphisms respectively on U and V . Hence (U, φ) and
(V, ψ) form the charts on S. Now the transition functions φ ◦ ψ −1 : ψ(U ∩ V ) →
φ(U ∩ V ) and ψ ◦ φ−1 : φ(U ∩ V ) → ψ(U ∩ V ) are the identity map on φ(U ∩ V )
and ψ(U ∩ V ) respectively, where φ(U ∩ V ) = (−1, 0] = ψ(U ∩ V ), which are not
open in R. Thus, A does not form an atlas on S (Fig. 2.8).
Problem 2.20 Let U = {(x, 0) : x ∈ R} and V = {(x, 0) : x < 0} ∪ {(x, 1) : x >
0}. Suppose the maps φ : U → R and ψ : V → R are defined respectively by
φ(x, 0) = x and ψ(x, 0) = x, ψ(x, 1) = x. Prove that A = {(U, φ), (V, ψ)} an atlas
on S = U ∪ V .
2.3 Smooth Manifold 55

Fig. 2.9 Atlas

Solution: 20 Endow S = U ∪ V with the subspace topology inherited from R2 .


Let S1 = {(x, 0) : x ∈ (−∞, 0)} and S2 = {(x, x) : x ∈ (0, 1)}. Then V = S1 ∪ S2 .
Here φ and ψ are both homeomorphisms respectively on U and V . Hence (U, φ) and
(V, ψ) form the charts on S. Now the transition functions φ ◦ ψ −1 : ψ(U ∩ V ) →
φ(U ∩ V ) and ψ ◦ φ−1 : φ(U ∩ V ) → ψ(U ∩ V ) are the identity map on φ(U ∩ V )
and ψ(U ∩ V ) respectively, where φ(U ∩ V ) = (−∞, 0) = ψ(U ∩ V ), which are
open in R. Also the map φ ◦ ψ −1 and ψ ◦ φ−1 are C ∞ . Thus, A does form an atlas
on S (Fig. 2.9).

Exercises

Exercise 2.7 Give an example of a topological manifold which does not admit dif-
ferential structure.

Exercise 2.8 Let (M, A) be a smooth manifold. Prove that there exists a chart (U, φ)
of A such that φ( p) = 0, p ∈ U .

Remark 2.18 The condition of second countability in the definition of smooth man-
ifold implies paracompactness which further implies metric structure in the manifold.
Since the present book is considering only the different aspects of smooth manifold,
the condition is redundant here.

From now onwards, unless otherwise stated, a manifold will mean a smooth
manifold.
56 2 Manifold Theory

2.4 Stereographic Projection

The stereographic projection is a particular mapping that projects a sphere onto a


plane. The projection is defined on the entire sphere, except at the projection point.
To begin with, let us consider the sphere S 1 = {(x, y) : (x, y) ∈ R2 , x 2 + y 2 =
1}. The
 stereographic projection, on the sphere  S 1 , from the North Pole N =
(0, 1) respectively, South Pole S = (0, −1) onto the line y = 0 (i.e. x-axis) is the
map which assigns any point p ∈ S 1 \ {N  } respectively, p ∈ S \ {S} to the point
1

where the straight line through p and N respectively, through S intersects the line
y = 0.
 The inverse of the stereographic
 projection is the map from the x-axis to S 1 \ {N }
respectively, S \ {S} assigning the point q in the
1
 line y = 0 to the point where the
straight line through N respectively, through S and q intersect S 1 .
Similarly, the stereographic projection, on the sphere S 2 = {(x, y, z) : (x, y, z) ∈
R , x 2 + y 2 + z 2 = 1}, from the North Pole N = (0, 0, 1) respectively, South Pole
3

S = (0, 0, −1) onto the plane z = 0 (i.e. x y-plane) is the map which assigns any
point p ∈ S 2 \ {N }(respectively, p ∈ S 2 \ {S}) to the point where the straight line
through p and N respectively, through S intersects the plane z = 0.
The inverse of the stereographic projection is the map from the x y-plane to S 2 \
{N }(respectively, S 2 \ {S})assigning the point q in the plane z = 0 to the point where
the straight line through N respectively, through S and q intersects S 2 .
On generalization, we can define the stereographic projection and its inverse for the

n+1
n-dimensional sphere S n = {(x1 , x2 , . . . , xn+1 ) : (x1 , x2 , . . . , xn+1 ) ∈ Rn+1 , (xi )2
i=1
= 1} as follows.
The stereographic projection, on the sphere S n , from the North Pole N =
(0, 0, . . . , 0, 1) respectively, South Pole S = (0, 0, . . . , 0, −1) onto the plane
x n+1 = 0 is the map which assigns any point p ∈ S n − {N } respectively, p ∈
S n − {S} to the point where the straight line through p and N (respectively, S)
intersects the plane x n+1 = 0.
The inverse of the stereographic projection is the map from the plane x n+1 = 0 to
S n \ {N } (respectively, S n \ {S}) assigning the point q in the plane x n+1 = 0 to the
point where the straight line through N (respectively, through S) and q intersect S n .
Problem 2.21 Using stereographic projection with x-axis as the image line, show
that S 1 is a smooth manifold (Fig. 2.10).

Solution: 21 Note that N = (0, 1) and S = (0, −1) are the North and South Poles of
S 1 . Let p = (x, y) ∈ S 1 and consider the set U = S 1 \ {N } and V = S 1 \ {S}. From
the definition of stereographic projection, φ : U → R and ψ : V → R are given by
x x
φ(x, y) = , ψ(x, y) = .
1−y 1+y
2.4 Stereographic Projection 57

Fig. 2.10 Stereographic


projection on S 1

Our claim is that {(U, φ), (V, ψ)} forms an atlas of S 1 . It is obvious that φ and ψ
 x
are homeomorphisms. Here x 2 + y 2 = 1 ⇒ x 2 = (1 − y)(1 + y). Let x = .
1−y
Then by virtue of foregoing equation, we have

  x 2 − 1
x 2 (1 − y)2 = x 2 ⇒ (x )2 (1 − y) = (1 + y) ⇒ y = .
1 + x 2

2x 
Similarly, x = . Hence, the inverse map φ−1 : R → U is given by
1 + x 2
 
−1  2x  x 2 − 1
φ (x ) = (x, y) = , .
1 + x 2 1 + x 2

Here φ(U ∩ V ) = R − {0} and ψ(U ∩ V ) = R − {0}. Now, ψ ◦ φ−1 : φ(U ∩ V ) →


1
ψ(U ∩ V ) is given by (ψ ◦ φ−1 )(t) = , is a C ∞ function. Similarly, φ ◦ ψ −1 is also
t
C ∞ function. This proves {(U, φ), (V, ψ)} forms a C ∞ atlas of S 1 . Hence S 1 is a
smooth manifold.

Problem 2.22 Using stereographic projection with an equatorial plane as the image
plane, prove that S 2 is a smooth manifold (Fig. 2.11).

Solution: 22 Note that N = (0, 0, 1) and S = (0, 0, −1) are the North and South
Poles of S 2 . Let p = (x, y, z) ∈ S 2 and consider the set U = S 2 − {N } and V =
S 2 \ {S}. Here φ : U → R2 and ψ : V → R2 are given by
   
x y x y
φ(x, y, z) = , , ψ(x, y, z) = , .
1−z 1−z 1+z 1+z

Our claim is that {(U, φ), (V, ψ)} forms an atlas of S 2 .


It is clear that φ and ψ are homeomorphisms. Moreover, the inverse map φ−1 :
R2 → U is given by
58 2 Manifold Theory

Fig. 2.11 Stereographic projection on S 2

 
2x  2y  x 2 + y 2 − 1
φ−1 (x  , y  ) = (x, y, z) = , , .
1 + x 2 + y 2 1 + x 2 + y 2 1 + x 2 + y 2

Note that φ(U ∩ V ) = R2 \ {(0, 0)} and ψ(U ∩ V ) = R2 − {(0, 0)}. Now, ψ ◦ φ−1 :
φ(U ∩ V ) → ψ(U ∩ V ) given by
 
x y
(ψ ◦ φ−1 )(x  , y  ) = ,
x 2 + y 2 x 2 + y 2

is a C ∞ function. Similarly, φ ◦ ψ −1 is also C ∞ function. This proves {(U, φ),


(V, ψ)} forms a C ∞ atlas of S 2 . Hence S 2 is a smooth manifold.

Problem 2.23 Show that the n-dimensional sphere S n is a smooth manifold.

Solution: 23 Here N = (0, 0, 0, . . . , 0, 1) and S = (0, 0, 0, . . . , 0, −1) are the North


     
n−zeroes n−zeroes
and South Poles of S n . Note that

U = S n \ {(0, 0, 0, . . . , 0, 1)}, V = S n − {(0, 0, 0, . . . , 0, −1)}


     
n−zeroes n−zeroes

and φ : U → Rn and ψ : V → Rn are given by


 
x1 x2 xn
φ(x1 , x2 , . . . , xn+1 ) = , ,..., ,
1 − xn+1 1 − xn+1 1 − xn+1
 
x1 x2 xn
ψ(x1 , x2 , . . . , xn+1 ) = , ,..., .
1 + xn+1 1 + xn+1 1 + xn+1
2.5 Orientable Surface 59

The inverse map φ−1 : Rn → U is


⎛ ⎞

⎜ 2x1 2x2 2xn xi 2 − 1 ⎟
φ−1 (x1 , x2 , . . . , xn ) = ⎜
⎝  ,  ...,  ,  ⎟
i
⎠.
1+ xi 2 1 + xi 2 1+ xi 2 1+ xi 2
i i i i

So, ψ ◦ φ−1 : φ(U ∩ V ) → ψ(U ∩ V ) is given by


⎛ ⎞
⎜ x1 x2 xn ⎟
(ψ ◦ φ−1 )(x1 , x2 , . . . , xn ) = ⎜
⎝  ,  , . . . ,  ⎟,
2⎠
(xi ) 2
(xi ) 2
(xi )
i i i

which is C ∞ in φ(U ∩ V ) = Rn − {(0, 0, 0 . . . , 0, 0)}. Hence {(U, φ), (V, ψ)} forms
a C ∞ atlas of S n . So S n is a smooth manifold.

2.5 Orientable Surface

A regular surface S is said to be orientable if there exists a family of surface patches


(coordinate neighbourhoods), which will cover S, in such a way that if a point p of
S belongs to the intersection of two surface patches of the family, then the transition
map between the surface patches of the family has positive Jacobian. The existence
of such a family of surface patches is called an orientation of S and S is called an
orientable surface. If such a family does not exist, the surface is called non-orientable.
S n (n > 1) forms an example of orientable surface. The sphere can be covered by
two coordinate neighbourhoods (using stereographic projection; refer to examples
on stereographic projection). If W is the intersection of these coordinate neighbour-
hoods, W is connected. Let p ∈ W be fixed. Since the Jacobian is positive at p ∈ W ,
it follows from the connectedness of W that the Jacobian is everywhere positive.
Hence S n (n > 1) is orientable.
Now, we are going to focus our attention on an example of a non-orientable
surface, the so-called Möbius Strip. From the geometrical point of view, a Möbius
Band has the property that a figure moving around on the surface can come back to
its starting point and transform into its mirror image, so it is impossible to decide
consistently which of the two possible rotational directions on the surface to call
clockwise and which counterclockwise, or which is the front and which is the back
side. Let us denote the Möbius band by M (Fig. 2.12).
60 2 Manifold Theory

Fig. 2.12 Möbius band

The Möbius band is obtained by rotating an open segment AB around its mid-
point C = (1, 0, 0) at the same time as C moves around a circle S 1 ≡ x 2 + y 2 = 1
in such a manner that as C moves once around S 1 , the segment AB makes a half
turn around C. After C has rotated by an angle v around the z-axis, AB should have
v
rotated by around C in the plane containing C and the z-axis. Initially, the point
2
of the segment AB is at (1, 0, u).
Let U1 = {(u, v) : (u, v) ∈ R2 , − 12 < u < 21 , 0 < v < 2π} and U2 = {(ũ, ṽ) :
(ũ, ṽ) ∈ R2 , − 21 < ũ < 21 , −π < ṽ < π}. Each of U1 and U2 are open in R2 . We
define φ : U1 → M and ψ : U2 → M by
2.5 Orientable Surface 61
 v  v v
φ(u, v) = 1 − u sin cos v, 1 − u sin sin v, u cos
 2  2  2 
ṽ ṽ ṽ
ψ(ũ, ṽ) = 1 − ũ sin cos ṽ, 1 − ũ sin sin ṽ, ũ cos .
2 2 2

Here, φ(U1 ∩ U2 ) is not connected but consists of two connected components given
by

1 1
W1 = {φ(u, v) : (u, v) ∈ R2 , − < u < , 0 < v < π},
2 2
1 1
W2 = {φ(u, v) : (u, v) ∈ R2 , − < u < , π < v < 2π}.
2 2
Hence, φ(U1 ∩ U2 ) = W1 ∪ W2 . Geometrically, φ(U1 ∩ U2 ) is the union of the rectan-
gles given by 0 < v < π and π < v < 2π, with − 21 < u < 21 . If 0 < v < π, then it is
clear that (φ−1 ◦ ψ)(u, v) = (u, v). If π < v < 2π, we have v − ṽ = 2π. Now com-
bining sin ṽ2 = − sin v2 , cos ṽ2 = − cos v2 and φ(u, v) = ψ(ũ, ṽ) implies ũ = −u.
Hence

(φ−1 ◦ ψ)(u, v) = (u, v), if 0 < v < π


= (−u, v − 2π), if π < v < 2π.

Hence, φ−1 ◦ ψ forms the transition map between the two surface patches φ(u, v)
and ψ(u, v) for M. This proves {(U1 , φ), (U2 , ψ)} forms an C ∞ atlas for M, hence
a smooth manifold of dimension 2. Here
  v v v
φu u=0 = − sin cos v, − sin sin v, cos ,
 2 2 2
φv u=0 = (− sin v, cos v, 0).

  v v v
Therefore, φu u=0 × φv u=0 = (− cos v cos , − sin v cos , − sin ). Now the unit
2 2 2
φ u × φv
normal is given by Nφ = = φu × φv . If possible, let us assume M to be
||φu × φv ||
orientable. Then ∃ a well-defined unit normal vector N : M → R3 at every point of
M, which varies smoothly over M. At φ(0, v) ∈ S 1 , we have N = μ(v)Nφ where
μ : (0, 2π) → R is smooth. Also, μ(v) = ±1 ∀ v. It follows that

μ(v) = +1, ∀ v ∈ (0, 2π) or μ(v) = −1, ∀ v ∈ (0, 2π).

Depending on the changes u → v, ũ → ṽ has to be made, suppose μ = 1. As N is


smooth, at φ(0, 0) = φ(0, 2π), we have
62 2 Manifold Theory

N = lim Nφ = (−1, 0, 0)
v→0
N = lim Nφ = (1, 0, 0).
v→2π

This leads to a contradiction, which proves M is non-orientable.


∂(u, v)
Remark 2.19 The Jacobian(J) of φ−1 ◦ ψ, i.e. J (φ−1 ◦ ψ) = = 1 > 0 in
∂(ũ, ṽ)
∂(u, v)
W1 and J (φ−1 ◦ ψ) = = −1 < 0 in W2 .
∂(ũ, ṽ)

2.6 Product Manifold

Let M1m and M2n be smooth manifolds, with differentiable structures A1 and A2 , of
dimensions m and n, respectively. First we prove that M1 × M2 , with the product
topology, is a (m + n)-dimensional topological manifold.
Since M1 and M2 is a smooth manifold, the topology of M1 and M2 is Hausdorff
and second countable. Therefore, the product topology of M1 × M2 is Hausdorff and
second countable.
Let (x1 , x2 ) ∈ M1 × M2 . Since x1 ∈ M1 and M1 is a m-dimensional smooth man-
ifold with differentiable structure A1 , ∃ a coordinate chart (U, φ) ∈ A1 such that
x1 ∈ U . Similarly, ∃ a coordinate chart (V, ψ) ∈ A2 such that x2 ∈ V . Hence, φ is a
homeomorphism from the open subset U of M1 onto the open subset φ(U ) of R m .
Similarly, ψ is a homeomorphism from the open subset V of M2 onto the open subset
ψ(V ) of R n . Since U is an open neighbourhood of x1 and V is an open neighbourhood
of x2 , the Cartesian product U × V is an open neighbourhood of (x1 , x2 ) ∈ M1 × M2 .
Furthermore, the Cartesian product φ(U ) × ψ(V ) is open in Rm × Rn ≡ Rm+n . Let
us define a function

φ × ψ : U × V → φ(U ) × ψ(V ), (x1 , x2 ) → (φ(x1 ), ψ(x2 )), ∀ (x1 , x2 ) ∈ U × V.

Then φ × ψ is homeomorphic to an open subset φ(U ) × ψ(V ) in Rm+n . Hence,


M1 × M2 , with the product topology, is a (m + n)-dimensional topological manifold.
Let us denote the collection of coordinate charts of M1 × M2 by C. Then

C = {(U × V, φ × ψ) : (U, φ) ∈ A1 , (V, ψ) ∈ A2 }.



Here {U × V : (U × V, φ × ψ) ∈ C} = M1 × M2 . To show that M1 × M2 is a
(m + n)-dimensional smooth manifold, it only remains to show that all pairs of
members in C are C k -compatible, for every k = 1, 2, 3, . . . , . . ..
Let (U × V, φ × ψ) ∈ C and (Ũ × Ṽ , φ̃ × ψ̃) ∈ C. Then the transition function
   
(φ × ψ) ◦ (φ̃ × ψ̃)−1 = (φ ◦ φ̃−1 ) × (ψ ◦ ψ̃ −1 ) : (φ̃ × ψ̃) (U × V ) ∩ (Ũ × Ṽ ) → (φ × ψ) (U × V ) ∩ (Ũ × Ṽ )
2.6 Product Manifold 63

Fig. 2.13 Product manifold

is C k , as both (φ ◦ φ̃−1 ) and (ψ ◦ ψ̃ −1 ) are C k compatible. By similar reasoning, the


transition function
   
(ψ × φ) ◦ (ψ̃ × φ̃)−1 = (ψ ◦ ψ̃ −1 ) × (φ ◦ φ̃−1 ) : (ψ̃ × φ̃) (U × V ) ∩ (Ũ × Ṽ ) → (ψ × φ) (U × V ) ∩ (Ũ × Ṽ )

is also so. Hence, C forms a C ∞ atlas on M1 × M2 , so it is a smooth manifold of


dimension (m + n) (Fig. 2.13).

Example 2.7 Since S 1 is a smooth manifold, the torus S 1 × S 1 is a smooth manifold.

Problem 2.24 Prove that the infinite cylinder is a smooth manifold.

Solution: 24 Since the infinite cylinder can be expressed as a product of smooth


manifolds S 1 and R, it is a smooth manifold.

Exercise

Exercise 2.9 Prove that the n-dimensional torus is a smooth manifold.


64 2 Manifold Theory

2.7 Smooth Function on Smooth Manifold

Let M be a smooth manifold of dimension n (Fig. 2.14).


Let us recall the definition of a smooth function on a smooth manifold. A function
f : M → R is said to be of class C ∞ or smooth at a point p ∈ M, if for every chart
(U, φ) of p in M, the function

f ◦ φ−1 : φ(U ) ⊂ Rn → R

is of class C ∞ at φ( p). The function f is said to be C ∞ on M if it is C ∞ at every


point of M.
Proposition 2.3 The notion of smoothness of a map is independent of the choice of
a coordinate chart.
Proof Let M be an n-dimensional smooth manifold and let f : M → R be smooth
at every point of M. Hence, for every chart (U, φ) of p ∈ M, f ◦ φ−1 : φ(U ) ⊂
Rn → R is of class C ∞ at φ( p). If (V, ψ) is any other chart of p, where p ∈ U ∩ V ,
then on ψ(U ∩ V )
f ◦ ψ −1 = ( f ◦ φ−1 ) ◦ (φ ◦ ψ −1 ).

Let A = {(Uα , φα )} be an atlas of M, where each (U, φ), (V, ψ) ∈ A. Thus, each
φ ◦ ψ −1 , ψ ◦ φ−1 is of class C ∞ and hence from above, f ◦ ψ −1 is C ∞ on ψ(U ∩ V ).

Fig. 2.14 Smooth function


2.7 Smooth Function on Smooth Manifold 65

Consequently, the smoothness of f on M is independent of the choice of any


coordinate chart on M. This completes the proof.

Proposition 2.4 Let M be an n-dimensional smooth manifold and f : M → R be


a map. Then the followings are equivalent:
(i) The map f : M → R is of class C ∞ .
(ii) f ◦ φ−1 : φ(U ) ⊂ Rn → R is of class C ∞ , for every chart (U, φ) ∈ A, A being
the atlas.
(iii) f ◦ ψ −1 : ψ(V ) ⊂ Rn → R is of class C ∞ , for every chart (V, ψ) ∈ A, A
being the atlas.

Proof (ii)⇒(i): Let (ii) hold. Then f = ( f ◦ φ−1 ) ◦ φ : M → R is of class C ∞


at R, as (U, φ) is given to be a chart. Thus (ii)⇒(i).
(i)⇒(iii): Let (i) hold. Then by definition, ∃ a chart (U, φ) at p ∈ M such that
f ◦ φ−1 : φ(U ) ⊂ Rn → R is of class C ∞ at φ( p).
If (V, ψ) is any chart at p on M, p ∈ U ∩ V , then

f ◦ ψ −1 = ( f ◦ φ−1 ) ◦ (φ ◦ ψ −1 )

is of class C ∞ on ψ(V ). Thus (i)⇒(iii).


(iii)⇒(ii): Let (iii) hold. Let A be an atlas on M, where (U, φ), (V, ψ) ∈ A. By
definition, φ ◦ ψ −1 and ψ ◦ φ−1 are C ∞ -related. Thus

f ◦ φ−1 = ( f ◦ ψ −1 ) ◦ (ψ ◦ φ−1 )

is of class C ∞ on φ(U ).

We shall often denote by F(M), the set of all C ∞ -functions on M and will sometime
denote by F( p), all the C ∞ -functions at p of M. It is to be noted that such F(M) is
(i) an algebra over R
(ii) a module over R,
where the defining relations are

⎨ (a) ( f + g)( p) = f ( p) + g( p), ;
(b) ( f g)( p) = f ( p)g( p), ;

(c) (λ f )( p) = λ f ( p), ∀ f, g ∈ F(M), λ ∈ R .

Now we are going to discuss smooth functions between smooth manifolds.


Let M be an n-dimensional and N be an m-dimensional manifold (Fig. 2.15).
A mapping f : M → N is said to be a differentiable mapping of class C k if for
every chart (U, φ) of p ∈ M n and every chart (V, ψ) of f ( p) ∈ N m ,

(i) f (U ) ⊂ V
(2.9)
(ii) ψ ◦ f ◦ φ−1 : φ(U ) ⊂ Rn → ψ(V ) ⊂ Rm , is of class C k .
66 2 Manifold Theory

Fig. 2.15 Differentiable mapping

By differentiable mapping, we shall mean unless otherwise stated, a mapping


of class C ∞ .
Let M and N be two n-dimensional smooth manifolds. A mapping f : M → N
is said to be a diffeomorphism if

⎨ (i) f is a differentiable mapping.
(ii) f is a bijection (2.10)

(iii) f −1 is of class C ∞ .

In such cases, M, N are said to be diffeomorphic to each other. A diffeomorphism


f of M onto itself is called a transformation on M.

Proposition 2.5 The notion of smoothness of a map between two smooth manifolds
is independent of the choice of a coordinate chart.

Proof Let M be an n-dimensional and N be an m-dimensional manifold. Let us


consider the map f : M → N to be differentiable of class C ∞ . Note that (U, φ) and
(V, ψ) are any charts about p ∈ M and f ( p) ∈ N respectively with f (U ) ⊂ V such
that the map ψ ◦ f ◦ φ−1 being C ∞ . Then for p ∈ M, suppose there exists charts
(Ũ , φ̃) of p and (Ṽ , ψ̃) of f ( p) with f (Ũ ) ⊂ Ṽ such that the map

ψ̃ ◦ f ◦ φ̃−1 : φ̃(Ũ ) ⊂ Rn → ψ̃(Ṽ ) ⊂ Rm , is of class C ∞ at φ̃( p).


2.7 Smooth Function on Smooth Manifold 67

However, φ̃ ◦ φ−1 and ψ ◦ ψ̃ −1 are C ∞ on open subset of Euclidean spaces. Hence,

(ψ ◦ ψ̃ −1 ) ◦ (ψ̃ ◦ f ◦ φ̃−1 ) ◦ (φ̃ ◦ φ−1 ) = ψ ◦ f ◦ φ−1

is C ∞ . This proves that the C ∞ structure of a map does not depend on any coordinate
chart chosen.

Proposition 2.6 If (U, φ) is a chart on a manifold M of dimension n, then the map


φ : U ⊂ M → φ(U ) ⊂ Rn is a diffeomorphism.

Proof Note that (U, φ) being a chart, φ is a homeomorphism.


Further, we use the atlas {(U, φ)} with a single chart on U and the atlas
{φ(U ), Iφ(U ) } with a single chart on φ(U ). Then Iφ(U ) ◦ φ ◦ φ−1 : φ(U ) ⊂ Rn →
φ(U ) is the identity map. Thus, Iφ(U ) ◦ φ ◦ φ−1 is of class C ∞ and by (2.9), φ is a
differentiable mapping. Furthermore, φ ◦ φ−1 ◦ Iφ(U ) : φ(U ) → φ(U ) is an identity
map, hence is of class C ∞ . Thus, all the conditions of (2.10) are satisfied by φ and
so is diffeomorphism.

Proposition 2.7 Let U ⊂ M be an open subset of a manifold M of dimension n.


If φ : U ⊂ M → φ(U ) ⊂ Rn is a diffeomorphism, onto an open subset of Rn , then
(U, φ) is a chart on M.

Proof For any chart (U, φα ) on M, both φα and φ−1 α are of class C

(refer to
Proposition 2.6).
Now φ being a diffeomorphism, it is C ∞ . Consequently, the composite mappings
φ ◦ φ−1
α and φα ◦ φ
−1
are of class C ∞ . Hence (U, φ) is compatible with an atlas on
M. Thus (U, φ) is a chart on M.

Problem 2.25 If (x 1 , x 2 , . . . , x n ) and (y 1 , y 2 , . . . , y m ) are respectively the local


coordinate systems defined in the neighbourhood U of p ∈ M n and V of f ( p) ∈ N m ,
then it can be shown that

y j ◦ f = g j (x 1 , x 2 , . . . , x n ), where (2.11)
−1
g (q) = (ψ ◦ f ◦ φ )(q), q ∈ φ(U ).
j
(2.12)

Solution: 25 Let φ( p) = q, p ∈ U ⊂ M n . Then

g(φ( p)) = (ψ ◦ f ◦ φ−1 )(φ( p)) = (ψ ◦ f )( p)


or u j (g(φ( p))) = u j (ψ( f ( p))) = (u j ◦ ψ)( f ( p))
or g j (φ( p)) = y j ( f ( p)) by (2.1), (2.3)
or g j (x 1 ( p), . . . , x n ( p)) = (y j ◦ f )( p) by (2.2)
i.e. y j ◦ f = g j (x 1 , x 2 , . . . , x n ).
68 2 Manifold Theory

Fig. 2.16 Differentiable


map between punctured
sphere and cylinder

Problem 2.26 Obtain a differentiable map between the punctured sphere at two
points (0, 0, 1), (0, 0, −1) and the cylinder N with infinite ends (Fig. 2.16).

Solution: 26 Let us consider the punctured sphere M = S 2 − {(0, 0, 1), (0, 0, −1)}
where S 2 = {(x, y, z) : (x, y, z) ∈ R3 , x 2 + y 2 + z 2 = 1}. Here, the coordinate
neighbourhood (U, φ) is given by U = S 2 − {(0, 0, 1), (0, 0, −1)}, φ : U → R2 ,
where φ(cos v cos u, cos v sin u, sin v) = (x, y). Therefore, φ−1 : R2 → U is
defined by φ−1 (x, y) = (cos v cos u, cos v sin u, sin v).
Let us consider the cylinder N = {(x̃, ỹ, z̃) : (x̃, ỹ, z̃) ∈ R3 , x̃ 2 + ỹ 2 = 1, 0 <
z̃ < 1}. Here, the coordinate neighbourhood (V, ψ) is given by V = N and ψ :
N → R2 , where ψ(cos u, sin u, z) = (x, z). Therefore, ψ −1 : R2 → V is defined by
ψ −1 (x, z) = (cos u, sin u, z).
Let us define a map f : M → N by (x, y, z) → (x̃, ỹ, z). In other words, we can
say that the line joining p and f ( p) is parallel to the x y-plane and orthogonal to
the z-axis. Since the point (cos v cos u, sin v sin u, sin v) is moving from the sphere
M parallel to the x y-plane and is orthogonal to the z-axis, therefore the x and y
components of (cos v cos u, sin v sin u, sin v) will take the coordinate of the cylinder
N but the z-component, i.e. sin v will remain unchanged. Hence,

f (φ−1 (x, y)) = (cos u, sin u, sin v) = ψ −1 (x, sin v)


i.e. (ψ ◦ f ◦ φ−1 )(x, y) = (x, sin v),

which shows ψ ◦ f ◦ φ−1 is differentiable.

Exercise
Exercise 2.10 Let M and N be two smooth manifolds with M = N = R. Let (U, φ)
and (V, ψ) be two charts on M and N respectively, where U = R, φ : U → R is
the identity mapping and V = R, ψ : V → R is the mapping defined by ψ(x) = x 3 .
Show that the two structures defined on R are not C ∞ -related even though M and
N are diffeomorphic where f : M → N is defined by f (t) = t 1/3 .
2.8 Differential Curve and Tangent Vector 69

2.8 Differential Curve and Tangent Vector

We are now in a position to introduce one of the important concepts of geometry, i.e.
tangent vector. Geometers prefer to define the tangent vector at a point with respect
to a curve. Hence, at first we shall define a curve on a manifold.
A differentiable curve at p on a manifold M is a differentiable mapping σ :
[a, b] ⊂ R → M n such that σ(t0 ) = p, a ≤ t0 ≤ b (Fig. 2.17).
Then by (1.1) and (2.1), we obtain

(x i ◦ σ)(t) = u i (φ(σ(t))) = u i (σ 1 (t), σ 2 (t), . . . , σ n (t)) = σ i (t). (2.13)

Often, we write it as
x i (t) = σ i (t). (2.14)

The tangent vector to the curve σ(t) at p of M is a function X p : F( p) → R defined


by
⎧ !
⎨ X f = lim f (σ(t + h)) − f (σ(t)) d 
p
= f (σ(t))t=t , ∀ f ∈ F( p)
h→0 h dt 0
⎩ t=t0
σ(t0 ) = p.
(2.15)
Note that
d 
X p (a f + bg) = (a f + bg)(σ(t))t=t , ∀ f, g ∈ F( p), a, b ∈ R; by (2.15)
dt 0

d  d 
= a f (σ(t))t=t + b g(σ(t))t=t , by (1.8)
dt 0 dt 0

X p (a f + bg) = a(X p f ) + b(X p g), by (2.15). (2.16)

Fig. 2.17 Differential curve


70 2 Manifold Theory

Again

d 
X p ( f g) = ( f g)(σ(t))t=t , by (2.15)
dt 0

d 
= f (σ(t))g(σ(t))t=t , by (1.8)
dt 0

d  d 
= f (σ(t))t=t g(σ(t0 )) + f (σ(t0 )) g(σ(t))t=t , by (1.9)
dt 0 dt 0

X p ( f g) = (X p f )g( p) + f ( p)X p g. (2.17)

Equations (2.16) and (2.17) are respectively known as linearity property and Leib-
nitz Product Rule. Thus, the tangent vector at a point on a manifold is a derivation
at that point.
Let Tp (M) denote the set of all tangent vectors at p of M. We define

(X p + Y p ) f = X p f + Y p f, ∀ X p , Y p ∈ T p (M)
(2.18)
(λX p ) f = λ(X p f ), λ ∈ R.

Clearly, Tp (M) is a real vector space (refer to any standard textbook of Linear Alge-
bra). Hence, Tp (M) must have a basis.
If (x 1 , x 2 , . . . , x n ) is the local coordinate system in a neighbourhood U of p ∈ M,

then for each i = 1, 2, . . . , n, we define a mapping : F( p) → R by
∂x i
 
 ∂  ∂f
f = ( p), ∀ f ∈ F( p). (2.19)
∂x i p ∂x i (t)
 

Clearly, each : i = 1, 2, 3, . . . , n satisfies (2.16) and (2.17).
∂x i p
Let us define a differentiable curve σ : [a, b] ⊂ R → M by

σ i (t) = σ i (t0 ) for fixed i
(2.20)
σ j (t) = 0, j = 1, 2, 3, . . . , i − 1, i + 1, . . . , n.

Then

d   ∂ f (σ(t)) dσ i (t) 
f (σ(t))t=t =  , by chain rule
dt 0
i
∂σ i (t) dt t=t0
∂ f (σ(t0 )) 
= , for fixed i, by (2.20)
∂σ i (t) t=t0
∂ f ( p)
= , by (2.14)
∂x i (t)
2.8 Differential Curve and Tangent Vector 71
 

= f, by (2.19).
∂x i p

 ∂ 
Thus, each : i = 1, 2, 3, . . . , n is a tangent vector to the curve σ defined by
∂x i p
(2.20). Further from (2.15), we have

d 
Xp f = f (σ(t))
dt t=t0

 ∂ f (σ(t0 ))  d x i (t) 
= 
i
∂x i (t) dt t=t0


 d x i (t)    
 ∂
=  f
t=t0 ∂x
dt i
i p
  

= ξ ( p)
i
f, say, where
i
∂x i p
⎧  
⎨ d x i (t)
ξ i ( p) = , i = 1, 2, 3, . . . , n
dt (2.21)
⎩ i
ξ : M → R, are differentiable functions on M.

Thus we write
  

Xp = ξ i ( p) , ∀ f ∈ F( p). (2.22)
∂x i p

     
∂ ∂
Finally, if we assume that ξ ( p)
i
= 0, then ξ ( p)
i
x k = 0,
∂x i p ∂x i p
where x k ∈ F( p). Then ξ k ( p) = 0, by (2.19). Proceeding in this manner, we can say
that
ξ 1 ( p) = ξ 2 ( p) = · · · = ξ n ( p) = 0.

∂ 
Thus, the set : i = 1, 2, 3, . . . , n is linearly independent. We can now state
∂x i
the following.
Theorem 2.4 If (x 1 , x 2 , . . . , x n ) is a local coordinate system in a neighbourhood U
of a point p in an n-dimensional manifold M, the basis of the tangent space T p (M)
∂ 
is given by : i = 1, 2, 3, . . . , n and every X p ∈ T p (M) can be expressed
∂x i
uniquely by (2.22).
72 2 Manifold Theory

Problem 2.27 Let X = (2, 3, 0) ∈ R3 . Find X p f for a fixed point p = (−2, π, 1)


where f = x 1 x 3 cos x 2 .

Solution: 27 Comparing with (2.22), we see that

ξ 1 ( p) = 2, ξ 2 ( p) = 3, ξ 3 ( p) = 0.

Further,
 
∂f
= (x 3 cos x 2 ) p = −1;
∂x 1 p
 
∂f
= (−x 1 x 3 sin x 3 ) p = 0;
∂x 2 p
 
∂f
= 2.
∂x 3 p

Thus X p f = −2.
∂ ∂
Problem 2.28 Let X = 2x − 2y be a vector in R2 . Find X p f where f =
∂x ∂y
2x + y 3 , p = (x, y).

Solution: 28 As done in the previous problem,

∂ f  ∂ f 
ξ 1 ( p) = 2x, ξ 2 ( p) = −2y, = 2, = 3y 2 .
∂x p ∂y p

Thus X p f = 4x − 6y 3 .

∂ ∂
Problem 2.29 Let X = + be a vector in R2 . Find X p f for a fixed point
∂x ∂z
p = (1, 1, 0) where f = x z cos y.

Solution: 29 In this case


  
X p f = 1(z cos y) p + 0(−x z sin y) p + 1(x cos y) p = cos 1.

Problem 2.30 If C is a constant function on a manifold M and X is a tangent vector


to some curve σ on M, show that X · C = 0.
2.8 Differential Curve and Tangent Vector 73

Solution: 30 Here by (2.16), we have X (1) = 0, i.e. X · 1 = 0. Again by virtue


of linearity, X · C = X (C · 1) = C(X · 1), as C is a constant function. Hence
X · C = 0.

Problem 2.31 Let f = ((x 1 )3 − 2)x 3 + (x 2 x 3 − 1)x 1 . Find X p f, p = (x 1 , x 2 , x 3 ).

Solution: 31 Here

X p f = ξ 1 ( p){3(x 1 )2 x 3 + (x 2 x 3 − 1)} p + ξ 2 ( p)(x 1 x 3 ) p + ξ 3 ( p){(x 1 )3 − 2 + x 1 x 2 } p , by (2.21).

Exercises
Exercise 2.11 Find the tangent vector
(i) to the curve σ ∈ Rn where σ i = a i + bi t, a i , bi ∈ R for every i.
(ii) to the curve σ(t) = (t2 , t 3 ) on R2 .  
cos 2t − sin 2t x
(iii) to the curve σ p (t) = at t = 0.
sin 2t cos 2t y

Exercise 2.12 (i) Consider the curve γ(t) = (cos t, sin t) ∈ R2 , t ∈ (0, π). Find
π
the vector X tangent to γ at . Calculate X f where f : R2 → R is defined by
4
f = 2x + y 3 .
(ii) Consider the curve ψ in R2 defined by x = sin t, y = cos t, t ∈ (−π, π) and the
2
map f : R → R defined by f (x, y) = x 3 y. Find the vector X tangent to ψ at
π
t = and compute X f .
2
Exercise 2.13 Let X = (2, −3, 4) ∈ R3 . For a fixed point p = (2, 5, 7), compute
X p f where
(i) f : R3 → R is defined by f = x 3 y.
(ii) f : R3 → R is defined by f = z 7 .
(iii) f : R3 → R is defined by f = e x cos z.

Answers
∂ ∂ ∂ ∂ ∂ ∂ ∂
2.11 (i) b1 1 + b2 2 + · · · + bn n (ii) 2t + 3t 2 or 2t 1 + 3t 2 2 .
∂x ∂x ∂x ∂x ∂y ∂x ∂x
∂ ∂
(iii) −2y + 2x .
∂x ∂y
1 ∂ 1 ∂ 1 ∂
2.12 (i) − √ +√ ; − √ (ii) − ; −1.
2 ∂x 2 ∂x 2 2 ∂y
2.13 (i) 96 (ii) 4 · 77 (iii) 2e2 (cos 7 − 2 sin 7).
74 2 Manifold Theory

2.9 Inverse Function Theorem for Smooth Manifold

In continuation with the Inverse Function Theorem for Rn , stated and proved in
Chap. 1, the following theorem deals with the study of Inverse Function Theorem
for arbitrary smooth manifolds.
Theorem 2.5 Let M and N be n-dimensional smooth manifolds and F : M → N
be a smooth map. Let p ∈ M. If F∗ : Tp (M) → TF( p) (N ) is invertible (i.e. 1 − 1 and
onto) at p, then ∃ an open neighbourhood U of p, and an open neighbourhood V of
F( p) such that F : U → V is a diffeomorphism.
Proof Since M is an n-dimensional smooth manifold and p ∈ M, therefore there
exists an admissible coordinate chart  (Ũ , φ) of M such that p ∈ Ũ . For every q ∈
∂  ∂  ∂ 
Ũ , let , , . . . , n q be a coordinate basis of Tq (M) corresponding
∂x 1 q ∂x 2 q ∂x
to (Ũ , φ). Again, since N is an n-dimensional smooth manifold and F( p) ∈ N ,
therefore there exists an admissible coordinate chart(Ṽ , ψ) of M such that F( p)  ∈ Ṽ
∂  ∂  ∂ 
and (ψ ◦ F)( p) = 0. For every r ∈ Ṽ , suppose , , . . . , n r is a
∂ y1 r ∂ y2 r ∂y
coordinate basis of Tr (N ) corresponding to (Ṽ , ψ). Here, the matrix representation
of the linear map f ∗ at the point p with respect to some basis, denoted by ( f ∗ ), is
the n × n order matrix. Since the linear map f ∗ is invertible, therefore det( f ∗ ) =
0. Furthermore, the map ψ ◦ F ◦ φ−1 : φ(Ũ ∩ F −1 (Ṽ )) → ψ(Ṽ ) is smooth. Also,
(ψ ◦ F ◦ φ−1 )(φ( p)) = ψ(F( p)) = 0. Moreover, it is clear that φ(Ũ ∩ F −1 (Ṽ )) is
an open neighbourhood of φ( p) ∈ Rn and ψ(Ṽ ) is an open neighbourhood of 0 ∈ Rn .
Now, by virtue of Inverse Function Theorem for Rn , ∃ an open neighbourhood Ū of
φ( p) satisfying
• Ū ⊂ φ(Ũ ∩ F −1 (Ṽ ));
• (ψ ◦ F ◦ φ−1 )(Ū ) is an open neighbourhood of 0;
• ψ ◦ F ◦ φ−1 has a smooth inverse on (ψ ◦ F ◦ φ−1 )(Ū ).
Set U = φ−1 (Ū ) and V = (F ◦ φ−1 )(Ū ). Our claim is that U is an open neigh-
bourhood of p ∈ M. Since Ū is open and contained in φ(Ũ ∩ F −1 (Ṽ ))(⊂ φ(Ũ )),
and φ(Ũ ) is open, Ū is open in φ(Ũ ), hence φ−1 (Ū ) is open in Ũ . Moreover,
φ−1 (Ū ) = U (say) is open in M. Since φ( p) ∈ Ū , therefore p ∈ U .
Since ψ ◦ F ◦ φ−1 has a smooth inverse, therefore ψ ◦ F ◦ φ−1 is continuous.
Also φ ◦ F −1 is continuous, as ψ is so. Furthermore, (F ◦ φ−1 )(Ū ) = V is open.
Since φ( p) ∈ Ū , therefore F( p) ∈ V . This shows V is an open neighbourhood of
F( p).
Since ψ ◦ F ◦ φ−1 has an inverse, therefore ψ ◦ F ◦ φ−1 is one-to-one and onto.
Thus, the composite map ψ −1 (ψ ◦ F ◦ φ−1 ) ◦ φ = F is also one-to-one and onto.
Since ψ ◦ F ◦ φ−1 has a smooth inverse, φ ◦ F −1 ◦ ψ −1 is smooth. This gives F −1 :
V → U is smooth. This completes the proof.
Let M and N be n-dimensional smooth manifolds and F : M → N be a smooth
map. Let p ∈ M. If ∃ an open neighbourhood U of p, such that the neighbourhood
2.9 Inverse Function Theorem for Smooth Manifold 75

F(U ) of F( p) is open in N and F : U → F(U ) is a diffeomorphism, then we say


that F is a local diffeomorphism.

Problem 2.32 Let M and N be n-dimensional smooth manifolds and F : M → N


be a local diffeomorphism. Prove that F is an open map.

Solution: 32 Let p ∈ U . Since F is a local diffeomorphism and p ∈ M, ∃ an open


neighbourhood V of p such that F(V ) is open in N , and the map F|V : V → F(V ) is
a diffeomorphism. Since U, V are open neighbourhoods of p, U ∩ V is also an open
neighbourhood of p, in V . Since F|V : V → F(V ) is a diffeomorphism, therefore
it is a homeomorphism. This implies F|V (U ∩ V ) is open in F(V ), as U ∩ V is
open in V . Moreover, F(V ) being open in N , therefore F|V (U ∩ V ) is open in N .
Since p ∈ U ∩ V , F( p) ∈ F(U ∩ V ). Thus, F(U ∩ V ) ⊂ F(V ) and F(U ∩ V ) is
an open neighbourhood of F( p).

Remark 2.20 Let M, N , P be smooth manifolds. Let F : M → N and G : N → P


be local diffeomorphisms. The composite map G ◦ F : M → P is a local diffeomor-
phism.

Let M and N be respectively n- and m-dimensional smooth manifolds. If F :


M → N is continuous, and for every p ∈ M, ∃ an open neighbourhood U of p such
that F(U ) is an open neighbourhood of F( p) and the map F|U : U → F(U ) is a
homeomorphism, then F : M → N is said to be a local homeomorphism.

Remark 2.21 If F : M → N is a local diffeomorphism, then it is a local homeo-


morphism.

Remark 2.22 Let M, N be smooth manifolds. Let U be a non-empty open subset


of M. Let F : M → N be a local diffeomorphism. Then F|U : U → N is a local
diffeomorphism.

Note that diffeomorphism implies local diffeomorphism, but the converse is not
always true in general.
Example 2.8 Consider the map f : R2 → R 2
defined by f (x, y) = (e x cos y,
 e cos y −e sin y
x x
e x sin y). Here f (x, y) = and the Jacobian of the matrix is
e x sin y e x cos y

non-zero, which shows f (x, y) is invertible. But f is not one-to-one, since it is of
period 2π. So f is a local diffeomorphism but not a diffeomorphism.

Problem 2.33 Let M and N be respectively n- and m-dimensional smooth manifolds


and F : M → N be 1 − 1 and onto. Suppose F is a local diffeomorphism. Then prove
that F is a diffeomorphism.

Solution: 33 Since F : M → N is a local diffeomorphism, it is a local homeomor-


phism and hence continuous. Also F is an open map. Since F : M → N is 1 − 1
and onto, F is continuous and F is an open map, F : M → N is a homeomorphism.
It remains to show that F : M → N is smooth and its inverse is also so.
76 2 Manifold Theory

For any p ∈ M, we have to find an admissible coordinate chart (U, φ) of M with


p ∈ U , and an admissible coordinate chart (V, ψ) of N with F( p) ∈ V such that
ψ ◦ F ◦ φ−1 : φ(U ∩ F −1 (V )) → ψ(V ) is smooth.
Since p ∈ M, and F : M → N is a local diffeomorphism, ∃ an open neighbour-
hood Ũ of p such that F(Ũ ) is an open neighbourhood of F( p) in N , and the
map F|Ũ : Ũ → F(Ũ ) is a diffeomorphism. Furthermore, ∃ an admissible coor-
dinate chart (Û , φ̂) of M with p ∈ Û . Also, ∃ an admissible coordinate chart
(V̂ ,ψ̂) of N with F(  p) ∈ V̂ . Moreover, Ũ ∩ Û is an open neighbourhood of p.
So (Ũ ∩ Û ), φ̂|Ũ ∩Û forms an admissible chart of M satisfying p ∈ Ũ ∩ Û . Since
F|Ũ : Ũ → F(Ũ ) is a diffeomorphism, the map F|Ũ ∩Û : Ũ ∩ Û → F(Ũ ∩ Û ) is
a diffeomorphism. Since F : M → N is a homeomorphism, and Ũ ∩ Û is an open
neighbourhood of p, F(Ũ ∩ Û ) is an open neighbourhood of F( p).
Since (V̂ , ψ̂) is an admissible coordinate chart of N with V̂ being an open neigh-
bourhood of F( p), and F(Ũ ∩ Û ) is an open neighbourhood of F( p), V̂ ∩ F(Ũ ∩
Û ) is an open neighbourhood of F( p). Hence, (V̂ ∩ F(Ũ ∩ Û ), ψ̂|V̂ ∩F(Ũ ∩Û ) ) forms
an admissible coordinate chart of N with F( p) ∈ V̂ ∩ F(Ũ ∩ Û ). 
Since the map F|Ũ : Ũ → F(Ũ ) is a diffeomorphism, it is smooth. As (Ũ ∩

Û ), φ̂|Ũ ∩Û forms an admissible chart of M satisfying p ∈ Ũ ∩ Û and (V̂ ∩ F(Ũ ∩
Û ), ψ̂|V̂ ∩F(Ũ ∩Û ) ) forms an admissible coordinate chart of N with F( p) ∈ V̂ ∩
F(Ũ ∩ Û ), therefore
(ψ̂|V̂ ∩F(Ũ ∩Û ) ) ◦ F|Ũ ◦ (φ̂|Ũ ∩Û )−1

is smooth. This proves F : M → N is smooth.


Finally, since F : M → N is 1 − 1 and onto, therefore F −1 : N → M exists
and is 1 − 1 and onto. Also F −1 : N → M is a local diffeomorphism, therefore
proceeding as above F −1 : N → M is smooth. This completes the solution.

Problem 2.34 Let M and N be respectively n- and m-dimensional smooth manifolds


and F : M → N be a local diffeomorphism. Then F is a smooth immersion and
smooth submersion.

Solution: 34 We wish to show that, for every p ∈ M, rank F = n = m. Since p ∈


M and F is a local diffeomorphism, ∃ an open neighbourhood U of p such that
F(U ) is open in N , and the map F|U : U → F(U ) is a diffeomorphism. Since
U (= φ) is an open subset of M, and M is an n-dimensional smooth manifold,
U is an n-dimensional smooth manifold, so dim Tp (U ) = dim U = n. Similarly,
dim TF( p) (F(U )) = dim F(U ) = m. Since F|U : U → F(U ) is a diffeomorphism,
p ∈ U , the linear map f ∗ : Tp (U ) → TF( p) (F(U )) at p is an isomorphism (refer
to Exercise 2.55), which implies rank F = dim Tp (U ) = n = m = dim TF( p) (F(U ))
at p.

Problem 2.35 Let M and N be respectively n- and m-dimensional smooth manifolds


and F : M → N be a smooth immersion and smooth submersion. Then F is a local
diffeomorphism.
2.10 Vector Field 77

Solution: 35 Let p ∈ M. Our claim is to find an open neighbourhood U of p ∈ M


such that F(U ) is open in N , and the map F|U : U → F(U ) is a diffeomorphism.
Since F is a smooth immersion and smooth submersion, the linear map F∗ :
Tp (M) → TF( p) (N ) is 1 − 1 and onto at p ∈ M, therefore dim Tp (M) = dim TF( p)
(M), i.e. n = m. By Inverse Function Theorem for manifolds, ∃ an open neigh-
bourhood V of F( p) with F|U : U → V being a diffeomorphism. Hence F(U ) =
(F|U )(U ) = V . Moreover, F(U ) being open, F|U : U → F(U ) is a diffeomor-
phism.

Exercises
Exercise 2.14 Let M and N be n-dimensional smooth manifolds and F : M → N
be a smooth immersion. Then F is a local diffeomorphism.
Exercise 2.15 Let M and N be n-dimensional smooth manifolds and F : M → N
be a smooth submersion. Then F is a local diffeomorphism.

2.10 Vector Field

In classical notation, if to each point p of R3 or in a domain U of R3 , a vector


α : p → α( p) is specified, then we say that a vector field is given on R3 or in a
domain of R3 . In the same manner, we will introduce a vector field in a manifold M.
A vector field X on M is a correspondence that associates with each point p of M,
a vector X p ∈ T p (M). In fact, if f ∈ F(M) then X f is defined to be a real-valued
function on M, as follows:
(X f )( p) = X p f. (2.23)

A vector field X is called differentiable if X f is so for every f ∈ F(M). From (2.22),


a vector field X can be expressed as
 ∂
X= ξi . (2.24)
i
∂x i

Let χ(M) denote the set of all differentiable vector fields on M. We define

(X + Y ) f = X f + Y f
(2.25)
(λX ) f = λ(X f ), ∀ X, Y ∈ χ(M), λ ∈ R.

It can be shown that χ(M) is a vector space over R. We also define f X to be a vector
field on M as follows:

( f X )( p) = f ( p)X p , ∀ p ∈ M. (2.26)

Let us define a mapping [ , ] : F(M) → F(M) as


78 2 Manifold Theory

[X, Y ] f = X (Y f ) − Y (X f ). (2.27)

Such a bracket is also known as Lie Bracket of X and Y .


Exercises
Exercise 2.16 Show that for every X, Y, Z in χ(M) and f, g in F(M)
(a) [X, Y ] ∈ χ(M)
(b) [bX, Y ] = b[X, Y ] = [X, bY ], ∀ b ∈ R
(c) [X + Y, Z ] = [X, Z ] + [Y, Z ]
(d) [X, Y + Z ] = [X, Y ] + [X, Z ]
(e) [X, [Y, Z ]] + [Y, [Z , X ]] + [Z , [X, Y ]] = θ: Jacobi Identity
(f) [X, X ] = θ
(g) [X, Y ] = −[Y, X ].

Hints
2.16 (a). Show that [X, Y ] satisfies Linearity and Leibnitz Product rule.
Remark
Remark 2.23 χ(M) with the product rule given by (2.27) is an algebra, also called
Lie Algebra.

Problem 2.36 Using [X, X ] = θ, show that [X, Y ] = −[Y, X ].

Solution: 36 For every X, Y ∈ χ(M), X + Y ∈ χ(M). Using the hypothesis, we


have

[X + Y, X + Y ] = θ
i.e. [X, Y ] + [Y, X ] = θ,
∴ [X, Y ] = −[Y, X ].

Problem 2.37 Prove that [X, f Y ] = f [X, Y ] + (X f )Y .

Solution: 37 Note that

(( f X )h)( p) = ( f X ) p h, ∀ h ∈ F(M) by (2.23)


= f ( p)X p h, by (2.26).
Also, ( f (X h))( p) = f ( p)(X h)( p), by (1.8)
= f ( p)X p h, by (2.23).

Thus
( f X )h = f (X h), ∀ p ∈ M. (2.28)

Again (2.27) yields


2.10 Vector Field 79

[X, f Y ] = X {( f Y )h} − ( f Y )(X h)


= X { f (Y h)} − f {Y (X h)}, by (2.28)
= (X f )(Y h) + f {X (Y h)} − f {Y (X h)}, by (2.17)
= {(X f )Y }h + f {[X, Y ]h}, by (2.27) and (2.28)
= {(X f )Y }h + { f [X, Y ]}h, by (2.28)
i.e., [X, f Y ] = f [X, Y ] + (X f )Y, ∀ h.

∂ ∂ ∂
Problem 2.38 If X = ,Y = + e x , compute [X, Y ](0,1,0) .
∂x ∂y ∂z
Solution: 38 Note that
∂ ∂f ∂f  ∂ ∂ ∂ f
[X, Y ] f = + ex − + ex
∂x ∂ y ∂z ∂y ∂z ∂x
∂f
= ex
∂z

∴, [X, Y ] = e x , ∀ f
∂z

∂ 
Hence, [X, Y ](0,1,0) = .
∂z (0,1,0)

Problem 2.39 Find the general expression for Z ∈ χ(R2 ) where


! !
∂ ∂
, Z = Z and , Z = Z.
∂x1 ∂x2

Solution: 39 Let us assume that


∂ ∂
Z = λ(x1 , x2 ) + μ(x1 , x2 ) , λ, μ ∈ F(R2 ).
∂x1 ∂x2

Substituting the expression of Z and using (2.27), one gets after a few steps
!
∂ ∂λ(x1 , x2 ) ∂ ∂μ(x1 , x2 ) ∂
,Z = + .
∂x1 ∂x1 ∂x1 ∂x1 ∂x2

Similarly, !
∂ ∂λ(x1 , x2 ) ∂ ∂μ(x1 , x2 ) ∂
,Z = + .
∂x2 ∂x2 ∂x1 ∂x2 ∂x2

From the given condition,


! !
∂ ∂
, Z = Z and , Z = Z.
∂x1 ∂x2
80 2 Manifold Theory

Thus on comparing, we have

∂λ(x1 , x2 ) ∂μ(x1 , x2 ) ∂λ(x1 , x2 ) ∂μ(x1 , x2 )


= λ(x1 , x2 ); = μ(x1 , x2 ); = λ(x1 , x2 ); = μ(x1 , x2 ).
∂x1 ∂x1 ∂x2 ∂x2

From the first equation, we see that

λ(x1 , x2 ) = A f (x2 )e x1 , A being constant.

Substituting in the third equation, we get


{A f (x2 )e x1 } = A f (x2 )e x1
∂x2
A f  (x2 )e x1 = A f (x2 )e x1
∴, f (x2 ) = Ce x2 , C is constant.
Thus λ(x1 , x2 ) = ACe x2 e x1 = De x1 +x2 , say D = AC being constant.

By similar computation, from the second and fourth equations, it can be found that
μ(x1 , x2 ) = Bg(x2 )e x1 and after a brief calculation, μ(x1 , x2 ) = Ee x1 +x2 , E being a
constant. Thus,
∂ ∂
Z = De x1 +x2 + Ee x1 +x2 .
∂x1 ∂x2

Problem 2.40 Write in cylindrical coordinates, the vector field on R3 defined by

∂ ∂ ∂
X= + + .
∂x ∂y ∂z

Solution: 40 If (ρ, θ, z) is the cylindrical coordinate, then the Cartesian coordinate


(x, y, z) is given by
x = ρ cos θ, y = ρ sin θ, z = z.

Therefore, |J | = ρ. Let us write

∂ ∂ ∂
X = ξ 1 (ρ, θ, z) + ξ 2 (ρ, θ, z) + ξ 3 (ρ, θ, z) .
∂ρ ∂θ ∂θ
⎛ ⎞ ⎛ ⎞
ξ1 1
Then J ⎝ ξ 2 ⎠ = ⎝ 1 ⎠ , i.e., ξ 1 cos θ − ξ 2 ρ sin θ = 1; ξ 1 sin θ + ξ 2 ρ cos θ =
ξ3 1
1; ξ = 1. After a few steps, one gets from above
3
2.10 Vector Field 81

ξ 1 = cos θ + sin θ
1
ξ 2 = (cos θ − sin θ).
ρ

Thus in cylindrical coordinate, we have

∂ 1 ∂ ∂
X = (cos θ + sin θ) + (cos θ − sin θ) + .
∂ρ ρ ∂θ ∂z

Problem 2.41 If X = (x − y) ∂x − ∂∂y , Y = x 2 ∂x∂
+ y ∂∂y are vector fields on R2 ,
show that X, Y are linearly independent differentiable vector fields on R2 , if

x 2 − y 2 + x y = 0. Further, if Z = (x 2 − y 2 ) ∂x + (x 2 + y 2 ) ∂∂y is any vector of R2 ,
express Z = f X + gY , where f, g ∈ F(R2 ).

Solution: 41 Clearly, X, Y are differentiable, as x − y, x 2 , y are also so. If X, Y


are linearly independent, then for λX + μY = θ ⇒ λ = μ = θ, ∀ λ, μ ∈ R. Again
λX + μY = θ gives

∂ ∂
{λ(x − y) + μx 2 } + (−λ + μy) = θ.
∂x ∂y

∂ ∂ 
As , is a basis of T(x,y) (R2 ), we must have
∂x ∂ y

λ(x − y) + μx 2 = 0 = −λ + μy.

Therefore, λ = μy and μ(x 2 − y 2 + x y) = 0. Thus X, Y are linearly independent if


x 2 − y 2 + x y = 0. Writing Z = f X + gY , we find on comparing

x 2 − y 2 = f (x − y) + gx 2 and x 2 + y 2 = − f + g y.

One gets, after a brief calculation,

x 2 y − y3 − x 4 − x 2 y2 (x − y)(x + y + x 2 + y 2 )
f = , g= .
x + xy − y
2 2 x 2 + x y − y2

Exercises
Exercise 2.17 Show that
(a) [ f X, Y ] = f [X, Y ] − (Y f )X
(b) [ f X, gY ] = f g[X, Y ] + { f (X g)}Y − {g(Y f )}X , where X, Y ∈ χ(M) and
f, g ∈ F(M).

Exercise 2.18 In terms of a local coordinate system (x 1 , x 2 , . . . , x n ) of a point in a


neighbourhood of a differential manifold M, show that
82 2 Manifold Theory
!
∂ ∂
(a) , = θ, i, j = 1, 2, 3, . . . , n.
∂x i ∂x j
 ∂η j ∂ξ j

∂ n n
i ∂ ∂
(b) [X, Y ] = ξi i − ηi i where X = ξ , Y = ηj j ,
i, j
∂x ∂x ∂x j
i=1
∂x i
j=1
∂x
each ξ i , η j ∈ F(M).
!
∂ ∂ ∂
Exercise 2.19 Compute y −x , .
∂x ∂ y ∂x
∂ ∂
Exercise 2.20 Let X = x + y . Compute X f where
∂x ∂y
(i) f : R3 → R is defined by f (x, y, z) = x 2 − y 2 − z 2
(ii) f : R2 → R is defined by f (x, y) = x y 7
(iii) f : R3 → R is defined by f (x, y, z) = e x cos y.

Exercise 2.21 (A). Compute [X, Y ]; (B). Compute [X, Y ](1,0) where
∂ ∂ ∂
(i) X = , Y = ex +
∂x ∂y ∂x
2 ∂ ∂
(ii) X = x ,Y = x
∂x ∂y
2 ∂ 2 ∂ ∂
(iii) X = x +y , Y = (y + 1) .
∂x ∂y ∂x
Exercise 2.22 (A). Compute [X, Y ]; (B). Compute [X, Y ](1,1,1) where
∂ ∂ ∂
(i) X = , Y = ex +
∂x ∂y ∂z
∂ ∂ ∂
(ii) X = y + x ,Y = y .
∂x ∂z ∂y

Exercise 2.23 Compute (A). ( f X )(1,1,1) and (B). (X f )(1,1,1) where f : R3 → R is


defined
∂ ∂ ∂ ∂
by f (x, y, z) = x 2 y 2 and (i) X = y + x , (ii) X = e x + and (iii) X =
∂x ∂z ∂y ∂z
∂ ∂
+ ez .
∂y ∂z

Answers

2.19. .
∂y
2.20. (i). 2(x 2 − y 2 ) (ii) 8x y 7 (iii) e x (x cos y) − y sin y.
∂ ∂ ∂ ∂
2.21. (A) (i) (ii) x 2 (iii) y 2 − 2x(y + 1) .
∂y ∂y ∂x ∂x
∂  ∂  ∂ 
(B) (i) (ii) (iii) −2 .
∂ y (1,0) ∂ y (1,0) ∂x (1,0)
2.11 Integral Curve 83

∂ ∂ ∂  ∂ 
2.22 (A) (i) (ii) −y (B)(i) (ii) − (1,1,1) .
∂y ∂x ∂y (1,1,1) ∂x
∂  ∂  ∂  ∂  ∂  ∂ 
2.23 (A) (i) + (ii) e (1,1,1) + (iii) +e  .
∂x (1,1,1) ∂z (1,1,1) ∂y ∂z (1,1,1) ∂ y (1,1,1) ∂z (1,1,1)
(B) (i) 2 (ii) 2e (iii) 2.

2.11 Integral Curve

We are going to state the geometrical interpretation of the vector field in this section.
In the last section, we have shown that a vector field is a rule that gives a tangent
vector at every point of the manifold M. Each point of M has its own tangent space.
The question now arises—for a given vector field, can we start from one point of
M and choose a curve whose tangent vector is always the given vector field? The
answer has been given in the affirmative sense.
At p ∈ U ⊂ M, suppose a vector field Y ∈ χ(M) is specified. A curve σ is an
integral curve of the vector field Y if the range of σ is contained in U and for
every a ≤ t0 ≤ b in the domain [a, b] of R of σ, the tangent vector to σ at σ(t0 ) = p
coincides with Y p , i.e.

Y p = Yσ(t0 )
i.e. Y p f = Yσ(t0 ) f, ∀ f ∈ F(M).

Using (2.15) and (2.22), we see that

   
n
 ∂  d  n
d x i (t)   ∂ 
ξ i ( p) f = ( f ◦ σ)(t)  = f.
∂x i p dt  dt t=t ∂x i p
i=1 t=t i=1 0 0

 ∂ 
Since { : i = 1, 2, 3, . . . , n} is a basis of Tp (M), we must have
∂x i p

d x i (t) 
ξ i ( p) =
dt t=t
  0
 d x i
(t) 
or ξ (σ(t)) =
i 
t=t0
dt t=t
  0
 d x i
(t) 
i.e. ξ (x (t), x (t), . . . , x (t)) =
i 1 2 n  , by (2.14).
t=t
dt t=t
0 0

Hence they are related by

d x i (t)
= ξ i (x 1 (t), x 2 (t), . . . , x n (t)). (2.29)
dt
84 2 Manifold Theory

A vector field X on M is said to be complete if at every point p of M, the integral


curve of X through p can be defined for all t ∈ R. Otherwise, it is said to be an
incomplete vector field.
An integral curve is said to be maximal if its domain cannot be extended to a
larger interval.
Remark 2.24 Note that the paths of different integral curves can never cross except
possibly at a point where ξ i = 0 for all t, because of the uniqueness of the solutions
of (2.29). Since some integral curve passes through each point p (it is found by
solving (2.29) with initial conditions at p), the integral curves “fill”M.
For instance, if M is a 3-dimensional manifold, then there exists a 2-dimensional
family of integral curves for each vector field on M and they cover all of M. Such a
manifold-filling set of integral curves is called congruence.
Remark 2.25 Let σ1 and σ2 be integral curves of a vector field X defined on open
intervals I1 and I2 respectively, containing 0. If σ1 (0) = σ2 (0), then σ1 = σ2 at each
point of I1 ∩ I2 .
Problem 2.42 Find the integral curve of the null vector field.
Solution: 42 For a given null vector field on Rn , the required differential equations
are
dx1 dx2 dxn
= = ··· = = 0, (2.30)
dt dt dt

∂ ∂ ∂
where θ = 0 + 0 2 + ··· + 0 n.
∂x 1 ∂x ∂x
If for initial condition t = 0, we have x 1 = p 1 , x 2 = p 2 , . . . , x n = p n , then we
get from (2.30) after integration

c1 = p 1 , c2 = p 2 , . . . , cn = p n ,

where c1 , c2 , . . . , cn are integrating constants.


Thus the integral curve, say σ, for the null vector field θ on Rn , is given by

σ = ( p 1 , p 2 , . . . , p n ), i.e. the point itself.

∂ ∂
Problem 2.43 Compute the integral curve of the vector field X = −y +x
∂x ∂y
on R2 , starting at the point (1, 0) ∈ R2 .
Solution: 43 The differential equations are

dx dy
= −y, = x.
dt dt
Thus ẋ = −y gives ẍ = − ẏ = −x from above. Hence x = A cos t + B sin t, where
A, B are to be determined. Therefore y = −ẋ gives y = A sin t − B cos t. It is given
2.11 Integral Curve 85

Fig. 2.18 Integral curves of


the vector field

X = −y ∂x + x ∂∂y

that, for t = 0, x = 1, y = 0. Hence we get from the above

A = 1, B = 0 i.e. x = cos t, y = sin t.

Thus the integral curve σ for the given vector field, starting at (1, 0), is σ =
(cos t, sin t), i.e. the curve is the unit circle.

Remark 2.26 In general, if for t = 0, x = p 1 and y = p 2 , then

p 1 = A, p 2 = B,

i.e.x(t) = p 1 cos t + p 2 sin t, y(t) = p 1 sin t − p 2 cos t. Hence, the integral curve
σ for the given vector field, starting from p = ( p 1 , p 2 ), is

σ(t) = ( p 1 cos t + p 2 sin t, p 1 sin t − p 2 cos t).

It is to be noted that σ(t) is defined for all t ∈ R and hence the given vector field X
is a complete vector field.
In this case,
x 2 (t) + y 2 (t) = ( p 1 )2 + ( p 2 )2 .

Thus the integral curves are circles with centre at the origin. The figure is given
(Fig. 2.18).

∂ x2 ∂
Problem 2.44 Let X = y ,Y = be two vector fields on R2 . Show that
∂x 2 ∂y
X, Y are complete but [X, Y ] is not.
86 2 Manifold Theory


Solution: 44 For X = y , the differential equations are
∂x
dx dy
= y, = 0.
dt dt
After integration, x = yt + A, y = B, where A, B are integrating constants. For
t = 0, if x = x0 , y = y0 , then A = x0 , B = y0 . Consequently the integral curve, say
σ(t) for X , through (x0 , y0 ) is

σ(t) = (y0 t + x0 , y0 ),

which is defined for all t ∈ R. Hence X is complete.


x2 ∂
Similarly, for Y = in χ(R2 ), the differential equations are
2 ∂y

dx dy x2
= 0, = .
dt dt 2

Consequently the integral curve, say σ̃(t) for Y , through (x0 , y0 ), is given by

 1 
σ̃(t) = x0 , x02 + y0 ,
2
which is defined for all t ∈ R. Hence Y is complete.
Now

" ∂ x2 ∂ # ∂  x2 ∂ f  x2 ∂  x2 ∂ f 
[X, Y ] f = y , f =y − y
∂x 2 ∂ y ∂x 2 ∂ y 2 ∂ y 2 ∂x
∂f yx 2 ∂ 2 f x2 ∂ f x 2 y ∂2 f
= yx + − −
∂y 2 ∂x∂ y 2 ∂x 2 ∂x∂ y
x2 ∂ ∂
i.e. [X, Y ] = − + xy .
2 ∂x ∂y

Thus, the differential equations are

dx x2 dy
=− , = x y. (2.31)
dt 2 dt
Integrating the foregoing equation, one finds

1 t
= + A, A being constant.
x 2
1 2x0
Thus, A = for t = 0, x = x0 and hence x = . From (2.31), one gets
x0 x0 t + 2
2.11 Integral Curve 87

dy 2x0
= dt.
y x0 t + 2

On integrating,
log y = 2 log(x0 t + 2) + log B,

where log B is the integrating constant. Therefore y = (x0 t + 2)2 . Hence y0 = 4B,
where y = y0 for t = 0. Consequently, the integral curve γ(t) for [X, Y ] is
 
2x0 y0
γ(t) = , (x t + 2) ,
2
x0 t + 2 4 0

2
which is not defined for t = − . Thus [X, Y ] is not complete.
x0
∂ ∂
Problem 2.45 Let X = y ,Y = x be two vector fields on R2 . Show that X, Y
∂x ∂y
are complete. Is [X, Y ] a complete vector field?

Solution: 45 For X = y , the differential equations are
∂x
dx dy
= y, = 0.
dt dt
After integration, x = yt + A, y = B, where A, B are integrating constants. For
t = 0, if x = x0 , y = y0 , then A = x0 , B = y0 . Consequently the integral curve, say
σ(t) for X , through (x0 , y0 ) is

σ(t) = (y0 t + x0 , y0 ),

which is defined for all t ∈ R. Hence X is complete.


x ∂
Similarly, for Y = in χ(R2 ), the differential equations are
2 ∂y

dx dy
= 0, = x.
dt dt

In a similar manner, we can show that the integral curve, say σ̂(t) for Y , through
(x0 , y0 ) is given by  
σ̂(t) = x0 , x0 t + y0 ,

which is defined for all t ∈ R. Hence Y is complete.


Again, after a brief calculation we obtain

∂ ∂
[X, Y ] = −x +y .
∂x ∂y
88 2 Manifold Theory

Hence, the differential equations are

dx dy
= −x, = y,
dt dt
which on solving, one gets

x = −x0 et , y = y0 et , where x = x0 , y = y0 for t = 0.

Thus, the integral curve γ(t) of [X, Y ] through (x0 , y0 ) is given by γ(t) = (−x0 et ,
y0 et ), which is defined for all t ∈ R. Thus [X, Y ] is a complete vector field.
∂ ∂
Problem 2.46 Find the integral curve for a given vector field X = x +y in
∂x ∂y
R2 . Is X complete? Give the geometrical interpretation of such X .
dx dy
Solution: 46 The differential equations are = x and = y. Integrating one
dt dt
gets log x = t + A and log y = t + B, A, B being integration constants. With ini-
tial condition, for t = 0, let x = p 1 , y = p 2 . Then

x = p 1 et , y = p 2 et .

Hence the integral curve, say σ, for X is given by

σ = ( p 1 et , p 2 et ),

x p1
which is defined for all t ∈ R. Thus X is complete. Also, = c, say where c = 2 .
y p
Therefore x = cy. This represents straight lines passing through the origin of R2 .

Problem 2.47 Let X be the vector field x 2 on the real line R. Find the integral
∂x
curve of X at 1. Is X complete?

Solution: 47 The differential equation is

dx
= x 2.
dt
Integrating, one gets

1
− = t + A, A being integration constant.
x
2.11 Integral Curve 89

1
When t = 0, then x = 1. Thus A = −1. Consequently, x = . Hence the integral
1−t
1
curve, say σ of X , is σ = which is not defined for t = 1. Thus X is not a
1−t
complete vector field.

Exercises
Exercise 2.24 Find the integral curve for the following vector fields. Also check
whether the given vector field is complete or not:

(a) X = e−x on R.
∂x
∂ ∂
(b) X = + (x 1 )2 2 on R2 .
∂x 1 ∂x
∂ ∂
(c) X = + ex on R3 .
∂y ∂z
∂ ∂
(d) X = y −x on R2 .
∂x ∂y
∂ ∂
(e) X = x 2 1 − (x 2 )3 2 on R2 .
∂x ∂x

(f) X = where X ∈ χ(R2 − {0}).
∂x
∂ ∂
Exercise 2.25 Compute the integral curve of the vector field X = + 2y +
∂x ∂y

3 on R3 passing through (x0 , y0 , z 0 ) at t = 0.
∂z
∂ ∂
Exercise 2.26 Compute the integral curve of X = +x on R2 passing through
∂x ∂y
(a, b) at t = 0.

Exercise 2.27 Let X be the vector field x on R. Find the integral curve of X
∂x
starting at p.
 
x+y ∂
Exercise 2.28 Find the integral curve of the vector field X = −
  r ∂y
y−x ∂
on R2 .
r ∂x

Answers
2.24. (a) log(t + e p ); No (b) ( p 1 + t, t ( p 1 + t)2 , p 2 ); Yes
(c) ( p , t + p , te + p ); Yes
1 2 1 2
(d) ( p 1 cos t + p 2 sin t, − p 1 sin t + p 2 cos t);
Yes  
t p2 p2
(e)  + p1 ,  ; No (f) (t + p 1 , p 2 ); No
1 − 2t ( p 2 )2 1 − 2t ( p 2 )2
2.25. (t + x0 , y0 e2t , 3t + z 0 ) 2.26. (t + a,  t (t + a), b) 2.27. pet
2.28. family of logarithmic spiral, where r = x + y . 2 2
90 2 Manifold Theory

2.12 Differential of a Mapping

Let M be an n-dimensional and N be an m-dimensional manifold and f : M → N


be a C ∞ map (Fig. 2.19).
Such f induces a map

f ∗ : F( f ( p)) → F( p), by

f ∗ (g) = g ◦ f, ∀ g ∈ F( f ( p)), p ∈ M (2.32)

and is called the pull-back of g by f . It satisfies



f ∗ (ag + bh) = a( f ∗ g) + b( f ∗ h)
(2.33)
f ∗ (gh) = f ∗ (g) f ∗ (h), ∀ h ∈ F( f ( p)), a, b ∈ R.

The map f also induces a mapping

f ∗ : Tp (M) → T f ( p) (N ), such that

{ f ∗ (X p )}g = X p ( f ∗ g) = X p (g ◦ f ) (2.34)

and is called the push-forward of X by f at p, denoted by f ∗, p . Such f ∗ is also


called the derived linear map or differential map of f on Tp (M). We write

f ∗ (X p ) = ( f ∗ X ) f ( p) . (2.35)

Fig. 2.19 Push-forward mapping


2.12 Differential of a Mapping 91

Remark 2.27 These notational conventions f ∗ , f ∗ as defined above reflect a similar


situation in linear algebra related to linear mappings of vector spaces and their duals,
respectively.

Problem 2.48 Prove that f ∗ is a linear map.

Solution: 48 Note that, for X p , Y p ∈ Tp (M) and λ ∈ R, we have

{ f ∗ (λX p + Y p )}g = (λX p + Y p )( f ∗ g) = (λX p + Y p )(g ◦ f ), by (2.34)


= λX p (g ◦ f ) + Y p (g ◦ f )
= λ{ f ∗ (X p )}g + { f ∗ (Y p )}g, by (2.34).

This proves f ∗ is a linear map.

Problem 2.49 Prove that f ∗ (X p ) is the derivation at f ( p).

Solution: 49 Note that, for all h, g, h + g ∈ F( f ( p)), we obtain

{ f ∗ (X p )}(h + g) = X p ( f ∗ (h + g)) = X p ((h + g) ◦ f )


= X p (h ◦ f ) + X p (g ◦ f )
= { f ∗ (X p )}(h) + { f ∗ (X p )}(g).

Also
{ f ∗ (X p )}(λh) = λX p (h ◦ f ) = λ{ f ∗ (X p )h}, λ ∈ R.

Thus f ∗ (X p ) is the derivation at f ( p).

Problem 2.50 If I is the identity map in the neighbourhood of a point p in a manifold


M, prove that (I∗ ) p is the identity map on Tp (M).

Solution: 50 Let I M denote the identity C ∞ map in the neighbourhood of a point p


of M. By (2.34), we obtain

{(I∗ ) p X p }g = X p (g ◦ I ) = X p g,

∴ (I∗ ) p X p = X p , ∀ g.

Thus (I∗ ) p is the identity differential of Tp (M).

Problem 2.51 If f is a smooth map from a manifold M into another manifold N


and g is a smooth map from N into another manifold L, then

(g ◦ f )∗ = g∗ ◦ f ∗ .

Solution: 51 Note that g ◦ f : M → L. Now f, g, g ◦ f induce the following linear


map:
92 2 Manifold Theory

f ∗ : T p (M) → T f ( p) (N ), g∗ : T f ( p) (N ) → Tg( f ( p)) (L), (g ◦ f )∗ : T p (M) → T(g◦ f )( p) (L).

Let h ∈ F(L). Then h ∈ F((g ◦ f )( p)). Now

((g ◦ f )∗ X p )h = X p (h ◦ (g ◦ f ))
= X p ((h ◦ g) ◦ f )
= ( f ∗ (X p ))(h ◦ g)
= {g∗ ( f ∗ (X p ))}h
∴ (g ◦ f )∗ X p = g∗ ( f ∗ (X p )), ∀ h
or (g ◦ f )∗ = g∗ ◦ f ∗ , ∀ X p ∈ Tp (M).

Problem 2.52 Let f : M → N be a diffeomorphism between two manifolds M and


N . Prove that
f ∗−1 (g X ) = (g ◦ f ) f ∗−1 X, ∀ g ∈ F(N ).

Solution: 52 Given that f : M → N is a diffeomorphism and hence by definition,


f −1 : N → M is C ∞ . Thus we can write

{ f ∗−1 (X f ( p) )}h = X f ( p) (h ◦ f −1 ), by (2.34), ∀ h ∈ F(M)


or ( f ∗−1 X ) p h = X f ( p) (h ◦ f −1 ), by (2.35).

Now for all X ∈ χ(N ), g X ∈ χ(N ), g ∈ F(N ) and hence replacing X by g X in the
above equation, we get

{ f ∗−1 (g X )} p h = (g X ) f ( p) (h ◦ f −1 )
= g( f ( p))X f ( p) (h ◦ f −1 ), by (2.26)
= (g ◦ f )( p)( f ∗−1 X ) p h, from above
or f ∗−1 (g X ) = (g ◦ f ) f ∗−1 X, ∀ h.

Problem 2.53 If f is a transformation of M and g is a differentiable function on


M, show that f ∗ (g X ) = (g ◦ f −1 ) f ∗ X .

Solution: 53 By virtue of (2.34),

{ f ∗ (g X ) p }h = (g X ) p (h ◦ f ), ∀ h ∈ F(M)
or { f ∗ (g X )} f ( p) h = g( p)X p (h ◦ f ), by (2.26), (2.35)
= g{( f −1 f )( p)}X p (h ◦ f ), as f is a transformation on M
= {(g ◦ f −1 ) f ( p)}{ f ∗ (X p )}h, by (2.34)
= (g ◦ f −1 ) f ( p)( f ∗ X ) f ( p) h
Thus f ∗ (g X ) = (g ◦ f −1 ) f X, ∀ h.
2.12 Differential of a Mapping 93

Problem 2.54 If f is a transformation of M and g is a differentiable function on


M, prove that f ∗ (( f ∗ X )g) = X ( f ∗ g).
Solution: 54 Note that f : M → M is a transformation and hence

f ( p) = q ⇒ p = f −1 (q), ∀ p, q ∈ M.

In view of (2.34), one gets

{ f ∗ (X p )}g = X p (g ◦ f ), ∀ g ∈ F(M)
or {( f ∗ X ) f ( p) }g = {X (g ◦ f )}( p), by (2.35)
or {( f ∗ X )g} f ( p) = {X (g ◦ f )}( p), by (2.23)
or {( f ∗ X )g}q = {X (g ◦ f )} f −1 (q)
or ( f ∗ X )g = {X (g ◦ f )} f −1 }, ∀ q
or {( f ∗ X )g} f = X (g ◦ f )
or f ∗ (( f ∗ X )g) = X ( f ∗ g), by (2.32).

Exercises
Exercise 2.29 If f is a smooth map from a manifold M into another manifold N
and g is a smooth map from N into another manifold L, then prove that (g ◦ f )∗ =
f ∗ ◦ g∗ .
Exercise 2.30 If f is a transformation of M and g is a differentiable function on
M, then f ∗ [X, Y ] = [ f ∗ X, f ∗ Y ].

Geometrical Interpretation of Differential Map


For X p ∈ Tp (M), we choose a curve σ(t) in M such that X p is the tangent vector to
the curve σ(t) at σ(t0 ) = p, a ≤ t0 ≤ b (Fig. 2.20).

Fig. 2.20 Geometrical interpretation of f ∗ (X p )


94 2 Manifold Theory

Then f ∗ (X p ) is defined to be the tangent vector to the curve f (σ(t)) at f ( p) =


f (σ(t0 )) and from (2.15), we have

d 
{ f ∗ (X p )}g = g( f (σ(t)))t=t , ∀ g ∈ F( f ( p))
dt 0

d 
= (g ◦ f )(σ(t))t=t
dt 0

= X p (g ◦ f ), by (2.15).

Theorem 2.6 If f is a mapping from an n-dimensional manifold M to an m-


dimensional manifold N where (x 1 , x 2 , . . . , x n ) is the local coordinate system in
a neighbourhood of a point p of M and (y 1 , y 2 , . . . , y n ) is the local coordinate
system in a neighbourhood of a point f ( p) of N , then

 ∂  m
∂ f j   ∂ 
f∗ = , where f j = y j ◦ f.
∂x i p j=1
∂x i p ∂ y j f ( p)

∂ 
Proof It is known that : i = 1, 2, 3, . . . , n is a basis of Tp (M) and in the
∂x i
∂ 
same manner : j = 1, 2, 3, . . . , m is a basis of T f ( p) (N ). Thus
∂y j

 ∂  m
 ∂ 
f∗ = aij , i = 1, 2, 3, . . . , n, (2.36)
∂x i p
j=1
∂y j

where aij ’s are to be determined. Therefore,

 ∂   k  j k
f∗ y = ai δ j = aik .
∂x i p j

By virtue of (2.34), we obtain

 ∂  k
(y ◦ f ) = aik
∂x i p
 ∂  k
or f = aik , by hypothesis
∂x i p
∂ f k 
or = aik .
∂x i p
Using in (2.36), the result follows immediately.

Corollary 2.1 Let (U, φ) be a chart about a point p in a manifold M. If (u 1 ,


u 2 , . . . , u n ) are the standard coordinates of Rn , then
2.12 Differential of a Mapping 95

 ∂   ∂ 
φ∗ = ,
∂x i p ∂u i φ( p)

where x i = u i ◦ φ, i = 1, 2, 3, . . . , n are the coordinates of p.

Proof Left to the reader.

Problem 2.55 Let f : M → N be a diffeomorphism between two smooth mani-


folds M and N . Then f ∗ : Tp (M) → T f ( p) (N ) is an isomorphism.

Solution: 55 Note that for f ∗ : Tp (M) → T f ( p) (N ), both Tp (M) and T f ( p) (N ) are


vector spaces over R. Hence we have to show
(i) f ∗ is a linear mapping and
(ii) f ∗−1 exists.
Now let f : M → N be a diffeomorphism and hence by definition f −1 exists and
is of class C ∞ . Now f ◦ f −1 = I M and f −1 ◦ f = I N . Again, in view of Problem
(2.51), we have
(I M )∗ = ( f ◦ f −1 )∗ = f ∗ ◦ f ∗−1 ,

where (I M )∗ is the identity differential on Tp (M). Furthermore,

(I N )∗ = ( f −1 ◦ f )∗ = f ∗−1 ◦ f ∗ ,

where (I N )∗ is the identity differential on T f ( p) (N ). Thus, f ∗−1 exists and in addition


to that it is of class C ∞ . Thus f ∗ is an isomorphism.

Remark 2.28 The matrix representation of f ∗ , denoted by ( f ∗ ), is given by


⎛ ⎞
∂ f1 ∂ f1
· · · ∂∂xf n
1

∂x 1 ∂x 2
⎜ ∂ f2 ∂ f2 2 ⎟
⎜ · · · ∂∂xf n ⎟
( f∗ ) = ⎜
⎜ ..
∂x 1 ∂x 2
.. .. .. ⎟ .
⎟ (2.37)
⎝ . . . . ⎠
∂ fm ∂ fm
· · · ∂∂xf n
m

∂x 1 ∂x 2

Problem 2.56 Find ( f ∗ ) where f : R2 → R2 is given by f =((x 1 )2 + (2x 2 )2 , 3x 1 x 2 ).


 1 2
2x 8x
Solution: 56 Here ( f ∗ ) = , where f 1 = (x 1 )2 + (2x 2 )2 , f 2 = (3x 1 x 2 ).
3x 2 3x 1
Exercise
Exercise 2.31 Find ( f ∗ ), where
(i) f : R2 → R2 is defined by f (x, y) = (xe y + y, xe y − y).
(ii) f : R → R is defined by f (x) = e x .
(iii) f : R2 → R3 is defined by f (x, y) = (x 2 y + y 2 , x − 2y 3 , ye x ).
(iv) f : R3 → R2 is defined by f (x, y, z) = (x 2 + y 2 + z 2 − 1, ax + by + cz).
96 2 Manifold Theory

(v) f : R4 → R2 is defined by f (x, y, z, t) = (x 2 + y 2 + z 2 + t 2 − 1, x 2 + y 2 +


z 2 + t 2 − 2y − 2z + 5).
(vi) f : R2 → R2 is defined by f (x, y) = (x 2 + y 2 , x 3 y 3 ).

Answers ⎛ ⎞
 y y  2x y x 2 + 2y  
e xe + 1 ⎝ 2 ⎠ 2x 2y 2z
(2.31)(i) (ii) (e ) (iii)
x
1 −6y (iv)
e y xe y − 1 a b c
ye x ex
   
2x 2y 2z 2t 2x 2y
(v) (vi) .
2x 2y − 2 2z − 2 2t 3x 2 y 3 3x 3 y 2
Problem 2.57 If f : R2 → R2 is given by (y 1 , y 2 ) = f (x 1 , x 2 ) = ((x 1 )2 + (2x 2 )2 ,
 ∂   ∂ 
3x 1 x 2 ), find f ∗ , f∗ .
∂x 1 ∂x 2
Solution: 57 From Theorem 2.6, we see that

 ∂   2
∂ f j  ∂
f∗ = , i = 1, 2.
∂x i
j=1
∂x i ∂ y j
 ∂  ∂ f 1  ∂ ∂ f 2  ∂
∴ f∗ = +
∂x 1 ∂x 1 ∂ y 1 ∂x 1 ∂ y 2
∂ ∂
= 2x 1 1 + 3x 2 2 and
∂y ∂y
 ∂  ∂ f 1  ∂ ∂ f 2  ∂
f∗ = +
∂x 2 ∂x 2 ∂ y 1 ∂x 2 ∂ y 2
∂ ∂
= 4x 2 1 + 3x 1 2 .
∂y ∂y

Alternative
Here $ %      1
 ∂  ∂ f1 ∂ f1
1 2x 1 4x 2 1 2x
f∗ = ∂x 1
∂ f2
∂x 2
∂ f2 = = .
∂x 1 0 3x 2 3x 1 0 3x 2
∂x 1 ∂x 2

 ∂  ∂ ∂ 
Since the vector f ∗ is the linear combination of the basis vectors , 2 ,
∂x 1 ∂y ∂y
1
we write from above
 ∂  ∂ ∂
f∗ = 2x 1 1 + 3x 2 2
∂x 1 ∂y ∂y
   1 2   2
 ∂  1 2x 4x 0 4x
Similarly, f ∗ = = =
∂x 2 0 3x 2 3x 1 1 3x 1
 ∂  ∂ ∂
∴ f∗ = 4x 2 1 + 3x 1 2 .
∂x 2 ∂y ∂y
2.12 Differential of a Mapping 97

Exercises
Exercise 2.32 A. Find ( f ∗ ) at (0, 0) in Exercise 2.31(i), (ii) and (iii).
B. Find ( f ∗ ) at (0, 0, 0) in Exercise 2.31(iv).
C. Find ( f ∗ ) at (0, 0, 0, 0) in Exercise 2.31(v).

Exercise 2.33 A. Let f : R2 → R2 be defined by


(i) f (x, y) = (x 2 − 2y, 4x 3 y 2 ). Find ( f ∗ )(1,2) .
(ii) f (x, y) = (x 2 + y 2 , x 3 y 3 ). Find ( f ∗ )(2,1) .
B. Let g : R2 → R3 be defined by g(x, y) = (x 2 y − y, 2x 3 − y, xe y ). Find (g∗ )(2,1) .
∂  ∂  ∂  ∂
Exercise 2.34 A. Find ( f ∗ ) , ( f∗ ) , ( f∗ ) , ( f∗ ) where
∂x ∂y ∂z ∂t
(i) f : R2 → R2 is defined by (u, v) = f (x, y) = (xe y + y, xe y − y).
(ii) f : R2 → R3 is defined by (u, v, w) = f (x, y) = (x 2 y + y 2 , x − 2y 3 , ye x ).
(iii) f : R3 → R2 is defined by (u, v) = f (x, y, z) = (x 2 + y 2 + z 2 − 1,
ax + by + cz).
(iv) f : R4 → R2 is defined by (u, v) = f (x, y, z, t) = (x 2 + y 2 + z 2 + t 2 − 1,
x 2 + y 2 + z 2 + t 2 − 2y − 2z + 5).
 ∂ 
B. Find ( f ∗ ) where f : R2 → R3 is defined by (u, v, w) = f (x, y)
∂x
= (x, y, x y).

Answers ⎛ ⎞
  00    
1 1 ⎝ ⎠ 000 0 0 0 0
2.32 A. (i) (ii) (1) (iii) 1 0 B. C. .
1 −1 abc 0 −2 −2 0
01
⎛ ⎞
    4 0
2 −2 4 1
2.33 A. (i) (ii) B. ⎝ 24 −1 ⎠.
48 16 12 24
e 2e
∂ ∂ ∂ ∂
2.34 A. (i) e y ∂u + e y ∂v , (xe y + 1) ∂u + (xe y − 1) ∂v .
∂ ∂ ∂ ∂ ∂ ∂
(ii) 2x y ∂u + ∂v + ye x ∂w , (x 2 + 2y) ∂u − 6y 2 ∂v + e x ∂w .
∂ ∂ ∂ ∂ ∂ ∂
(iii) 2x ∂u + a ∂v , 2y ∂u + b ∂v , 2z ∂u + c ∂v .
∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂
(iv) 2x ∂u + 2x ∂v , 2y ∂u + 2(y − 1) ∂v , 2z ∂u + 2(z − 1) ∂v , 2t ∂u + 2t ∂v .
∂ ∂ ∂ ∂
B. ∂u + y ∂w , ∂v + x ∂w .


Problem 2.58 Let f : R3 → R be defined by f (x, y, z) = x 2 y. If X = x y +
∂x

x2 ; compute f ∗ (X )(1,1,0) OR ( f ∗ )(1,1,0) (X )(1,1,0) .
∂z
Solution: 58 Taking into consideration (2.35), we know that

f ∗ (X )(1,1,0) = ( f ∗ X ) f (1,1,0) = ( f ∗ X )1 .

Now
98 2 Manifold Theory
⎛ ⎞
∂ f ∂ f ∂ f  xy
f ∗ (X )(1,1,0) = ⎝ 0 ⎠
∂x ∂ y ∂z (1,1,0) x 2
(1,1,0)
⎛ ⎞
1
= (2 1 0) ⎝ 0 ⎠
1
= 2.

d 
Thus, f ∗ (X )(1,1,0) = 2 , where t denotes the canonical coordinate of R.
dt 1

Exercises

Exercise 2.35 If X = x 2 , compute ( f ∗ )(1,1) (X )(1,1) for Exercise 2.31 (i), (iii).
∂y

Exercise 2.36 Let f : R2 → R2 be defined by f (x, y) = (x 2 + y, x 3 y 3 ) and g :


R2 →R3 be defined by g(x,
 y) = (x y − y, 2x − y, 
2 3
xe y ). Compute the following:
 ∂ ∂   ∂ ∂ 
(i) f ∗ 2 + (ii) g∗ 2 + .
∂x ∂ y (0,1) ∂x ∂ y (0,1)

Answers  
∂  ∂  ∂ ∂ ∂
2.35. (i) (e + 1) ∂u (e+1,e−1)
+ (e − 1) ∂v (e+1,e−1)
(ii) (3 ∂u − 6 ∂v + e ∂w )(2,−1,e) .
∂ ∂ ∂ ∂
2.36. (i) ( ∂x )(1,0) (ii) (− ∂x − ∂y
+ 2e ∂z )(−1.−1,0) .

Problem 2.59 Let f : R2 → R2 be defined by


  
cos θ − sin θ x1
(u, v) = f (x1 , x2 ) = .
sin θ cos θ x2

∂ ∂
Let X = −x2 + x1 be a vector field on R2 . If p = (x1 , x2 ) ∈ R2 and
∂x1 ∂x2

 ∂ ∂ 
f∗ X p = a +b ,
∂u ∂v f ( p)
find a, b.

Solution: 59 Here (u, v) = f (x1 , x2 ) = (x1 cos θ − x2 sin θ, x1 sin θ + x2 cos θ). Now
2.12 Differential of a Mapping 99

∂ ∂ 
{ f ∗ X p }u = a +b =a
∂u ∂v
or, X p (u ◦ f ) = a
 ∂ ∂ 
or, − x2 + x1 (x cos θ − x2 sin θ) = a
∂x1 ∂x2 1
or, − x2 cos θ − x1 sin θ = a.
 ∂ ∂ 
Similarly, − x2 + x1 (x sin θ + x2 cos θ) = b
∂x1 ∂x2 1
or, − x2 sin θ + x1 cos θ = b.

The values of a, b are therefore calculated.


Alternative    
cos θ − sin θ −x2
Here ( f ∗ ) = and (X ) p = . So
sin θ cos θ p=(x ,y)
x1 (x1 ,x2 )
1

 
cos θ − sin θ −x2
( f ∗ )(X ) p =
sin θ cos θ x1 f ( p)= f (x ,x )
  1 2

−x2 cos θ − x1 sin θ


=
−x2 sin θ + cos θ (x cos θ−x sin θ,x sin θ+x cos θ)
 1 2 1 2
&
∂ ∂
= (−x2 cos θ − x1 sin θ) + (−x2 sin θ + cos θ) .
∂u ∂u f ( p)

Note that, for the linear map f ∗ : Tp (M) → T f ( p) (N ) at the point p ∈ M, the Kernel
of f ∗ at p is given by

ker f ∗ = {X p ∈ Tp (M) f ∗ (X p ) = θ, θ ∈ T f ( p) (N )}.

Here ker f ∗ is a subspace of Tp (M). Also, the image of f ∗ at p is



Image f ∗ = {Y f ( p) ∈ T f ( p) (N ) f ∗ (X p ) = Y f ( p) },

which is a subspace of T f ( p) (N ).
Problem 2.60 Let f : R4 → R2 be defined by f (x1 , x2 , x3 , x4 ) = (u, v) = (x12 +
x22 + x32 + x42 − 1, x12 + x22 + x32 + x42 − 2x2 − 2x3 + 5).
(i) Find a basis of ker f ∗ at (0, 1, 2, 0).
(ii) Find the image by ( f ∗ ) of (1, 0, 2, 1) ∈ T(1,2,0,1) R4 .
Solution: 60 (i) As f : R4 → R2 is defined by f (x1 , x2 , x3 , x4 ) = (x12 + x22 +
x32 + x42 − 1, x12 + x22 + x32 + x42 − 2x2 − 2x3 + 5), then for p = (0, 1, 2, 0), f ∗ :
T(0,1,2,0) (R4 ) → T(4,4) (R2 ) is a differential map such that ker f ∗ at (0, 1, 2, 0) is a
vector subspace of T(0,1,2,0) (R4 ).
100 2 Manifold Theory

∂ ∂ ∂ ∂
Now any X ∈ χ(R4 ) can be expressed as X = a +b +c +d ,
∂x1 ∂x2 ∂x3 ∂x4
where a, b, c, d ∈ R. Here
 
2x1 2x2 2x3 2x4
( f∗ ) = .
2x1 2(x2 − 1) 2(x3 − 1) 2x4
 
0240
Therefore ( f ∗ )(0,1,2,0) = . Thus, ( f ∗ )(0,1,2,0) X ∈ T f ( p) (R2 ) is a linear
0020
∂ ∂ 
combination of , , where p = (0, 1, 2, 0), i.e. ( f ∗ )(0,1,2,0) X p = (2b +
∂u ∂v
∂ ∂
4c) + 2c . But ker f ∗ is such that ( f ∗ )(0,1,2,0) X p = θ, θ ∈ T(4,4) (R2 ), where
∂u ∂v
p = (0, 1, 2, 0). Consequently, we must have b = 0 = c. Thus

∂ ∂
Xp = a +d , p = (0, 1, 2, 0).
∂x1 ∂x2

Consequently,

ker( f ∗ )(0,1,2,0) = {X (0,1,2,0) ∈ T(0,1,2,0) (R4 )( f ∗ )(0,1,2,0) X (0,1,2,0) = θ},

 ∂   ∂  
and the basis is , such that ker( f ∗ )(0,1,2,0) is a subspace
∂x1 (0,1,2,0) ∂x2 (0,1,2,0)
of T(0,1,2,0) (R4 ).  
24 0 2
(ii) Also ( f ∗ )(1,2,0,1) = and f (1, 2, 0, 1) = (5, 7). Thus, ( f ∗ )(1,2,0,1) X p ∈
2 2 −2 2
T(5,7) (R ) can be expressed as
2

∂  ∂ 
( f ∗ )(1,2,0,1) X p = (2a + 4b + 2d) + (2a + 2b − 2c + 2d)  , p = (1, 2, 0, 1) and X = (1, 0, 2, 1).
∂u f ( p) ∂v f ( p)

∂ 
Therefore ( f ∗ )(1,2,0,1) X p = 4 .
∂u (5,7)
Problem 2.61 Let f : R2 → R3 be defined by f (x, y) = (x 2 y − y, 2x 3 − y, xe y ).
Calculate the conditions that the constants A, B, C must satisfy for the vector
 
∂ ∂ ∂
A +B +C
∂x ∂y ∂z f (0,0)

to be the image of some vector by f ∗ .


2.12 Differential of a Mapping 101
⎛ ⎞
0 −1
∂ ∂
Solution: 61 Note that ( f ∗ )(0,0) = ⎝ 0 −1 ⎠. Let us choose X = λ +μ ∈
1 0 ∂x ∂y
χ(R2 ) such that
⎛ ⎞ ⎛ ⎞
0 −1   −μ  ∂
λ ∂ ∂ 
( f ∗ )(0,0) X (0,0) = ⎝ 0 −1 ⎠ = ⎝ −μ ⎠ = A +B +C .
μ ∂x ∂y ∂z f (0,0)
1 0 λ f (0,0)

Compare A = −μ, B= − μ, C = λ. Note that ( f ∗ )(0,0) : T(0,0) (R2 ) → T(0,0,0) (R3 ).


Here Image f ∗ is a subspace of T(0,0,0) (R3 ) by definition. Thus, the image of f ∗ of
T(0,0) (R2 ) is a vector subspace of T(0,0,0) (R3 ) of vectors of type (−A, −A, C).

Exercises
Exercise 2.37 A. Let f : R4 → R2 be defined by

f (x1 , x2 , x3 , x4 ) = (u, v) = (x12 + x22 + x32 + x42 + 1, x12 + x22 + x32 + x 24 − 2x1 − 2x4 + 6).

(i) Find the basis of ker f ∗ at (1, 0, 1, 0).


(ii) Find the image by ( f ∗ ) of (1, 0, 1, 1) ∈ T(1,1,0,1) R2 .
B. Let g : R2 → R2 be defined by g(x1 , x2 ) = (x12 + x2 , x13 x23 ). Calculate the con-
ditions that the constants A, B must satisfy for the vector

 ∂ ∂ 
A +B
∂x1 ∂x2 g(0,0)

to be the image of some vector by g∗ .

Exercise
 2.38   us fix θ and define
Let f : R4 → R4 by f θ (x, y, z, t) =
cos θ − sin θ x z
.
sin θ cos θ yt
(i) Compute ( f θ )∗ .
∂ ∂ ∂ ∂
(ii) Compute ( f θ )∗ X , where X = cos θ − sin θ + cos θ − sin θ .
∂x ∂y ∂z ∂t

Answers 
∂ 
2.37 A. (i) {( ∂∂y )(1,0,1,0) , ( ∂w

)(1,0,1,0) } (ii) 4 ∂u B. (A, 0).
⎛ ⎞ (4,7)

cos θ − sin θ 0 0
⎜ sin θ cos θ 0 0 ⎟
2.38 (i) ⎜
⎝ 0
⎟ (ii) ∂ + ∂ .
0 cos θ sin θ ⎠ ∂x ∂z

0 0 sin θ cos θ
A smooth map f : M → N is said to be a smooth submersion (or simply sub-
mersion) at p ∈ M, if f ∗ at p, i.e. f ∗, p is surjective. Equivalently, we can say that

dim Image f ∗, p = dim T f ( p) (N ) = m.


102 2 Manifold Theory

Moreover, f is called a smooth immersion (or simply immersion) at p ∈ M, if its


differential f ∗, p is injective. In other words, we have

ker f ∗, p = {θ} =⇒ dim ker f ∗, p = 0.

Hence, using the well-known theorem in linear algebra, viz.

dim ker f ∗, p + dim Image f ∗, p = dim M = n,

we find dim Image f ∗, p = n. Thus, if a smooth map f : M n → N m is a submersion


at a point p in M, then n ≥ m and if immersion at a point p in M then n ≤ m. A
simple question arises: Is an injective or surjective map f , respectively, an immersion
or submersion? For this, we need to explain immersion and submersion using the
rank of the differential map f ∗ at the point p in M.
We know that the dimension of Image f ∗, p is said to be the rank of f ∗ at p,
denoted by rank f ∗, p . We define the rank of f at p, denoted by rank f | p , to be of r
if rank f ∗ at p = r . In other words, we can also say that the rank f ∗ is the rank of
the Jacobian matrix of f with respect to any smooth chart. If f has the same rank r
at every point of M, we say that it has constant rank r , and write rank f = r .
Since rank f ∗, p ≤ min{dim Tp (M), dim T f ( p) (N )} = min{n, m}, therefore

rank f | p ≤ min{n, m}.

A smooth map f : M → N is said to be a submersion at p ∈ M, if

rank f ∗, p = dim Image f ∗, p = dim T f ( p) (N ) = m i.e. rank f | p = m.

Moreover, f is called an immersion at p ∈ M, if

rank f | p = dim M = n.

If we consider a smooth map f : R2 → R3 , then its differential at p ∈ R2 is given


by
f ∗, p : Tp (R2 ) → T f ( p) (R3 ).

Here
rank f ∗, p ≤ min{2, 3} = 2 ⇒ rank f ∗, p = 1 or 2.

Here, f fails to be submersion at p, for if f is submersion at p then rank f ∗ = 3


which is not possible. Thus we can say that if

dim M = n < dim N = m,

then f fails to be a submersion at any point p in M. However, if rank f ∗ = 2, then


rank f | p = 2 implies f is immersion at p in R2 . In a similar manner for a smooth map
2.12 Differential of a Mapping 103

f : R3 → R2 , to be immersion at p ∈ R3 , we have rank f ∗ = 3, i.e. rank f | p = 3


which is not possible. Thus if

dim M = n > dim N = m,

then f fails to be an immersion at any point p in M.


A smooth map f : M → N is said to be a submersion and immersion if its dif-
ferential is respectively surjective and injective at every point of M.
Remark 2.29 Let f : M → N be a smooth map of constant rank. Then
(i) if f is surjective, it is a submersion.
(ii) if f is injective, it is an immersion.
(iii) if f is bijective, it is a diffeomorphism.

Examples
Example 2.9 Let us consider the curve γ : (−π, π) → R2 defined by γ(t) =

(sin 2t, sin t). Here, γ is injective but γ (t) does not vanish for any t. Hence γ is
an injective immersion.
Example 2.10 Let us consider the function f : R → R2 by t → (t 2 − 1, t (t 2 − 1)).

Here f is not injective as f (1) = f (−1) = (0, 0). But f does not vanish for any t,
so f is an immersion but not injective.
Example 2.11 Suppose M1 , M2 , . . . , Ms are the smooth manifolds. Then each of
the projection maps
πi : M1 × M2 × · · · × Ms → Mi

is a submersion. In particular, if π is a projection map from Rn+k to Rn by

(x 1 , x 2 , . . . , x m , x m+1 , . . . , x n ) → (x 1 , x 2 , . . . , x m ),

then π is a submersion.
Example 2.12 If U is an open subset of a manifold M, then the inclusion map
i : U → M is both an immersion and submersion. Moreover, here the map is not
surjective. So this example shows that a submersion need not be surjective.
Example 2.13 Let us consider the map f : R → R, f (x) = x 3 . Here f is surjective

but at x = 0, d f = f (x) = 3x 2 is not surjective. Hence f fails to be a submersion.
So this example shows that a surjective map need not be a submersion.
Let f : M → N be a smooth map. A point p ∈ M is said to be a critical point
of f if f ∗, p is not surjective. A point q ∈ N is said to be a critical value of f if the
set f −1 (q) contains a critical point of f . In other words, a point in N is a critical
value if it is the image of some critical point in M.
In particular, let f : M → R be a smooth map on M. A point p ∈ M is said to
be a critical point of f if f ∗, p = 0.
104 2 Manifold Theory

Proposition 2.8 Let f : N n → R be a C ∞ map. A point p ∈ N is a critical point


if and only if it is relative to a coordinate system (U, x 1 , x 2 , x 3 , . . . , x n ) at p ∈ N ,
∂ f 
= 0.
∂x i p
Proof For f : N n → R, the map f ∗ : Tp (N ) → T f ( p) (R) is such that
 
∂f ∂f ∂f
( f∗ ) p = . . . .
∂x 1 ∂x 2 ∂x n p

It is known from the previous discussion that Image f ∗ is a subspace of T f ( p) (R).


Thus it is either 0-dimensional or 1-dimensional. Hence either f ∗ is a zero-map
or a surjective map.
∂ f 
Thus, f ∗ will not be surjective if and only if = 0.
∂x i p
It means that the real number f ( p) is called a critical value of f . A critical point
is called non-degenerate if

 ∂2 f 
det ( p) = 0.
∂x ∂x
i j

Remark 2.30 Non-degeneracy is independent of the choice of coordinate system.


A point p ∈ M is a regular point of f : M → N if f ∗ , p is surjective. In other
words, we can say that p ∈ M is a regular point of f if and only if f is a submersion
at p, i.e. rank f ∗ = dim M.

Example 2.14 The function f (x) = x + e−x has a critical point at c = 0. The
derivative is zero at this point. So
 
f (x) = (x + e−x ) = 1 − e−x .

Now f (c) = 1 − e−c = 0 ⇒ c = 0.

Example 2.15 The function f (x) = 2x − x 2 has a critical point at c = 1. The


 
derivative is zero at this point. Here f (x) = 2 − 2x. So, f (c) = 0 ⇒ c = 1.

Problem 2.62 Find the critical points of the map f : R3 → R2 given by

(x, y, z) → (x z, y).

Solution: 62 Let p ∈ R3 . The Jacobian matrix f ∗, p is given by


 ∂   
∂x
(x z) ∂∂y (x z) ∂z

(x z) z0x
f ∗, p = ∂ ∂ ∂ = .
∂x
(y) ∂x (y) ∂x (y) 010
2.13 Submanifolds 105

Here f ∗, p fails to be surjective if rank f ∗ < 2 if and only if x = z = 0. Hence the set
of critical points of f is the y-axis.

f :M→N

Injective Surjective Injective Neither Injective


Not Surjective Not Injective And Surjective Nor Surjective

f∗ f∗ f∗ f∗ f∗ f∗ f∗ f∗
Injective Surjective Injective Surjective Injective Surjective Injective Surjective

Immersion Submersion Immersion Submersion Immersion Submersion Immersion Submersion

2.13 Submanifolds

Let us consider a map f : M → N , where M and N are topological manifolds. f


is said to be a topological embedding (topological imbedding) if f is a homeo-
morphism onto its image f (M)(⊂ N ), where f (M) is endowed with the subspace
topology inherited from N .
The smooth map f : M → N , where M and N are smooth manifolds, is a smooth
embedding (or smooth imbedding) if f is an immersion together with topological
embedding.
Let N be a smooth manifold and M ⊆ N . Let M be a manifold endowed with the
subspace topology inherited from N . Then M is said to be an embedded submanifold
(or regular submanifold) of N if M is endowed with a smooth structure with respect
to which the inclusion map of M in N is a smooth embedding.
Let N be an m-dimensional smooth manifold and M ⊂ N . Let M be a manifold
equipped with a topology, not necessarily the subspace topology inherited from N ,
with respect to which it is a topological manifold of dimension n. Then M is said to
be an immersed submanifold of N if M is endowed with a smooth structure with
respect to which the topological manifold M becomes an n-dimensional smooth
manifold such that the inclusion map i : M → N is a smooth immersion. Also, the
immersed submanifold M has co-dimension m − n.
It is evident that every embedded submanifold is an immersed submanifold (refer
to Problem 2.63). For the sake of simplicity, embedded submanifold and immersed
submanifold are always of the smooth kind.
Example 2.16 If M1 , M2 , . . . , Ms are the smooth manifolds and qi ∈ Mi are arbi-
trary points, each of the maps

ζi : Mi → M1 × M2 × · · · × Ms ,

given by ζi ( p) = (q1 , q2 , . . . , q j−1 , p, q j+1 , . . . , qs ), is a smooth embedding.


106 2 Manifold Theory

Example 2.17 The map γ : R → R2 given by γ(t) = (t 5 , 0) is a smooth map but



is not a topological embedding because γ (0) = 0.

Example 2.18 Let f : R → R2 be defined by f (t) = (cos πt, sin πt), ∀ t ∈ R.


Here f is an immersion but not an injective map. So it fails to be a topological
embedding.

Example 2.19 The circle S 1 is a 1-dimensional embedded submanifold of R2 .


1
Example 2.20 The graph y = (x) 3 is an embedded submanifold of R2 .

Example 2.21 The sphere S n is an embedded submanifold of Rn with dimension


n − 1.

Example 2.22 Let f : Rm → Rn be a smooth map. Then its graph



γ = {(x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ) ∈ Rm+n  f (x1 , x2 , . . . , xm ) = (y1 , y2 , . . . , yn )}

is a smooth m-dimensional embedded submanifold of Rm+n .

Lemma 2.1 Let X, Y be topological spaces. Let F : X → Y be 1 − 1, continuous


and open map. Then F is a topological embedding.

Proof Let U be an open subset of X . Since F is open, F(U ) is open in Y . Hence


F(U ) ∩ F(X ) is open in F(X ). Since F(U ) ⊂ F(X ), therefore F(U ) is open in
F(X ). This proves F is a topological embedding.

Lemma 2.2 Let X, Y be topological spaces. Let F : X → Y be 1 − 1, continuous


and closed map. Then F is a topological embedding.

Proof Left to the reader.

Lemma 2.3 Closed map lemma: Let X be a compact space and Y be a Haus-
dorff space. Let F : X → Y be a 1 − 1, continuous map. Then F is a topological
embedding.

Proof Let A be any closed subset of X . Since A is closed in the compact space X ,
A is compact in X . Since F is continuous, therefore F(A) is compact in Y , and Y
being Hausdorff, therefore F(A) is closed in Y . This shows F is a closed map. Since
F : X → Y is a 1 − 1, continuous map, therefore F is a topological embedding (refer
to Lemma 2.2).

Now we are going to define a proper map between two topological spaces as
follows.
Let X, Y be topological spaces. Let F : X → Y be a mapping. If for every compact
subset W of Y , the inverse image F −1 (W ) is compact in X , then we say that F :
X → Y is a proper map.
2.13 Submanifolds 107

Lemma 2.4 Let X be a Hausdorff topological space and Y be a Hausdorff locally


compact space. Let F : X → Y be a continuous map. Let F be a proper mapping.
Then F is a closed mapping.

Proof Left to the reader.

Lemma 2.5 Let M be an n-dimensional smooth manifold and N be an


m-dimensional smooth manifold. Let F : M → N be a 1 − 1 smooth immersion.
If F is a proper mapping, then F : M → N is a smooth embedding.

Proof Since M is an n-dimensional smooth manifold, M is a Hausdorff locally


compact space. Similarly, N is a Hausdorff locally compact space. Since F is a
smooth map, it is continuous. Further, since F is a proper map, F is a closed map
(refer to Lemma 2.4). Since F is a 1 − 1 smooth immersion, and F is closed, therefore
F is a smooth embedding.

Lemma 2.6 Let M be an n-dimensional compact smooth manifold and N be an


m-dimensional smooth manifold. If F : M → N is a 1 − 1 smooth immersion, then
F : M → N is proper mapping.

Proof Let W be any compact subset of N . Our claim is to show that the inverse
image F −1 (W ) is compact in M. Since W is compact in N and N is Hausdorff,
therefore W is closed in N . Since F is smooth, it is continuous. Hence F −1 (W ) is
closed in M. As M is compact, F −1 (W ) is compact in M. This proves F is proper,
hence a smooth embedding (refer to Lemma 2.5).

Problem 2.63 Let M be an n-dimensional smooth manifold and S(= φ) ⊂ M. Let


S be an embedded submanifold of M with co-dimension k, k = 1, 2, 3, …, n − 1.
Prove that S is an immersed submanifold of M with co-dimension k.

Solution: 63 Let us set τS = {G ∩ S : G is open in M}. Since S is an embedded


submanifold of M with co-dimension k, τS is a topology over S (called the subspace
topology of S) with respect to which S becomes a (n − k)-dimensional topological
manifold. Moreover, ∃ an C ∞ -atlas A on S, with respect to which the topological
manifold S becomes an (n − k)-dimensional smooth manifold such that the map
i : S → M is a smooth embedding, and hence i is a smooth immersion. Thus, S is
an immersed submanifold of M with co-dimension n − (n − k) = k. This completes
the solution.

Problem 2.64 Let M be an n-dimensional smooth manifold with C ∞ -atlas A and


N be an m-dimensional smooth manifold with C ∞ -atlas B. The map F : M → N
is a 1 − 1 smooth immersion. Then F(M) is a smooth submanifold of N with co-
dimension m − n.

Solution: 64 Let us set τ = {F(U ) : U is open in M}. It is obvious that τ is a


topology over F(M), and F is a homeomorphism from M onto F(M). Since M
is an n-dimensional topological manifold and M is homeomorphic onto F(M),
108 2 Manifold Theory

therefore
 F(M) is an n-dimensional topological manifold. Moreover, {(F(U ), φ ◦
(F −1  F(U ) )) : (U, φ) ∈ A} forms an C ∞ -atlas on F(M). Hence, F(M) forms a
smooth manifold, and F forms a diffeomorphism from M onto F(M).
Now we want to prove that the map i : F(M) → N is a smooth map. Here i =
F ◦ F −1 . Since F : M → F(M) is a diffeomorphism, F −1 : F(M) → M is also so.
Furthermore, as F is a 1 − 1 smooth immersion, the composition map i = F ◦ F −1
is a smooth immersion. It follows that F(M) is a smooth submanifold of N with
co-dimension m − n.

Problem 2.65 Let M be an n-dimensional smooth manifold and S(= φ) ⊂ M. Let


S be a smooth submanifold of M with co-dimension 0. Prove that S is an embedded
submanifold of M with co-dimension 0.

Solution: 65 Since S is a smooth submanifold of M with co-dimension 0, by its


definition, there exists a topology τ over S with respect to which S becomes an n-
dimensional topological manifold, and there exists a C ∞ structure on S with respect
to which the topological manifold S becomes an n-dimensional smooth manifold
such that the map i : S → M is a 1 − 1 smooth immersion. Thus i is a smooth
embedding, hence a topological embedding. Thus i is a homeomorphism from S
onto i(S)(= S), where i(S) has the subspace topology inherited from M. It follows
that τ is the subspace topology of S inherited from M. Thus combining all the facts,
we conclude that S is an embedded submanifold of M with co-dimension 0.

Problem 2.66 Let M be an n-dimensional smooth manifold and S(= φ) ⊂ M. Let


S be a smooth submanifold of M with co-dimension k. If the map i : S → M is
proper, then prove that S is an embedded submanifold of M with co-dimension k.

Solution: 66 Since S is a smooth submanifold of M with co-dimension k, by virtue


of its definition, ∃ a topology τ over S with respect to which S becomes an (n − k)-
dimensional topological manifold, and ∃ a smooth structure on S with respect to
which the topological manifold S becomes an (n − k)-dimensional smooth manifold
with the inclusion map i : S → M which is a 1 − 1 smooth immersion. Since the
map i is proper, it follows that i is a smooth embedding (refer to Lemma 2.5), and
hence a topological embedding. So i is a homeomorphism from S onto i(S)(= S),
where i(S) has the subspace topology inherited from M. It follows that τ is the
subspace topology of S inherited from M. Thus combining all the facts, we conclude
that S is an embedded submanifold of M with co-dimension k.

Problem 2.67 Let M be an n-dimensional smooth manifold and S(= φ) ⊂ M. Let


S be a compact smooth submanifold of M with co-dimension k. Prove that S is an
embedded submanifold of M with co-dimension k.

Solution: 67 Since S is a smooth submanifold of M with co-dimension k, by virtue


of its definition, ∃ a topology τ over S with respect to which S becomes an (n − k)-
dimensional topological manifold, and ∃ a smooth structure on S with respect to
which the topological manifold S becomes an (n − k)-dimensional smooth manifold
2.13 Submanifolds 109

with the map i : S → M which is a 1 − 1 smooth immersion. Further, since S is


compact, it follows that i is a smooth embedding (refer to Lemma 2.5), and hence
a topological embedding. So i is a homeomorphism from S onto i(S)(= S), where
i(S) has the subspace topology inherited from M. It follows that τ is the subspace
topology of S inherited from M. Thus combining all the facts, we conclude that S is
an embedded submanifold of M with co-dimension k.

Remark 2.31 Let M be an n-dimensional smooth manifold and N be an m-


dimensional smooth manifold. The map F : M → N is a smooth map. Let S(=
φ) ⊂ M be a smooth submanifold of M. Then the restricted map F|S : S → M is
also smooth.

Problem 2.68 Let M be an n-dimensional smooth manifold and N be an m-


dimensional smooth manifold with S(= φ) ⊂ N . Let S be a smooth submanifold
of N with co-dimension k. Let F : M → N be a smooth map, and F(M) ⊂ S. Let
F : M → S be continuous. Then prove that F : M → S is a smooth map.

Solution: 68 Let us fix any p ∈ M. Then F( p) ∈ F(M) ⊂ S ⇒ F( p) ∈ S. Since


S is a smooth submanifold of N with co-dimension k, by virtue of its definition,
∃ a topology τ over S with respect to which S becomes an (n − k)-dimensional
topological manifold, and ∃ a C ∞ structure on S with respect to which the topological
manifold S becomes an (n − k)-dimensional smooth manifold with the map i : S →
N which is a 1 − 1 smooth immersion. Taking advantage of Exercise 2.14, we can
say that i is a local diffeomorphism. Hence, ∃ an open neighbourhood V of F( p) ∈ S
such that i(V ) is open in N , and the map i|V : V → i(V ) is a diffeomorphism, and
hence the map (i|V )−1 : i(V ) → V is smooth. Since V is an open neighbourhood
of F( p) ∈ S, and F is continuous, ∃ an open neighbourhood U of p ∈ M such that
F(U ) ⊂ V . Since U is open in M, U is an open submanifold of M, hence U is
an embedded submanifold of M. Thus U is a smooth submanifold of M. Further,
since F is a smooth map, by Remark 2.31, F|U : U → N is a smooth map. Thus,
the composite map (i|V )−1 ◦ F|U = F|U : U → V is smooth. Consequently, as U
is an open neighbourhood of p ∈ M, F : M → S is smooth at p, it follows that
F : M → S is a smooth map.

i : M(⊆ N ) → N
(Inclusion Map)

i ∗ Injective i ∗ Injective
+ +
M endowed with subspace topology inherited from N M endowed with any topology other than subspace topology

Embedded Submanifold Immersed Submanifold


110 2 Manifold Theory

Exercise
Exercise 2.39 Let M be an n-dimensional smooth manifold, N be an m-dimensional
smooth manifold and S(= φ) ⊂ N . Let S be an embedded submanifold of N with
co-dimension k. Let F : M → N be a smooth map, and F(M) ⊂ S. Let F : M → S
be continuous. Then prove that F : M → S is smooth.

2.14 f -Related Vector Fields

Let f : M → N be a smooth map. For each p ∈ M, let X p ∈ Tp (M) and X f ( p) ∈


T f ( p) (N ) be such that
f ∗ (X p ) = Y f ( p) . (2.38)

In such a case, we say that X, Y are f -related vector fields. Now

{ f ∗ (X p )}g = Y f ( p) g, ∀ g ∈ F( f ( p)).

Using (2.23) and (2.34), we find

X p (g ◦ f ) =(Y g)( f ( p))


or {X (g ◦ f )}( p) =(Y g) f ( p) by (2.23)
X (g ◦ f ) =(Y g) f. (2.39)

If f is a transformation on M and

f ∗ (X p ) = X f ( p)
i.e. ( f ∗ X ) f ( p) = X f ( p) ,

we say that X is f -related to itself or X is invariant under f . Thus

f ∗ X = X. (2.40)

Proposition 2.9 Let f : M → N be a smooth map. If the vector fields X 1 , X 2 on


M are f -related to the vector fields Y1 , Y2 respectively on N , then the Lie bracket
[X 1 , X 2 ] is f -related to the Lie bracket [Y1 , Y2 ].

Proof Given that X 1 (g ◦ f ) = (Y1 g) f and X 2 (g ◦ f ) = (Y2 g) f . Now


2.14 f -Related Vector Fields 111

[X 1 , X 2 ](g ◦ f ) = X 1 {X 2 (g ◦ f )} − X 2 {X 1 (g ◦ f )}, f ∈ F(M), g ∈ F(N )


= X 1 {(Y2 g) f } − X 2 {(Y1 g) f }, from above
= {Y1 (Y2 g)} f − {Y2 (Y1 g)} f, Y2 g ∈ F(N ), Y1 g ∈ F(N )
= {Y1 (Y2 g) − Y2 (Y1 g)} f = {[Y1 , Y2 ]g} f.

Thus by (2.39), we can claim that [X 1 , X 2 ] is f -related to [Y1 , Y2 ].


Problem 2.69 Let f : R2 → R4 be a differential map, and f (x1 , x2 ) = (u, v, ω, t)

be such that u = x12 − x22 , v = x12 + x22 , ω = x1 + x2 , t = x1 − x2 . Let X = x1 +
∂x1
∂ ∂ ∂
x2 and Y = −x2 + x1 be two vector fields on R2 . Find vector fields on
∂x2 ∂x1 ∂x2
R4 , f -related to X, Y respectively.
Solution: 69 By definition f ∗ : Tp (R2 ) → T f ( p) (R4 ) and if f ∗ X = X̄ and f ∗ Y = Ȳ ,
then X̄ and Ȳ are respectively the f -related vector fields of X, Y . Let

∂ ∂ ∂ ∂
X̄ = a +b +c +t , (2.41)
∂u ∂v ∂ω ∂t

where a, b, c, d are from F(R4 ), to be determined. Now

{ f ∗ X }u = X̄ u = a.

Again by (2.34), the left-hand side of the foregoing equation reduces to

X (u ◦ f ) = a.

Applying the hypothesis,

∂ ∂
a = (x1 + x2 )(x12 − x22 ) = u(2x1 ) − x2 (2x2 ) = 2(x12 − x22 ).
∂x1 ∂x2
∂ ∂
Similarly, b = (x1 + x2 )(x12 + x22 ) = x1 (2x1 ) + x2 (2x2 ) = 2(x12 + x22 ).
∂x1 ∂x2
∂ ∂ ∂ ∂
c = (x1 + x2 )(ω ◦ f ) = (x1 + x2 )(x1 + x2 ) = x1 · 1 + x2 · 1 = x1 + x2 .
∂x1 ∂x2 ∂x1 ∂x2
∂ ∂ ∂ ∂
d = (x1 + x2 )(t ◦ f ) = (x1 + x2 )(x1 − x2 ) = x1 − x2 .
∂x1 ∂x2 ∂x1 ∂x2

Consequently, from (2.41) we write

∂ ∂ ∂ ∂
X̄ = 2(x12 − x22 ) + 2(x12 + x22 ) + (x1 + x2 ) + (x1 − x2 )
∂u ∂v ∂u ∂t
∂ ∂ ∂ ∂
i.e. X̄ = 2u + 2v +ω +t
∂u ∂v ∂ω ∂t

is f -related vector field of X , in R4 . Again, we write


112 2 Manifold Theory

∂ ∂ ∂ ∂
Ȳ = a  + b + c + d , (2.42)
∂u ∂v ∂ω ∂t

where a  , b , c , d  are from F(R4 ), to be determined. In a similar manner, it can be


shown that
∂ ∂ ∂
Ȳ = (t 2 − ω 2 ) +t −ω
∂u ∂ω ∂t

is f -related vector field of Y , in R4 .


Problem 2.70 Let f : M → N be a C ∞ map. Prove that two vector fields X, Y
respectively on M, N are f -related if and only if

f ∗ (( f ∗ X )g) = X ( f ∗ g), g ∈ F(N ).

Solution: 70 Given that X, Y are f -related, hence by (2.38) we obtain

f ∗ (X p ) = Y f ( p)
or ( f ∗ X ) f ( p) = Y f ( p) , by (2.35)
or ( f ∗ X ) f ( p) g = Y f ( p) g, g ∈ F(N )
i.e. {( f ∗ X )g} f ( p) = (Y g) f ( p), by (2.23)
i.e. ( f ∗ X )g = Y g. (2.43)

In view of (2.39), we have

X (g ◦ f ) = (Y g) ◦ f
or X ( f ∗ g) = f ∗ (Y g). by (2.32)

Using (2.43) above, on the right-hand side, we get

X ( f ∗ g) = f ∗ (( f ∗ X )g).

The converse follows immediately.


Problem 2.71 Let f : M → N be a C ∞ map. Let X, Y be two f -related vector
fields. If σ is the integral curve of X , prove that f ◦ σ is the integral curve of Y .
Solution: 71 If σ is the integral curve of X , then by definition, we have

X p = X σ(t0 ) ,

where σ : [a, b] ⊂ R → M is such that σ(t0 ) = p, p ∈ M, t ∈ [a, b] and X p is


tangent vector to the curve σ at p, i.e.

d 
Xp f = f (σ(t))t=t .
dt 0
2.14 f -Related Vector Fields 113

Again by virtue of (2.34), one gets

d
{ f ∗ (X p )}g = X p (g ◦ f ) = (g ◦ f )(σ(t)), g ∈ F(N ).
dt
Using (2.38) on the left-hand side of the last equation, we get

d 
{Y f ( p) }g = g(( f ◦ σ)(t))t=t .
dt 0

Thus, we can say that Y f ( p) is tangent vector to the curve f ◦ σ at f ( p) = ( f ◦ σ)(t0 ).


Thus, f ◦ σ is the integral curve of the vector field Y at f ( p), i.e.

Y f ( p) = Y f (σ(t0 )) ,

where X, Y are f -related vector fields satisfying (2.38).

Problem 2.72 Let the projection map π : R2 → R be defined by π(x, y) = x. Find


the condition that a vector field of R2 is π-related to some vector field of R.
∂ ∂
Solution: 72 Let X ∈ χ(R2 ) be such that X = ξ +η where ξ : R2 → R, η :
∂x ∂y
R2 → R are c∞ functions.
Let X be π-related to vector field Y ∈ χ(R) and hence by (2.38), we can write

π∗ X = Y. (2.44)

d
Let us write Y = θ , where t denotes the canonical coordinates on R. Again
dt
π(x, y) = x, so we have

 ∂π1 ∂π1 
(π∗ ) = = (1 0) and
∂x ∂ y
 
ξ
(π∗ )(X ) = (1 0) = ξ.
η

If (2.44) holds, then we must have ξ = θ and this is the required condition.

Exercises

Exercise 2.40 If f is a transformation on M, prove that for every X ∈ χ(M), there


exists a unique f -related vector field of X .

Exercise 2.41 Let f : R2 → R3 be a differential map f (x1 , x2 ) = (u, v, ω) be


∂ ∂ ∂
such that u = x1 x2 , v = x2 + 1, ω = x1 + 1. Let X = x12 + x2 , Y = x1
∂x1 ∂x2 ∂x1
be two vector fields on R2 . Find vector fields on R3 , f -related to X, Y respectively.
114 2 Manifold Theory

Answer
∂ ∂ ∂ ∂ ∂
2.41. uw + (v − 1) + (ω − 1)2 ; u + (ω − 1) .
∂u ∂v ∂ω ∂u ∂ω

2.15 One Parameter Group of Transformations on a


Manifold

In this section, we wish to interpret the algebraic interpretation of the vector field.
Let a mapping φ : R × M → M be such that

⎨ (i) for each t ∈ R, φ(t, p) → φt ( p) is a transformation on M;
(ii) for all t, s, t + s in R, (2.45)

φt (φs ( p)) = φt+s ( p).

Then the family {φt |t ∈ R} of mappings is called a one-parameter group of transfor-


mations on M.
Problem 2.73 Let {φt |t ∈ R} be one-parameter group of mappings of M. Show that
(i) φ0 is the identity mapping.
(ii) φ−t = (φt )−1 .

Solution: 73 (i) Taking t = 0 ∈ R in (2.45) (ii), one gets φ0 (φs ( p)) = φs ( p). Thus
φ0 is the identity mapping.
(ii) For every t, −t ∈ R

φt (φ−t ( p)) = φ0 ( p), by (2.45)(ii)


= p, by (2.45)(i)
or φ−t ( p) = (φt )−1 ( p).
∴ φ−t = (φt )−1 , ∀ p ∈ M.

Exercise
Exercise 2.42 Prove that {φt |t ∈ R} forms an Abelian group.

Remark 2.32 Exercise 2.42 gives the algebraic interpretation of the vector field X
on a manifold.

Let us set
ψ(t) = φt ( p). (2.46)

Then ψ : R → M is a differentiable curve on M such that ψ(0) = φ0 ( p) = p (refer


to Problem 2.73). Such a curve is called the orbit through p of M. The tangent
vector, say X p , to the curve ψ(t) at p is therefore
2.15 One Parameter Group of Transformations on a Manifold 115

d f (φt ( p)) − f (φ0 ( p)))


Xp f = f (ψ(t))|t=0 = lim , ∀ f ∈ F(M). (2.47)
dt t→0 t

In this case, we say that {φt |t ∈ R} induces the vector field X and X is called the
generator of φt . The curve ψ(t) defined by (2.46) is called the integral curve of X .
Problem 2.74 Show that the mapping φ : R × R3 → R3 defined by

φ(t, p) = ( p 1 + t, p 2 + t, p 3 + t)

is a one-parameter group of transformations on R3 and the generator is given by

∂ ∂ ∂
+ 2 + 3 where p = (x 1 , x 2 , x 3 ) ∈ R3 .
∂x 1 ∂x ∂x

Solution: 74 Clearly, φ is a transformation on R3 and

φt (φs ( p)) = φt ( p 1 + s, p 2 + s, p 3 + s); as defined


= φt+s ( p).

Thus, {φt |t ∈ R} is a one-parameter group of transformations on R3 . Again,

φt ( p) − φ0 ( p)
X p = lim = (1, 1, 1).
t→0 t

∂ ∂ ∂ 
Now , 2 , 3 is a basis of Tp (R3 ) and hence X p ∈ Tp (R3 ) is given by
∂x ∂x ∂x
1

∂ ∂ ∂
+ 2+ 3
∂x 1 ∂x ∂x

which is the generator of {φt }.


Problem 2.75 Let φ : R × R2 → R2 be defined by φt ( p) = (x cos t − y sin t,
x sin t + y cos t).
(i)Show that {φt |t ∈ R} defines a one-parameter group of transformations on R2 .
(ii)Find its generator.
(iii)Describe the orbit.
(iv) Prove that X is invariant under φt , i.e. (φt )∗ X p = X φt ( p) .
 
 cos t − sin t 
Solution: 75 (i) Note that |J | =   = 1 = 0. Hence, φ−1 exists and
sin t cos t 
using Problem 2.73 (ii), we have
       
φ−1
t
(x , y ) = φ−t (x , y ) = (x cos t + y sin t, −x sin t + y cos t).

It can be shown that


116 2 Manifold Theory

   
φt (φ−t (x , y )) = (x , y ) and φ−t (φt (x, y)) = (x, y),

and hence we claim that {φt } is a transformation of R2 . Finally

φt (φs (x, y)) = (x cos(t + s) − y sin(t + s), x sin(t + s) + y cos(t + s))


= φt+s (x, y).

Thus, {φt |t ∈ R} is a one-parameter group of transformations on R2 .


(ii) From the definition,

d  
Xp = φt ( p)t=0 = (−x sin t − y cos t, x cos t − y sin t)t=0 = (−y, x).
dt

∂ ∂
Thus, the generator is given by −y +x .
∂x ∂y
(iii) The orbit through p = (x0 , y0 ) while t = 0 is the image of the map R → R2
given by
t → (x0 cos t − y0 sin t, x0 sin t + y0 cos t).

∂ ∂
(iv) Again, X = −y +x (refer to (ii) above). Now
∂x ∂y

φt ( p) = φt (x0 , y0 ) = (x0 cos t − y0 sin t, x0 sin t + y0 cos t).

Thus
 
∂ ∂
X φt ( p) = − y +x
∂x ∂ y (x cos t−y0 sin t,x0 sin t+y0 cos t)
0

∂ ∂
= (−x0 sin t − y0 cos t) + (x0 cos t − y0 sin t) .
∂x ∂y

Again
  
cos t − sin t −y 
(φt )∗ X p =
sin t cos t x p
= (−y cos t − x sin t, −y sin t + x cos t)(x0 ,y0 )
= (−y0 cos t − x0 sin t, −y0 sin t + x0 cos t).

Thus
∂ ∂
(φt )∗ X p = (−x0 sin t − y0 cos t) + (x0 cos t − y0 sin t) = X φt ( p) .
∂x ∂y

Thus X is invariant under φt .


2.15 One Parameter Group of Transformations on a Manifold 117

Problem 2.76 Let φ : R × R2 → R2 be defined by φt (x, y) = (x, yet ).


(i)Show that {φt |t ∈ R} defines a one-parameter group of transformations on R2 .
(ii)Find its generator.
(iii)Describe the orbit.
(iv) Prove that X is invariant under φt .
 
1 0 
Solution: (i) Note that |J | =  t  = et = 0, ∀ t. Hence φ−1 exists and
0e
     
φ−1
t
(x , y ) = φ−t (x , y ) = (x , y /et ).

It can be shown that


   
φt (φ−t (x , y )) = (x , y ) and φ−t (φt (x, y)) = (x, y),

and hence we claim that {φt } is a transformation of R2 . Finally

φt (φs (x, y)) = (x, yet+s )


= φt+s (x, y).

Thus, {φt |t ∈ R} is a one-parameter group of transformations on R2 .


(ii) From the definition,

d  
Xp = φt ( p)t=0 = (0, yet )t=0 = (0, y).
dt


Thus the generator is given by y .
∂y
(iii) The orbit through p = (x0 , y0 ) while t = 0 is the image of the map R → R2
given by
t → (x0 , y0 et ).

 ∂   t ∂
(iv) Now φt ( p)=φt (x0 , y0 )=(x0 , y0 et ). Hence X φ(t) ( p) = y (x ,y t ) = y0 e .
∂y 0 0 e ∂y
Again
   
1 0 0 
(φt )∗ X p =
0 et y  (x0 ,y0 )

= (0, yet )(x0 ,y0 )


  ∂
= y0 et = X (φt )( p) .
∂y

Thus X is invariant under φt .


118 2 Manifold Theory

Exercises
Exercise 2.43 Show that the following families of maps φ : R × R2 → R2 form a
one-parameter group of transformations and find their generators.
(i) φt ( p) = (x + at, y + bt), a, b ∈ R
(ii) φt ( p) = (xe2t , ye−2t )
where p = (x, y).

Exercise 2.44 Let φ : R × R2 → R2 be defined by

φt ( p) = (x cos t + y sin t, −x sin t + y cos t), p = (x, y).

Show that {φt |t ∈ R} defines a one-parameter group of transformations on R2 .


(i) Find its generator X .
(ii) Describe the orbit.
(iii) Prove that X is invariant under φt , i.e. (φt )∗ X p = X φt ( p) .
 
1t
Exercise 2.45 Let M = G L(2, R) and a mapping φt (A) = · A,
01
A ∈ G L(2, R) with the dot denoting matrix multiplication. Find the generator.

Answers
∂ ∂ ∂ ∂
2.43 (i) a +b (ii) 2x − 2y .
∂x ∂y ∂x ∂y
∂ ∂
2.44 (i) y −x (ii) circle centred at the origin.
∂x ∂y
∂ ∂
2.45 a21 + a22 .
∂x ∂y

Problem 2.77 Let M = R3 and a mapping φ : R × M → M be such that X =
∂x
is its generator. Find φ.

Solution: 76 For X , the differential equations are

dx dy dz
=1 =0= , where (x, y, z) ∈ R3 .
dt dt dt
On solving, we get
x = t + A, y = B, z = C

where A, B, C are integrating constants. If for t = 0, x = x0 , y = y0 , z = z 0 , then


A = x0 , B = y0 , C = z 0 . Consequently, the integral curve is given by ψ(t) = (x0 +
t, y0 , z 0 ) and φt : R × R3 → R3 is defined by φt (x, y, z) = (x + t, y, z) where
φ0 (x, y, z) = ψ(0) = p = (x0 , y0 , z 0 ).
In this case
2.15 One Parameter Group of Transformations on a Manifold 119

φt ( p) = φt (x0 , y0 , z 0 ) = (x0 + t, y0 , z 0 ),

and hence
 ∂  ∂
X φt ( p) = (x +t,y ,z )
= (x0 + t)
∂x 0 0 0 ∂x
and
⎛ ⎞⎛ ⎞
100 1 

(φt )∗ X p = 0 1 0 ⎠ ⎝ 0 ⎠ = (1, 0, 0)(x0 +t,y0 ,z0 ) = (x0 + t, 0, 0)
001 0 p

= (x0 + t, 0, 0) = X (φt )( p) .
∂x

∂ ∂
Exercise 2.46 Let M = R2 , the x y-plane and X = y − x . Find the domain
∂x ∂y
W and the one-parameter group φ : W → M.
Exercise 2.47 Let M = R2 and a mapping φ : R × M → M be such that
∂ ∂
(i) X = x +y is its generator;
∂x ∂y
∂ ∂
(ii) X = −y +x is its generator;
∂x ∂y
∂ ∂
(iii) X = +y is its generator;
∂x ∂y
Find φ in each case.

Answers
2.46 (yt + a, −xt + b).
(2.47)(i) φt (x, y) = (xet , yet ) (ii) φt (x, y) = (x cos t − y sin t, x sin t + y cos t).
(iii) φt (x, y) = (t + x, yet ).
Since every one-parameter group of transformations generates a vector field, the
question now arises whether every vector field induces a one-parameter group of
transformations or not. The question has been answered in negative.
∂ ∂
Example 2.23 Let X = −e x + be defined on R2 . As done earlier, it can be
∂x ∂y  
1
shown that the integral curve ψ(t) of X is ψ(t) = log , t + p 2
, not
(t + e− p1 )
defined ∀t ∈ R, where x(0) = p 1 , y(0) = p 2 .
Consequently, by (2.46), if we define ψ(t) = φt ( p) then, X does not induce one-
parameter group of transformations on R2 .
The above observation leads to the following definition:
Local one-parameter group of transformations: Let I be an open interval (−, )
on R and U be a neighbourhood of a point p of M (Fig. 2.21).
120 2 Manifold Theory

Fig. 2.21 Local


one-parameter group of
transformations

Let us define a mapping φ : I × U → φt (U ) ⊂ M by φ(t, p) → φt ( p) be such


that
(i) U is an open cover of M.
(ii) for each t ∈ I , φt ( p) is a transformation of U onto an open subset φt (U ) of
M.
(iii) if t, s, t + s are in I and if φs ( p) ∈ U then

φt (φs ( p)) = φt+s ( p).

Such a family {φt |t ∈ I } of mappings is called a local one-parameter group of


transformations defined on I × U .
Now, we are going to prove the following theorem.
Theorem 2.7 Let X be a vector field on a manifold M. Then X generates local
one-parameter group of transformations in a neighbourhood of a point in M.

Proof Let (U, φ) be a chart of p of M. By Exercise 2.8 we can write

φ( p) = (0, 0, 0, . . . , 0) ∈ Rn .

If (x 1 , x 2 , . . . , x n ) is the local coordinate system of p, then x i ( p) = 0, i = 1, 2,


3, . . . , n (Fig. 2.22).

Fig. 2.22 Existence theorem of local one-parameter group of transformations


2.15 One Parameter Group of Transformations on a Manifold 121


Let X = ξ i be a given vector field on U of p with each ξ i : U ⊂ Rn → R, i =
∂x i
1, 2, 3, . . . , n being differentiable. Then φX , the φ-related vector field on Rn , is
defined in a neighbourhood U1 = φ(U ) at φ( p) = (0, 0, 0, 0 . . . , 0) ∈ Rn . We write

φX = η i i where each η i : φ(U ) ⊂ Rn → R is differentiable. Then by virtue of
∂x
the existence theorem of the ordinary differential equation, for each φ( p) ∈ U1 ⊂ Rn ,
there exists δ1 > 0 and a neighbourhood V1 of φ( p), V1 ⊂ U1 such that, for each
q = (q 1 , q 2 , . . . , q n ) ∈ V1 , φ(r ) = q, say, r ∈ U ⊂ M, there exists n-tuple of C ∞
functions f 1 (t, q), f 2 (t, q), . . . , f n (t, q) defined on Iδ1 ⊂ I1 ,

f i : Iδ1 → V1 ⊂ U1 ⊂ Rn , i = 1, 2, 3, . . . , n

which satisfies the system of first-order differential equations

d i
f (t) = η i (t, φ( p)), i = 1, 2, 3, . . . , n (2.48)
dt
with the initial condition
f i (0, q) = f i (0) = q i . (2.49)

Let us write
θt (q) = ( f 1 (t, q), f 2 (t, q), . . . , f n (t, q)). (2.50)

Then θ : Iδ1 × V1 → θt (q) ∈ V1 is a transformation of V1 onto an open set θt (V1 ) of


Rn .
Let us set

(g 1 (t), g 2 (t), . . . , g n (t)) = ( f 1 (t + s, q), f 2 (t + s, q), . . . , f n (t + s, q)),

where each f i (t + s, q), f i (t, s(q)) are defined on Iδ1 × V1 if θs (q) ∈ V1 ⊂ U1 and
t, s, t + s are in Iδ1 . Componentwise, we write

(g i (t)) = ( f i (t + s, q)),

where each g i (t) is defined on Iδ1 × V1 , V1 ⊂ U1 with initial condition

(g i (0)) = ( f i (s, q)). (2.51)

Similarly, if we write
(h i (t)) = ( f i (t, θs (q))),

then each h i (t) is defined on Iδ1 × V1 , V1 ⊂ U1 and hence satisfies (2.48) with initial
condition
122 2 Manifold Theory

(h i (0)) = ( f i (0, θs (q))) = ( f i (0)) = (θs (q))i = ( f i (s, q)), by (2.49), (2.50)

as constructed. Thus (h i (0)) = (g i (0)) (by (2.51)). Hence, from uniqueness we have

(h i (t)) = (g i (t)) i.e. ( f i (t, θs (q))) = ( f i (t + s, q)),

which can be written as θt (θs (q)) = θt+s (q) (refer to (2.50)). We can also write it
as θ(t, θ(s, q)) = θ(t + s, q). Thus, {θt |t ∈ Iδ1 } is the local one-parameter group of
transformations induced by the vector field φX at U1 of φ( p) of Rn .
Let us now set φ−1 (V1 ) = V ⊂ U of p of M and define

ψ : I × V → ψt (V ) ⊂ M,

as ψt (r ) = φ−1 (θt (q)) with q = φ(r ), i.e.

ψt (r ) = φ−1 (θ(t, φ(r )). (2.52)

Then
(i) V is an open cover of M
(ii) for each t ∈ I , ψ(t, p) → ψt ( p) is a transformation of V onto an open set
ψt (V ) of M and
(iii) if t, s, t + s are in I and if ψt (r ) ⊂ ψt (V ), then

ψt (ψs (r )) = φ−1 (θ(t, φ(ψs (r )))), by (2.52)


= φ−1 (θ(t, φφ−1 (θ(s, φ(r ))), by (2.52)
= φ−1 (θ(t, θ(s, q)))
= φ−1 (θ(t + s, q)), as {θt } is the local 1-parameter group of transformations
= ψt+s (r ), by (2.52),

i.e. {ψt |t ∈ I } is the local one-parameter group of transformations defined on I ×


V, V ⊂ U ⊂ M for the vector field X defined in the neighbourhood U of a point of
M.
Finally, if we write

γ(t) = ψt (r ) = φ−1 (θt (q)), q = φ(r )


= φ−1 (σ(t)), say

then φ−1 (σ(t)) is the integral curve of X , where σ(t) is the integral curve of the
vector field φX of Rn . This completes the proof.

Problem 2.78 Prove that the integral curve always gives rise to a vector field, but
the converse is not true.
2.15 One Parameter Group of Transformations on a Manifold 123

Solution: 77 Let {φt |t ∈ R} be one-parameter group of transformations on a mani-


fold M. From (2.46), we see that if ψ : R → M is a differentiable curve on M such
that
ψ(t) = φt ( p), ψ(o) = φ0 ( p) = p,

d  d 
then X p = (ψ(t))t=0 = (φt ( p))t=0 is called the generator of {φt |t ∈ R} and
dt dt
the curve ψ(t) is the integral curve of X .
Thus, every one-parameter group of transformations or the integral curve on a
manifold induces a vector field on a manifold.
Conversely, by Example 2.23, the vector field on a manifold does not in general
induce an integral curve on a manifold.

Problem 2.79 Let φ be a transformation on M. If a vector field X generates {φt |t ∈


I } as its local one-parameter group of transformations, prove that the vector field φ∗ X
will generate {φφt φ−1 |t ∈ I } as its local one-parameter group of transformations on
M.

Solution: 78 Let X be given vector field on a manifold M. Then by Theorem 2.7,


X generates {φt |t ∈ I } as its local one-parameter group of transformations on M.
Let ψ(t) = φt ( p), then X p is the tangent vector to the curve ψ(t) at ψ(0) =
φ0 ( p) = p (refer to Problem (2.73)(i)). Thus

d  d 
Xp = ψ(t)t=0 = (φt ( p))t=0 .
dt dt

Now by definition, φ∗ : Tp (M) → Tφ( p) (M) and φ∗ (X p ) is defined to be the tangent


vector to the curve φ(ψ(t)) at φ(ψ(0)) = φ( p), i.e.

φ(ψ(t)) = φ(φt ( p))


= φ(φt (φ−1 (q))), say where φ( p) = q, φ being a transformation on M
= (φφt φ−1 )(φ( p)), and
d 
φ∗ (X p ) = φ(ψ(t))t=0
dt
d 
or (φ∗ X )φ( p) = (φφt φ−1 )(φ( p))t=0 , from above and by (2.35).
dt

Comparing with (2.47), we can now say that the vector field φ∗ X generates
{φφt φ−1 |t ∈ I } as its local one-parameter group of transformations on M.

Exercise
Exercise 2.48 Show that a vector field X on a manifold M is invariant under a
transformation φ on M if and only if φ ◦ φt = φt ◦ φ where {φt |t ∈ I } is the local
one-parameter group of transformations on M, generated by X .
124 2 Manifold Theory

Now, we are going to give the geometrical interpretation of the Lie Bracket [X, Y ]
for every vector field X, Y on M.

Theorem 2.8 If X generates {φt |t ∈ I } as its local one-parameter group of trans-


formations, then for every vector field Y on M

1
[X, Y ] = lim {Y − (φt )∗ Y }.
t→0 t

We also write
1
[X, Y ]q = lim {Yq − ((φt )∗ Y )q }, ∀ q ∈ M, q = φt ( p), p ∈ M.
t→0 t

To prove the theorem, we require a few lemmas.


Lemma 2.7 If ψ(t, p) is a function on I × M, I = (−, ) on R such that ψ(0, p)
= 0, ∀ p ∈ M, then there exists a function h(t, p) on I × M such that th(t, p) =
ψ(t, p). Moreover
  dψ
h(0, p) = ψ (0, p), ψ = .
dt
' 1
 d(ts)
Proof Let us define h(t, p) = ψ (ts, p) . In view of the fundamental the-
0 t
orem of calculus, we have

1" #1
h(t, p) = ψ(ts, p) 0 =⇒ t h(t, p) = ψ(t, p).
t
' 1
 
Also, h(0, p) = ψ (0, p)ds = ψ (0, p).
0

Lemma 2.8 If f is a function on M and X is a vector field on M which induces a


local one-parameter group of transformations {φt |t ∈ I }, then there exists a function
gt defined on I × V, V being the neighbourhood of a point p of M where gt ( p) =
g(t, p) such that f (φt ( p)) = f ( p) + t gt ( p). Moreover, X p f = g(0, p) = g◦ ( p).
Symbolically, X f = g◦ .

Proof Let us set

f˜(t, p) = f (φt ( p)) − f (φ◦ ( p)), ∀ p ∈ M.

Then f˜(t, p) is a function on I × M such that f˜(0, p) = 0, ∀ p ∈ M. Hence by


Lemma 2.7, there exists a function, say, g(t, p) on I × V, V ⊂ M such that

t g(t, p) = f˜(t, p)
or t gt ( p) = f (φt ( p)) − f ( p).
2.15 One Parameter Group of Transformations on a Manifold 125

Hence, the result follows. Also from above

1
g(0, p) = lim { f (φt ( p)) − f (φ◦ ( p))} = X p f.
t→0 t

Proof of the Main theorem:


Let us write φt ( p) = q. Therefore, p = φ−t (q). Then by (2.34), we get

{((φt )∗ Y ) f }(q) = {Y ( f ◦ φt )}( p) = (Y f + t Y gt )( p), by Lemma 2.8.

Therefore,

(Y f )(q) − {((φt )∗ Y ) f }(q) = (Y f )(q) − (Y f )( p) − t (Y gt )φ−t (q)


1 (Y f )(q) − (Y f )( p)
or, lim {Yq − ((φt )∗ Y )q } f = lim − lim (Y gt )φ−t (q)
t→0 t t→0 t t→0
(Y f )(q) − (Y f )( p)
= lim − Yq (X f ), by Lemma 2.8.
t→0 t

Now from (2.47),

1
X q f = lim { f (φt (q)) − f (q)}
t→0 t
1
∴ −X q f = lim { f ( p) − f (q)}, as p = φ−t (q).
t→0 t

Replacing f by Y f , one obtains

1
X q (Y f ) = lim {(Y f )(q) − (Y f )( p)}.
t→0 t

Thus, we write

1
lim {Yq − ((φt )∗ Y )q } f = X q (Y f ) − Yq (X f )
t→0t
= {X (Y f ) − Y (X f )}(q)
= {[X, Y ] f }(q)
= [X, Y ]q f.

1
Therefore, [X, Y ]q = lim {Yq − ((φt )∗ Y )q }, ∀ f . The result follows immediately.
t→0 t

1
Corollary 2.2 Show that (φs )∗ [X, Y ] = lim {(φs )∗ Y − (φs+t )∗ Y }.
t→0 t

Proof In view of Theorem 2.8, one gets


126 2 Manifold Theory

1
(φs )∗ [X, Y ] = lim (φs )∗ {Y − (φt )∗ Y }
t→0 t
1
= lim {(φs )∗ Y − (φs )∗ (φt )∗ Y )}, as (φs )∗ is linear
t→0 t
1
= lim {(φs )∗ Y − (φs ◦ φt )∗ Y }, by Problem 2.51
t→0 t
1
= lim {(φs )∗ Y − (φs+t )∗ Y }, by (2.45)(ii).
t→0 t


d(φt )∗ Y 
Corollary 2.3 Show that (φs )∗ [X, Y ] = − .
dt t=s

Proof Note that


 
d  (φt+h )∗ Y − (φt )∗ Y 

((φt )∗ Y ) = lim
dt h→0 h 
t=s t=s

(φs+h )∗ Y − (φs )∗ Y
= lim
h→0 h
= −(φs )∗ [X, Y ], using Corollary 2.2.

Corollary 2.4 Let X, Y generate {φt } and {ψs } respectively as its local one-
parameter group of transformations. Then φt ◦ ψs = ψs ◦ φt if and only if [X, Y ] = 0.

Proof Let φt ◦ ψs = ψs ◦ φt . Then from Exercise 2.48, we can say that the vector
field Y is invariant under φt . Consequently, by (2.40), we find (φt )∗ Y = Y . Hence,
taking advantage of Theorem 2.8, we find [X, Y ] = 0.
Conversely, let [X, Y ] = 0. Then in view of the foregoing corollary, we have

d
((φt )∗ Y ) = 0
dt
i.e. (φt )∗ Y = Constant, ∀ t
i.e. (φt )∗ Y = (φ0 )∗ Y = Y.

Finally, taking into consideration Exercise 2.48, we must have the desired result.

A vector field X on a manifold is said to be complete if it generates a one-parameter


group of transformations on M.
Theorem 2.9 Every vector field on a compact manifold is complete.

Proof Left to the reader.


2.15 One Parameter Group of Transformations on a Manifold 127

Hint
Theorem 2.9: Use Theorem 2.7 and then use the compactness property.
Remark 2.33 If φ is a transformation on a compact manifold and the vector field
X is complete, then φ∗ X is also so.
Chapter 3
Differential Forms

3.1 Cotangent Space

A mapping ω : χ (M) → F(M) that satisfies



ω(X + Y ) = ω(X ) + ω(Y )
(3.1)
ω(bX ) = bω(X ), ∀ X, Y ∈ χ (M), b ∈ F(M)

is called a linear mapping over R, where χ (M), F(M) are vector spaces over R.
A linear mapping ω : χ (M) → F(M) denoted by X → ω(X ) is also called a
1-form on M. 
Let D1 (M) = {ω, μ, . . . , . . .  ω : χ (M) → F(M)} be the set of all 1-forms on
M. Let us define 
(ω + μ)(X ) = ω(X ) + μ(X )
(3.2)
ω(bX ) = bω(X )

It can be shown that D1 (M) is a vector space over R, called the dual of χ (M). We
write
{ω(X )}( p) = ω p (X p ), where ω(X ) ∈ F(M). (3.3)

Thus ω p : Tp (M) → R and hence ω p ∈ dual of Tp (M).


We denote the dual of Tp (M) by Tp∗ (M) and is called the cotangent space of M
at p ∈ M. Elements of Tp∗ (M) are also called the co-vectors at p ∈ M.
For every f ∈ F(M), we denote the total differential of f by d f and is defined
as
(d f ) p (X p ) = (X f )( p) = X p f, ∀ p ∈ M. (3.4)

We also write it as
(d f )(X ) = X f. (3.5)

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 129
M. Majumdar and A. Bhattacharyya, An Introduction to Smooth Manifolds,
https://doi.org/10.1007/978-981-99-0565-2_3
130 3 Differential Forms

Problem 3.1 Show that for every f ∈ F(M), d f is a 1-form on M.

Solution: For every X, Y ∈ χ (M), X + Y ∈ χ (M) and

d f (X + Y ) = (X + Y ) f, by (3.5)
= d f (X ) + d f (Y ), by (3.5).
Also d f (bX ) = (bX ) f, by (3.5)
= b(X f )
= bd f (X ), by (3.5)

Thus d f is a 1-form.

Exercise

Exercise 3.1 If (x 1 , x 2 , . . . , x n ) are co-ordinate functions defined in a neighbour-


hood of p of M, show that each d x i , i = 1, 2, 3, 4, . . . , n is a 1-form on M.

From Exercise 3.1, we claim that each d x i ∈ Tp∗ (M), i = 1, 2, 3, . . . , n. We


define
  
∂ 1, i = j
(d x ) p
i
= δj =
i
(3.6)
∂x j p 0, i = j

Let ω p ∈ Tp∗ (M) be such that


 

ωp = ( f j ) p , where each ( f j ) p ∈ R. (3.7)
∂x j p

If possible, let μ p ∈ Tp∗ (M) be such that

μ p = ( f 1 ) p (d x 1 ) p + ( f 2 ) p (d x 2 ) p + · · · + ( f n ) p (d x n ) p .
 

Then, μ p = ( f i ) p , i = 1, 2, . . . , n by (3.6)
∂xi p
 

= ωp , , i = 1, 2, . . . , n see (3.7).
∂xi p
 

and hence μ p = ω p as : i = 1, 2, 3, 4, . . . , n is a basis of Tp (M). Thus any
∂xi

ω p ∈ Tp (M) can be expressed uniquely as
3.1 Cotangent Space 131


n 
ωp = ( f i ) p (d x i ) p i.e. ω = fi d x i (3.8)
i=1

and hence Tp∗ (M) = span{(d x 1 ) p , (d x 2 ) p , . . . , (d x n ) p }. Finally, if ( f i ) p (d x i ) p = 0


 

then ( f i ) p (d x ) p
i
= 0 yields ( f k ) p = 0 [refer to (3.6)].
∂xk p
Similarly, it can be shown that ( f 1 ) p = ( f 2 ) p = ( f 3 ) p = · · · = ( f n ) p = 0 and the
set {(d x 1 ) p , (d x 2 ) p , . . . , (d x n ) p } is linearly independent. We state
Theorem 3.1 If (x 1 , x 2 , . . . , x n ) are local co-ordinate system in a neighbourhood
U of p of M, then the set {(d x 1 ) p , (d x 2 ) p , . . . , (d x n ) p } is a basis of Tp∗ (M) or
D1 (M n ).

Remark 3.1 A zero-form is nothing but a function, by convention.

Remark 3.2 We say that the form ω in (3.8) is differentiable if each f i is of


class C ∞ .
  ∂ 
Remark 3.3 From ω = fi d x i , X = ξ i j , we see that ω(X ) = fi ξ i
∂x
[refer to (3.6)].

Remark 3.4 D1 (M) is a F(M)-module.

Unless Otherwise Stated, by a form, we will mean Differential form. From


(3.5), we see that
 ∂
d x i (X ) = X x i = ξ j j x i , say
∂x
Thus
d x i (X ) = ξ i . (3.9)

Consequently, from (3.5), we have


 ∂
d f (X ) = X f = ξj f
∂x j
 ∂f
= d x i (X ), by (3.9)
∂xi
 ∂f
df = dxi . (3.10)
∂xi

The transition formula for a 1-form: Let (U, x 1 , x 2 , . . . , x n ) and (V, y 1 , y 2 , . . . ,


y n ) be two charts on M where U ∩ V = φ. Now from (3.8), we have
132 3 Differential Forms
 
ω= fi d x i = g j dy j , say, where each f i , gi ∈ F(M)
i j
 ∂y j i
= gj d x [refer to (3.10)]
i j
∂xi
 ∂y j
fi = gj , i = 1, 2, 3, 4, . . . , n (3.11)
∂xi

as {d x i : i = 1, 2, 3, 4, . . . , n} is a basis.
For every f ∈ F(M), ω ∈ D1 (M), we define f ω ∈ D1 (M) as follows:

( f ω)(X ) = f ω(X )
(3.12)
{( f ω)(X )}( p) = f ( p)ω p (X p ) [refer to (3.3)].

Problem 3.2 Show that ω( f X ) = f ω(X ), where ω is a 1-form on M, f ∈ F(M)


and X ∈ χ (M).

Solution: Here

{ω( f X )}( p) = ω p ( f X )( p), by (3.3)


= ω p f ( p)X p , by (2.26)
= f ( p){ω(X )}( p), by (3.3)
= { f ω(X )}( p), by (2.26)

Thus ω( f X ) = f ω(X ), ∀ p ∈ M.
∂ ∂ ∂
Problem 3.3 Let X = y −x + be the vector field and ω = zd x + xdz
∂x ∂y ∂z
be the 1-form on R3 . Compute ω(X ).

Solution: Taking into consideration Remark 3.3 and also (3.6), we obtain
 
∂ ∂ ∂
ω(X ) = (zd x + xdz) y −x +
∂x ∂y ∂z
= zy + x.

∂ ∂
Problem 3.4 Let X = (x 2 + 1) + (y − 1) be the vector field and
∂x ∂y
ω = (2x y + y + 1)d x + (x − 1)dy be the 1-form on R2 . Compute ω(X ) at (0, 0).
2 2

Solution: Note that


3.1 Cotangent Space 133
 
∂ ∂
ω(X ) = {(2x y + y 2 + 1)d x + (x 2 − 1)dy} (x 2 + 1) + (y − 1)
∂x ∂y
= (2x y + y 2 + 1)(x 2 + 1) + (x 2 − 1)(y − 1).

Thus, ω(X ) = 2.


(0,0)

∂ ∂ ∂
Problem 3.5 Let X = x + 2y , Y = x y be the vector field and ω = (x +
∂x ∂y ∂y
y 2 )d x + (x 2 + y)dy be the 1-form on R2 . Compute ω([X, Y ]).

Solution: Here
    
∂ ∂ ∂ ∂ ∂ ∂
[X, Y ] = x + 2y xy − xy x + 2y
∂x ∂y ∂y ∂y ∂x ∂y

= xy .
∂y

Thus

ω([X, Y ]) = {(x + y 2 )d x + (x 2 + y)dy}x y
∂y
= x y(x 2 + y), by (3.6).

Problem 3.6 In Problem 3.5, compute ω([X, Y ]) at (1, 1).



Solution: Here ω([X, Y ])(1,1) = 2.

∂ ∂ ∂ ∂
Problem 3.7 Let X = −y −x and Y = e x −y be the two vector fields
∂x ∂y ∂x ∂y
on R2 . Find a 1-form ω on R2  {(0, 0)} such that ω(X ) = 1 and ω(Y ) = 0.

Solution: Here
 
−y −x
det = y 2 + xe x = 0 on R2  {(0, 0)}.
e x −y

Let ω = A(x, y)d x + B(x, y)dy, where A, B ∈ F(R2 ) are functions to be deter-
mined.
Given that
 
∂ ∂
1 = ω(X ) = {A(x, y)d x + B(x, y)dy} − y −x
∂x ∂y
∴ 1 = ω(X ) = −Ay − Bx and
0 = ω(Y ) = Ae x − By.
134 3 Differential Forms

After a brief calculation, one gets from the last two equations

y ex
A=− , B=− .
xe x + y2 xe x + y2

Consequently
y ex
ω=− d x − dy.
xe x + y 2 xe x + y 2

NOTE: The 1-form ω is on subset U and it is not a 1-form on R2 , as it is not defined


at the origin.
∂ ∂ ∂
Problem 3.8 Let X = 2 − , Y = ex be two vector fields on R2 . Find a
∂x ∂y ∂y
1-form ω on R2 such that ω(X ) = 1, ω(Y ) = 0.
 
2 −1
Solution: Note that = 2e x = 0, ∀ (x, y) ∈ R2 . Let ω = A(x, y)d x +
0 ex
B(x, y)dy, where A, B ∈ F(R2 ). Now

 ∂ ∂
1 = ω(X ) = {A(x, y)d x + B(x, y)dy} 2 − = 2 A − B and
∂x ∂y
0 = ω(Y ) = −Be x .

1 1
Thus B = 0, A = . Therefore, ω = d x.
2 2
Problem 3.9 Find a 1-form ω on R3 such that ω(X ) = 1, ω(Y ) = 0, ω(Z ) = 0
∂ ∂ ∂ ∂ ∂ ∂
where X = x y + , Y = e−x + ,Z =2 + are vector fields on R3 .
∂x ∂z ∂x ∂y ∂y ∂z
⎛ ⎞
xy 0 1
Solution: Note that ⎝ e−x 1 0 ⎠ = x y + 2e−x = 0, ∀ (x, y, z) ∈ R3 . Let
0 21

ω = A(x, y, z)d X + B(x, y, z)dy + C(x, y, z)dz, wher e A, B, C ∈ F(R3 ).

Now

1 = ω(X ) ⇒ Ax y + C = 1
0 = ω(Y ) ⇒ Ae−x + B = 0
0 = ω(Z ) ⇒ 2B + C = 0
3.1 Cotangent Space 135

1 e−x 2e−x
Solving, we find A = −x
,B =− −x
,C = . Thus
x y + 2e x y + 2e x y + 2e−x

dx e−x dy 2e−x dy
ω= − + .
x y + 2e−x x y + 2e−x x y + 2e−x

Problem 3.10 Find the subset of R3 where the vector fields

∂ ∂ ∂ ∂ ∂ ∂
X= , Y = − , Z= − − (1 − x 2 )
∂x ∂x ∂y ∂x ∂y ∂z

are linearly independent. Write the basis {α, β, γ } dual to {X, Y, Z } in terms of the
basis {d x, dy, dz}.

Solution: Here
⎛ ⎞
1 0 0
det ⎝ 1 −1 0 ⎠ = (1 − x 2 ) = 0 on R3  {(x, y, z)|x = ±1}.
1 −1 −(1 − x )
2

Let us write α = Ad x + Bdy + Cdz, where A, B, C are functions to be determined.


Now from (3.6)

1 = α(X ) = A ⇒ A = 1
0 = α(Y ) = A − B ⇒ B = A = 1
0 = α(Z ) = A − B − C(1 − x 2 ) ⇒ A − B = C(1 − x 2 ) ⇒ C = 0 as x = ±1.

Thus α = d x + dy.
Let us write β = A d x + B dy + C dz, where A , B , C are functions to be deter-
mined. Now

0 = β(X ) = A ⇒ A = 0
1 = β(Y ) = A − B ⇒ B = A − 1 = −1
1
0 = β(Z ) = A − B − C (1 − x 2 ) ⇒ C = .
1 − x2

1
Thus, β = −dy + dz.
1 − x2
dz
Proceeding as above we get, γ = .
x2 −1
Problem 3.11 Find the subset of R2 where the differential forms α = d x + dy, β =
−d x + (x 2 − 1)dy are linearly independent and determine the dual frame {X, Y }
on it.
136 3 Differential Forms

Solution: Here
 
1 1
det = x 2 = 0 on R2 /{(x, y)|x = 0}.
−1 x 2 − 1

Let us write
∂ ∂ ∂ ∂
X =a +b , Y =a +b
∂x ∂y ∂x ∂y

where a, b, a , b are all functions to be determined. Now

1 = X (α) = a + b, a + b = 1
0 = X (β) = −a + b(x 2 − 1), −a + b(x 2 − 1) = 0.

x2 − 1
From the last two equations, one gets a = . Thus
x2

x2 − 1 ∂ 1 ∂
X= + 2 .
x 2 ∂x x ∂y

1 1
Similarly, one gets a = − 2
, b = 2 . Thus
x x
1 ∂ 1 ∂
Y =− + 2 .
x2 ∂x x ∂y

Problem 3.12 Let f be given in spherical co-ordinate system by f (r, θ, φ) =


 π
r tan θ . Consider the point (r, θ, φ) = 1, , 0 . Find the constants A, B, C such
 π 4
that d f 1, , 0 = Adr + Bdθ + Cdφ.
4
Solution: Now
∂f ∂f
= tan θ,  =1
∂r ∂r (1, π ,0)
4

∂f ∂f
= r sec2 θ,  =2
∂θ ∂θ (1, π ,0)
4

∂f
= 0.
∂φ

Thus, d f 1, π4 , 0 = dr + 2dθ .
3.1 Cotangent Space 137

Problem 3.13 Let us write


∂ ∂ ∂
e1 = (1 + y 2 )e z , e2 = (2x y) + (1 + y 2 ) , and
∂x ∂x ∂y
2 ∂ 2 ∂ 2 ∂
e3 = −(x y ) − y(1 + y ) − (1 + y ) .
∂x ∂y ∂z

(i) Prove that {e1 , e2 , e3 } forms a basis of χ (R3 ).


(ii) Find the dual basis {e1 , e2 , e3 } in terms of d x, dy, dz.
(iii) Find [e1 , e2 ], [e1 , e3 ].

Solution:
 
 (1 + y 2 )e z 0 0 
 
(i) Note that  2x y (1 + y 2 ) 0  = − (1 + y 2 )3 e z = 0. Thus

 −x y 2
−y(1 + y ) −(1 + y ) 
2 2

{e1 , e2 , e3 } forms a basis of χ (R3 ).


(ii) Let us write e1 = Ad x + Bdy + Cdz, where A, B, C are functions to be deter-
mined. Now from (3.6), we have

1 = e1 (e1 ) = A(1 + y 2 )e z
0 = e1 (e2 ) = 2x y A + B(1 + y 2 )
0 = e1 (e3 ) = −A(x y 2 ) − By(1 + y 2 ) − C(1 + y 2 ).

On solving, one gets

1 2x y x y2
A= , B = − , C = .
(1 + y 2 )e z (1 + y 2 )e z (1 + y 2 )2 e z

Thus
1 2x y x y2
e1 = dx − dy + dz.
(1 + y )e
2 z (1 + y ) e
2 2 z (1 + y 2 )2 e z

Proceeding as above, we get

1 y
e2 = dy − dz
(1 + y 2 ) (1 + y 2 )
1
e3 = − dz.
(1 + y 2 )

(iii) It is to be noted that


 
∂ ∂ ∂
[e1 , e2 ] = (1 + y 2 )e z , 2x y + (1 + y 2 )
∂x ∂x ∂y
 
∂ ∂ ∂ ∂
= (1 + y 2 )e z , 2x y ] + [(1 + y 2 )e z , (1 + y 2 ) , by linearity
∂x ∂x ∂x ∂y
138 3 Differential Forms

Taking help of the relation [ f X, gY ] = ( f g)[X, Y ] + { f (X g)}Y − {g(Y f )}X ,



it can be shown that [e1 , e2 ] = 0. Similarly, [e1 , e3 ] = (1 + y 2 )2 e z =
∂x
(1 + y 2 )e1 .

Exercises

Exercise 3.2 If ω = zd x + ydz, compute ω(X ) on R3 where


∂ ∂
(i) X = x y + x2
∂x ∂z

(ii) X =y
∂y
∂ ∂ ∂
(iii) X =2 − +3
∂x ∂x ∂x

(iv) X = e−x
∂x
∂ ∂
(v) X = + ex
∂y ∂z
∂ x2 ∂
(vi) X = xy − .
∂y 2 ∂x
Exercise 3.3 Compute ω(X ) at (1, 0, 1) in each cases of Exercise 3.2.

Exercise 3.4 Find a 1-form ω on R2 \ {(0, 0)}, such that ω(X ) = 1 and ω(Y ) = 0
where
∂ ∂ ∂
(i) X = x y + x2 , Y = y
∂x ∂y ∂y
∂ ∂ ∂
(ii) X = 2 − , Y = e−x
∂x ∂y ∂x
∂ ∂ x2 ∂ ∂
(iii) X = + ex , Y = − + xy .
∂y ∂x 2 ∂x ∂y

Exercise 3.5 Let ω = (2x y + y 2 + 1)d x + (x 2 − 1)dy + xdz be the 1-form on R3 .


Compute ω(X ) and hence ω(X ) at (1, 0, 0) where
∂ ∂ ∂
(i) X = 2x y 2 + (2 + y 2 ) + (2 − z) .
∂x ∂y ∂z
2 ∂ z ∂ 2 ∂
(ii) X = 2y +e + (2 − y ) .
∂x ∂y ∂z
2 ∂ ∂ ∂
(iii) X = −y(2 + y ) − 2z + (2 + x 2 ) .
∂x ∂y ∂z
∂ ∂ ∂ ∂
Exercise 3.6 Let X = x + 2y and Y = x y +y be two vector fields on
∂x ∂y ∂x ∂y
R2 . Find a 1-form ω on R2 \ {(0, 0)} such that ω(X ) = 1 and ω(Y ) = 0.
3.1 Cotangent Space 139

Exercise 3.7 Find the basis {α, β} dual to {X, Y } where


∂ ∂ ∂ ∂
(i) X = −y − x , Y = ex −y .
∂x ∂y ∂x ∂y
∂ ∂ ∂
(ii) X = , Y = − .
∂x ∂x ∂y
∂ ∂ ∂
(iii) X = 2y 2 + ex , Y = .
∂x ∂y ∂y
∂ ∂ 1 ∂ ∂
(iv) X = e x + , Y =− +y .
∂x ∂y 2 ∂x ∂y
Exercise 3.8 Show that the given differential forms
dx 2x y
α= − dy,
(2 + y )e
2 z (2 + y 2 )2 e z
dy y
β= + dz,
(2 + y 2 ) (2 + y 2 )
dz
γ = ,
(2 + y 2 )
are linearly  Hence find the dual basis {X, Y, Z } dual to {α, β, γ } in
 independent.
∂ ∂ ∂
terms of , , .
∂ x ∂ y ∂z

Exercise 3.9 Find the subset of R2 where the differential forms

x dx y dy y dx x dy
α= + 2 , and β = − 2 + 2
x2 +y 2 x +y 2 x +y 2 x + y2

 independent. Hence find the dual basis {X, Y } dual to {α, β} in terms of
are linearly

∂ ∂
, .
∂x ∂y

Exercise 3.10 Find the subset of R3 where the differential forms

dz dz
α = d x + dy, β = −dy + , γ = 2
1 − x2 x −1

are linearly  Hence find the dual basis {X, Y, Z } dual to {α, β, γ } in
 independent.
∂ ∂ ∂
terms of , , .
∂ x ∂ y ∂z

Exercise 3.11 Let f : R3 → R be given by


3
f (x, y, z) = (x 2 + y 2 + z 2 ) 2 + x + 3y.

(i) Write down d f (1, 0, 0) in terms of {d x, dy, dz}.


140 3 Differential Forms

(ii) Express f in spherical co-ordinate (r, θ, φ), where

x = r sin θ cos φ, y = r sin θ sin φ, z = r cos θ.



(iii) Find constants A, B, C such that d f 1, π2 , 0 = Adr + Bdθ + Cdφ.

Exercise 3.12 A. If ω = e x cos yd x + e x sin ydy, compute ω([X, Y ]) on R2 where


∂ ∂ ∂
(i) X = x + y2 , Y = y
∂x ∂y ∂y
∂ y2 ∂ ∂ ∂
(ii) X = x y + ,Y = 2 −
∂x 2 ∂y ∂x ∂y
∂ ∂ ∂
(iii) X = y , Y = e x + .
∂y ∂x ∂y
B. Compute ω([X, Y ]) at (1, 2) ∈ R2 .

Exercise 3.13 Find a 1-form ω on R3 such that ω(X ) = 1, ω(Y ) = 0, ω(Z ) = 0


where
∂ ∂ ∂ ∂ ∂ ∂
(i) X = ,Y = − ,Z = − −
∂x ∂x ∂y ∂x ∂y ∂z
2 ∂ ∂ ∂ ∂ ∂ ∂
(ii) X = (2 + y ) + ,Y = + ,Z = + .
∂x ∂y ∂y ∂z ∂x ∂z

Answers

x2z
3.2. (i) x yz + x 2 y (ii) θ (iii) 2z + 3y (iv) ze−x (v) ye x (vi) − .
2
1
3.3. (i) 0 (ii) 0 (iii) 2 (iv) (v) − 21 .
e
1 ∂ ∂ 2y ∂ x ∂
3.4. (i) ω = (ii) ω = − (iii) ω = + .
xy ∂x ∂y (x + 2ye ) ∂ x
x (x + 2ye ) ∂ y
x
dx dy
3.5. (i) 2 (ii) 2 (iii) 3. 3.6. ω = − +
x(2y − 1) 2y − 1
y ex x y
3.7. (i) α = − x dx − x dy, β = x dx − x dy.
xe + y 2 xe + y 2 xe + y 2 xe + y 2
1
(ii) α = d x + dy, β = −dy (iii) α = 2 d x, β = dy.
2y
2y 1 2 2e x
(iv) α = d x + dy, β = − d x + dy.
1 + 2ye x 1 + 2ye x 1 + 2ye x 1 + 2ye x
∂ ∂ ∂ ∂ ∂
3.8. X = (2 + y 2 )e z , Y = 2x y + (2 + y 2 ) , Z = −2x y 2 − y(2 + y 2 ) +
∂x ∂x ∂y ∂x ∂y

(2 + y 2 )
∂z
∂ ∂ ∂ ∂
3.9. X = x + y , Y = −y +x
∂x ∂y ∂x ∂y
3.2 r -form, Exterior Product 141

∂ ∂ ∂ ∂ ∂ ∂
3.10. X = , Y = − , Z= − + (x 2 − 1)
∂x ∂x ∂y ∂x ∂y ∂z
3.11. (i) 4d x + 3dy (ii) r 3 + r sin θ cos φ + 3r sin θ sin φ (iii) 4dr + 3dφ.
3.12. A. (i) −y 2 e x sin y (ii) (x − 2y)e x cos y (iii) −e x sin y
B. (i) −4e sin 2 (ii) −3e cos 2 (iii) −e sin 2.
dx dy dz
3.13. (i) d x + dy (ii) + − .
3+y 2 3+y 2 3 + y2

3.2 r-form, Exterior Product

An r -form ω is a skew-symmetric mapping

ω : χ (M) × · · · χ (M) → F(M)


  
r −times

such that
(i) ω is R-linear
(ii) if σ is a permutation of 1, 2, 3, . . . , r with (1, 2, 3, . . . , r ) → (σ (1), σ (2),
. . . , σ (r )), then

1 
ω(X 1 , X 2 , . . . , X r ) = (sgn σ ) ω(X σ (1) , X σ (2) , . . . , X σ (r ) ) (3.13)
r! σ

where sgn σ (pronounced as signum σ ) is +1 or −1 according as σ is even or


odd permutation.
The product of two skew-symmetric form is called the exterior product or Grass-
mann product, as introduced by H.G Grassmann or Wedge Product, as a wedge
‘∧’ is used to denote this product. We are going to give the formal definition.
Remark 3.5 By convention, a zero-form is a function.
If ω is a r -form and μ is a s-form, then the exterior product or wedge product of ω
and μ, denoted by ω ∧ μ is a (r + s)-form defined as

(ω ∧ μ)(X 1 , X 2 , . . . , X r , X r +1 , . . . , X r +s ) (3.14)
1 
= (sgn σ ) ω(X σ (1) , X σ (2) , . . . , X σ (r ) )μ(X σ (r +1) , X σ (r +2) , . . . , X σ (r +s) )
(r + s)! σ

where σ ranges over the permutation (1, 2, 3, . . . , r + s), X i ∈ χ (M), i = 1, 2,


3, . . . , r + s.
For convenience, we write

f ∧ g = f g; ∀ f, g ∈ F(M). (3.15)
142 3 Differential Forms

It can be shown that, for a r -form ω



( f ∧ ω)(X 1 , X 2 , . . . , X r ) = f ω(X 1 , X 2 , . . . , X r )
(3.16)
(ω ∧ f )(X 1 , X 2 , . . . , X r ) = f ω(X 1 , X 2 , . . . , X r )

Again, if ω and μ are 1-forms, then

1
(ω ∧ μ)(X 1 , X 2 ) = {ω(X 1 )μ(X 2 ) − ω(X 2 )μ(X 1 )}. (3.17)
2
The exterior product obeys the following properties:


⎪ ω ∧ ω = 0, ω ∧ μ = (−1)r s μ ∧ ω, μ being s-form

f ω ∧ μ = f (ω ∧ μ) = ω ∧ f μ
(3.18)

⎪ f ω ∧ gμ = f g ω ∧ μ

(ω + μ) ∧ γ = ω ∧ γ + μ ∧ γ , ω ∧ (μ + γ ) = ω ∧ μ + ω ∧ γ .

Problem 3.14 Given 1-form ω = f d x − gdy + hdz and μ = f d x + g dy in R3 .


Compute ω ∧ ω and show that ω ∧ μ = −μ ∧ ω. Prove that each ω ∧ μ, μ ∧ ω is
a 2-form.

Solution: Taking help of (3.18), the result follows immediately.

Problem 3.15 Let V be a vector space of dimension 3 over R. Let {e1 , e2 , e3 } be


a basis of V and {e1 , e2 , e3 } be its dual basis. Let α, β be co-vectors, so that α =
a1 e1 + a2 e2 + a3 e3 , β = b1 e1 + b2 e2 + b3 e3 where each ai , bi ∈ F(V ). Show that
the components of α ∧ β = α × β.

Solution: Here

α ∧ β = (a1 e1 + a2 e2 + a3 e3 ) ∧ (b1 e1 + b2 e2 + b3 e3 )
= (a1 b2 − a2 b1 )e1 ∧ e2 + (a2 b3 − a3 b2 )e2 ∧ e3 + (a3 b1 − a1 b3 )e3 ∧ e1 .

Thus the components of α ∧ β = (a1 b2 − a2 b1 , a2 b3 − a3 b2 , a3 b1 − a1 b3 ). Again

α × β = (a1 b2 − a2 b1 , a2 b3 − a3 b2 , a3 b1 − a1 b3 ).

Hence the proof.

Problem 3.16 Let a 1-form α and 2-form β be given on R3 as α = ad x + bdy +


cdz,
β = a d x ∧ dy + b dy ∧ dz + c dz ∧ d x. Compute α ∧ β.
3.2 r -form, Exterior Product 143

Solution: Note that

α ∧ β = (ad x + bdy + cdz) ∧ (a d x ∧ dy + b dy ∧ dz + c dz ∧ d x)


= ab d x ∧ dy ∧ dz + bc dy ∧ dz ∧ d x + ca dz ∧ d x ∧ dy.
As d x ∧ dz ∧ d x = −d x ∧ d x ∧ dz = 0 and so on, we have
α ∧ β = (ab + bc + ca )d x ∧ dy ∧ dz,

which is a 3-form.

Theorem 3.2 In terms of a local co-ordinate system (x 1 , x 2 , . . . , x n ) in a neigh-


bourhood U of p of M, an r -form can be expressed uniquely as

ω= f i1 i2 ···ir d x i1 ∧ d x i2 ∧ · · · ∧ d x ir , i 1 < i 2 < . . . < ir (3.19)
i 1 ,i 2 ,...,ir

where f i1 i2 ···ir are differentiable functions.

Proof As the set {d x i1 ∧ d x i2 ∧ · · · ∧ d x ir } is a basis of Dr (M n ), from (3.14) we


find

1 
(d x i1 ∧ d x i2 ∧ · · · ∧ d x ir )(X 1 , X 2 , . . . , X r ) = (sgn σ )d x i1 (X σ (1) ) . . . d x ir (X σ (r ) ),
r! σ
where i 1 < i 2 < . . . < ir .

Let us write,
 j ∂
Xi = ξi m , i = 1, 2, 3, 4, . . . , r,
im
∂ x jm

where each ξiim is C ∞ function. Using this in the foregoing equation, we obtain

(d x i1 ∧ d x i2 ∧ · · · ∧ d x ir )(X 1 , X 2 , . . . , X r )
   
1  j ∂ j ∂
= (sgn σ )d x i1 ξσ m(1) j · · · d x ir ξσ m(r ) j
r! ∂x m ∂x m
(σ )
1 
= (sgn σ )ξσi1(1) · · · ξσir(r ) , i 1 < i 2 < . . . < ir (by (3.6)).
r!
(σ )

From (3.13), we see that


144 3 Differential Forms
 
1  j ∂  j ∂
ω(X 1 , X 2 , . . . , X r ) = (sgn σ ) ω ξσ m(1) j , . . . , ξσ m(r ) j
r! ∂x m ∂x m
(σ )
 
1   ∂ ∂
= (sgn σ ) ξσi1(1) · · · ξσir(r ) ω , . . . ,
r! ∂ x i1 ∂ x ir
(σ ) i 1 ,i 2 ,...,ir

1  
= (sgn σ ) ξσi1(1) · · · ξσir(r ) f i1 i2 ···ir , say, as defined in (3.7)
r!
(σ ) i 1 ,i 2 ,...,ir

= (d x i1
∧ d x i2 ∧ · · · ∧ d x ir )(X 1 , X 2 , . . . , X r ) f i1 i2 ···ir , from above.
i 1 ,i 2 ,...,ir
i 1 <i 2 <...<ir


Thus ω = f i1 i2 ···ir d x i1 ∧ d x i2 ∧ · · · ∧ d x ir , for each X i , i = 1, 2,
i 1 ,i 2 ,...,ir
i 1 <i 2 <...<ir
3, . . . , r .

Theorem 3.3 In terms of a local co-ordinate system (x 1 , x 2 , . . . , x n ) in a neigh-


bourhood U of p of a manifold, let f i , i = 1, 2, 3, . . . , r , be smooth functions on U .
Then
 ∂( f 1 , f 2 , . . . , f r ) i
df1 ∧ df2 ∧ ··· ∧ dfr = d x 1 ∧ d x i 2 ∧ · · · ∧ d x ir .
i 1 ,i 2 ,...,ir
∂(x i1 , x i2 , . . . , x ir )
i 1 <i 2 <...<ir

Proof Each d f i , i = 1, 2, 3, . . . , r , is a 1-form and d f 1 ∧ d f 2 ∧ · · · ∧ d f r is a r -


form and in view of Theorem 3.2, we write

df1 ∧ df2 ∧ ··· ∧ dfr = F j1 j2 ··· jr d x j1 ∧ d x j2 ∧ · · · ∧ d x jr , j1 < j2 < . . . < jr ,
j1 , j2 ,..., jr

where F j1 j2 ··· jr are differentiable functions to be determined. In view of (3.10), we


find
 ∂f  ∂fi
df = d x j
⇒ d f i
= dxk.
j
∂ x j
k
∂ x k

Now    ∂fi    i
∂ ∂ ∂f k ∂fi
df i
= d x k
= δ = .
∂x j k
∂xk ∂x j k
∂xk j
∂x j

Hence from the left hand side of the above expression, we have
 
∂ ∂ ∂ ∂( f 1 , f 2 , . . . , f r )
(d f 1 ∧ d f 2 ∧ · · · ∧ d f r ) , ,..., i = .
∂ x i1 ∂ x i2 ∂x r ∂(x i1 , x i2 , . . . , x ir )

Furthermore, from the right hand side, we get


3.2 r -form, Exterior Product 145

   
∂ ∂ ∂ i ...i
F j1 j2 ··· jr d x j1 ∧ d x j2 ∧ · · · ∧ d x jr , ,..., = F j1 j2 ··· jr δ j1 ... jr
j1 , j2 ,..., jr
∂ x i1 ∂ x i2 ∂ x ir j1 , j2 ,..., jr
1 r

= Fi1 i2 ···ir .

Thus
∂( f 1 , f 2 , . . . , f r )
= Fi1 i2 ···ir .
∂(x i1 , x i2 , . . . , x ir )

Consequently,

 ∂( f 1 , f 2 , . . . , f r )
df1 ∧ df2 ∧ ··· ∧ dfr = d x j1 ∧ d x j2 ∧ · · · ∧ d x jr
j1 , j2 ,..., jr
∂(x j1 , x j2 , . . . , x jr )
j1 < j2 <...< jr
 ∂( f 1 , f 2 , . . . , f r )
i.e. d f 1 ∧ d f 2 ∧ · · · ∧ d f r = d x i 1 ∧ d x i 2 ∧ · · · ∧ d x ir .
i 1 ,i 2 ,...,ir
∂(x i1 , x i2 , . . . , x ir )
i 1 <i 2 <...<ir

This completes the proof.

Proposition 3.1 If ω and μ are C ∞ -forms on M, then ω ∧ μ is also C ∞ .

Proof From (3.8) and Theorem 3.1, we know that


 
ω= fi d x i , μ = gj dx j ,
i j

where f i , g j ∈ F(M) for each i, j. By Remark 3.2, we say that ω, μ are C ∞ -forms
on M, if each f i , g j are C ∞ -functions on M. Now

ω∧μ= f i g j d x i ∧ d x j , Theorem 3.3.
i, j

 ∞
Since i, j f i g j are C -functions on M, by Remark 3.2, we can claim that

ω ∧ μ is also C -forms on M.

Exercises

Exercise 3.14 If ω is a 1-form and μ is a 2 form, show that

1
(ω ∧ μ)(X 1 , X 2 , X 3 ) = {ω(X 1 )μ(X 2 , X 3 ) + ω(X 2 )μ(X 3 , X 1 ) + ω(X 3 )μ(X 1 , X 2 )}.
3

Exercise 3.15 Compute ω ∧ μ where


(i) ω = xd x − ydy, μ = yd x + hdy
(ii) ω = xd x + ydy, μ = gd x
146 3 Differential Forms

(iii) ω = zd x + xdz, μ = xd x + ydy + zdz


(iv) ω = x dy ∧ dz, μ = gd x + hdz.
Exercise 3.16 Compute
(i) (2du 1 + du 2 ) ∧ (du 1 − du 2 ).
(ii) (6du 1 ∧ du 2 + 27du 1 ∧ du 3 ) ∧ (du 1 + du 2 + du 3 ).
(iii) θ ∧ φ ∧ ψ where θ = zdy, φ = xd x + ydy, ψ = zd x − ydz.
Exercise 3.17 If α = −du 1 + du 2 − 2du 3 is a 1-form and β is a 2-form given by
β = 2 du 1 ∧ du 3 − du 2 ∧ du 3 , compute α ∧ β ∧ α.

Answers

3.14. Use (3.14).


3.15. (i) (xh + y 2 ) d x ∧ dy. (ii) −yg d x ∧ dy
(iii) (z 2 − x 2 ) d x ∧ dz + zy d x ∧ dy − x y d x ∧ dy (iv) xg d x ∧ dy ∧ dz.
3.16. (i) −3 du 1 ∧ du 2 . (ii) −21 du 1 ∧ du 2 ∧ du 3 . (iii) x yz d x ∧ dy ∧ dz.
3.17. 0 (Zero).
Remark 3.6 Let Dr (M n ) denote the collection of all r -forms in M n . By virtue of
Theorem 3.1, the set {d x i1 ∧ d x i2 ∧ · · · ∧ d x ir : 1 ≤ i 1 < i 2 < · · · < ir ≤ n} forms
a basis of Dr (M n ).
Remark 3.7 The collection of all differential forms with respect to wedge product,
forms an algebra, called the EXTERIOR ALGEBRA. Now a days it is also termed as
GRASSMANN ALGEBRA, as R.G. Grassmann developed this powerful concept.
Remark 3.8 For a manifold M 4 , every point p ∈ M, has coordinates of the form
p = (x 1 , x 2 , x 3 , x 4 ). Thus the basis set of D1 (M 4 ) contains 4 C1 elements viz
{d x 1 , d x 2 , d x 3 , d x 4 };
that of D2 (M 4 ) is 4 C2 i.e. {d x 1 ∧ d x 2 , d x 1 ∧ d x 3 , d x 1 ∧ d x 4 , d x 2 ∧ d x 3 , d x 2 ∧
d x 4 , d x 3 ∧ d x 4 };
that of D3 (M 4 ) is 4 C3 i.e. {d x 1 ∧ d x 2 ∧ d x 3 , d x 1 ∧ d x 2 ∧ d x 4 , d x 1 ∧ d x 3 ∧ d x 4 ,
d x 2 ∧ d x 3 ∧ d x 4 };
and that of D4 (M 4 ) is 4 C4 given by {d x 1 ∧ d x 2 ∧ d x 3 ∧ d x 4 }. However, D5 (M 4 )
does not exist. Hence any,


4
ω ∈ D1 (M 4 ) is of the form ω = fi d x i ;
i=1

ω ∈ D2 (M ) is of the form ω =
4
f i j d x i ∧ d x j , 1 ≤ i < j ≤ 4;
ij

ω ∈ D3 (M ) is of the form ω =
4
f i jk d x i ∧ d x j ∧ d x k , 1 ≤ i < j < k ≤ 4;
i, j,k

ω ∈ D4 (M ) is of the form ω = f 1234 d x 1 ∧ d x 2 ∧ d x 3 ∧ d x 4 .


4
3.2 r -form, Exterior Product 147

In Sect. 3.1, the transition formula for a 1-form has been given by the (3.11). Now
we will focus to find the Transition Formula for a 2-We write it asform.
Let (U ; x 1 , x 2 , . . . , x n ) and (V ; y 1 , y 2 , . . . , y n ) be two charts on a manifold M
such that U ∩ V = φ. If ω is a 2-form, then by (3.19), we have
 
ω= fi j d x i ∧ d x j = gmn dy m ∧ dy n , where each f i j , gmn ∈ F(M)
i, j m,n
i< j m<n
 ∂(y m , y n ) i
= gmn d x ∧ d x j , by Theorem 3.3
i, j m,n
∂(x i , x j )
i< j m<n
 ∂(y m , y n )
fi j = gmn , i, j = 1, 2, 3, . . . , n, (3.20)
m,n ∂(x , x )
i j
m<n

as {d x i ∧ d x j : i, j = 1, 2, 3, . . . , n; 1 ≤ i < j ≤ n} is a basis of 2-form.


Problem 3.17 Let {ω1 , ω2 , ω3 } be a set of linearly independent 1-form on a smooth
 3
manifold M. Define 1-form μi , i = 1, 2, 3 as μi = ai j ω j , i = 1, 2, 3. Show that
j=1

μ1 ∧ μ2 ∧ μ3 = det(ai j )ω1 ∧ ω2 ∧ ω3 .

Solution: We write
  
μ1 = a1 j ω j , μ2 = a2 j ω j μ3 = a3 j ω j .

Thus

μ1 = a11 ω1 + a12 ω2 + a13 ω3


μ2 = a21 ω1 + a22 ω2 + a23 ω3
μ3 = a31 ω1 + a32 ω2 + a33 ω3 .

Now
μ1 ∧ μ2 = (a11 a22 − a12 a21 )ω1 ∧ ω2 + (a11 a23 − a13 a21 )ω1 ∧ ω3
+ (a12 a23 − a13 a22 )ω2 ∧ ω3 ,

as ω1 ∧ ω1 = 0, ωi ∧ ω j = −ω j ∧ ωi . Again Now
μ1 ∧ μ2 ∧ μ3 = a33 (a11 a22 − a12 a21 )ω1 ∧ ω2 ∧ ω3
+ a32 (a11 a23 − a13 a21 )ω1 ∧ ω2 ∧ ω3
+ a31 (a12 a23 − a13 a22 )ω1 ∧ ω2 ∧ ω3
= det(ai j )ω1 ∧ ω2 ∧ ω3 , as ωi ∧ ω j = −ω j ∧ ωi , i, j = 1, 2, 3.
148 3 Differential Forms

Problem 3.18 (Cartan Lemma) Let k < n and {ω1 , ω2 , . . . , ωk } be 1-forms on M n


which are linearly independent pointwise. Let μi be k number of 1-forms on M
satisfying
k
μi ∧ ωi = 0.
i=1

Prove that there exists C ∞ functions Ai j on M n such that


k
μi = Ai j ω j with Ai j = A ji , i = 1, 2, 3, 4, . . . , k.
j=1

Solution: As {ω1 , ω2 , . . . , ωk } is k-independent 1-forms on M, we complete the


basis of D1 (M) by taking 1-forms ωk+1 , . . . , ωn . Consequently, any 1-form μi , i =
1, 2, . . . , k can be expressed as


k 
n
μi = Aim ωm + Bi p ω p , i = 1, 2, 3, . . . , k,
m=1 p=k+1


k
where each Aim , Bi p is C ∞ -function. Given that μi ∧ μi = 0 i.e.
i=1

   
   
A1m ωm + B1 p ω p ∧ ω1 + · · · + Akm ωm + Bkp ω p ∧ ωk = 0.
m p m p

Using the properties

ωi ∧ ωi = 0 and ωi ∧ ω j = −ω j ∧ ωi ,

one gets  
(Ai j − A ji )ωi ∧ ω j + Bi j ωi ∧ ω j = 0.
i, j i≤k
i< j≤k j>k

As {ω1 , ω2 , . . . , ωn } is a basis of D1 (M), we must have Ai j − A ji = 0 and Bi j = 0.


Consequently, μi , i = 1, 2, . . . , k can be expressed as


k
μi = Aim ω j , with Ai j = A ji .
j=1
3.3 Exterior Differentiation 149

Exercise

Exercise 3.18 Show that a set of 1-forms {ω1 , ω2 , . . . , ωk } is linearly dependent if


and only if ω1 ∧ ω2 ∧ · · · ∧ ωk = 0.

Exercise 3.19 Let {ω1 , . . . , ωn } be a set of linearly independent 1-forms on a smooth


manifold. Let μi , i = 1, 2, 3, . . . , n be 1-forms on M satisfying


n
μi = ai j ω j .
j=1

Show that μ1 ∧ μ2 ∧ μ3 ∧ . . . ∧ μn = det(ai j ) ω1 ∧ ω2 ∧ ω3 ∧ . . . ∧ ωn .

3.3 Exterior Differentiation



As discussed in Remark 3.7 of the last section, we write D = rn=0 Dr (M n ), as the
Exterior Algebra, with respect to the wedge product in M n . It is interesting to note
that

dim D =n C0 +n C1 +n C2 + . . . +n Cn
= (1 + 1)n , by Binomial Theorem
= 2n .

We are now going to define the exterior derivative which is a linear mapping, denoted
by d, on D as follows:


⎪ (i) d(Dr ) ⊂ Dr +1 ;

(ii) for f ∈ D0 , d f is the total differential;
(3.21)

⎪ (iii) if ω ∈ Dr , μ ∈ Ds , then d(ω ∧ μ) = dω ∧ μ + (−1)r ω ∧ dμ;

(iv) d 2 = 0.

Hence from (3.19), we see that



dω = d f i1 i2 ...ir ∧ d x i1 ∧ · · · ∧ d x ir , i 1 < i 2 < . . . < ir . (3.22)
i 1 ,i 2 ,...,ir

Problem 3.19 Find the exterior derivative of the following:


(i) f = x 2 y 3 z (ii) f g.
150 3 Differential Forms

Solution:
(i) Here d f = 2x y 3 z d x + 6x 2 y 2 z dy + x 2 y 3 dz. Thus d f is a 1-form, verifying
(3.21)(ii)
(ii) We know d f (X ) = X f . Thus

d( f g)(X ) = X ( f g)
= g(X f ) + f (X g)
= g d f (X ) + f dg(X ),

i.e. d( f g) = g d f + f dg.
Problem 3.20 Verify (3.21)(iv) for ω = x 2 y d x.
Solution: Note that

dω = d(x 2 y d x)
= d(x 2 y) ∧ d x, see (3.22)
= (2x y d x + x 2 dy) ∧ d x
= x 2 dy ∧ d x, as d x ∧ d x = 0.

Again

d(dω) = d(x 2 dy ∧ d x)
= d(x 2 ) ∧ dy ∧ d x, see (3.22)
= 2x d x ∧ dy ∧ d x
= −2x d x ∧ d x ∧ dy
= 0.

Problem 3.21 Find the exterior derivative of:


(i) ω1 = 2xd x + (x + y)dy
(ii) ω2 = x 2 yd x − x z 3 dy + 3x ydz
(iii) ω3 = (x 2 − y 2 )d x ∧ dy
(iv) ω4 = x 2 yz dy ∧ dz − 2x yz dz ∧ d x + x yz 3 d x ∧ dy.
Solution:
(i) Here ω1 = 2xd x + (x + y)dy. Therefore

dω1 = d(2xd x) + d{(x + y)dy}, by linearity


= d(2x) ∧ d x + (−1)0 2xd(d x) + d(x + y) ∧ dy + (−1)0 (x + y)d(dy), by (iii) of (3.21),
as the function is assumed to be a 0 − form.
= (2d x) ∧ d x + (d x + dy) ∧ dy, by (iv) of (3.21)
= 0 + d x ∧ dy + 0
= d x ∧ dy.
3.3 Exterior Differentiation 151

(ii) Note that ω2 = x 2 yd x − x z 3 dy + 3x ydz. Hence

dω2 = d(x 2 y) ∧ d x − d(x z 3 ) ∧ dy + d(3x y) ∧ dz, by (3.21)


= (2x yd x + x 2 dy) ∧ d x − (z 3 d x + 3x z 2 dz) ∧ dy + (3yd x + 3xdy) ∧ dz
= x 2 dy ∧ d x − z 3 d x ∧ dy − 3x z 2 dz ∧ dz + 3yd x ∧ dz + 3xdy ∧ dz
= (−x 2 − z 3 )d x ∧ dy + (3x z 2 + 3x)dy ∧ dz − 3ydz ∧ d x.

(iii) Here ω3 = (x 2 − y 2 )d x ∧ dy. Therefore

dω3 = d(x 2 − y 2 ) ∧ d x ∧ dy, by (3.21)


= (2xd x − 2ydy) ∧ d x ∧ dy
= 2xd x ∧ d x ∧ dy − 2ydy ∧ d x ∧ dy
= 0 + 2ydy ∧ dy ∧ d x
= 0.

(iv) Similarly
dω4 = (2x yz − 2x z + 3x yz 2 ) d x ∧ dy ∧ dz.

Problem 3.22 Let ω be a 1-form on a manifold M. Consider a nowhere vanishing


function f : M → R such that d( f ω) = 0. Prove that ω ∧ dω = 0.
Solution: From the definition,

d( f ω) = d f ∧ ω + f ∧ dω, by (3.21)
∴ 0 = d f ∧ ω + + f dω, by (3.16)
1
or dω = − (d f ∧ ω).
f
1
Now ω ∧ dω = ω ∧ − (d f ∧ ω)
f
1
= ω ∧ (ω ∧ d f ), by (3.18)
f
= 0, by (3.18).

Problem 3.23 For any 1-form ω, is ω ∧ dω = 0? Justify your answer with an exam-
ple.
Solution: Let ω ∈ D1 (R3 ) be such that

ω = f d x 1 + gd x 2 + hd x 3 , where f, g, h ∈ F(R3 ).
152 3 Differential Forms

Then
dω = d f ∧ d x 1 + dg ∧ d x 2 + dh ∧ d x 3 , by (3.21).

Thus

ω ∧ dω = f d x 1 ∧ dg ∧ d x 2 + f d x 1 ∧ dh ∧ d x 3 + gd x 2 ∧ d f ∧ d x 1
+ gd x 2 ∧ dh ∧ d x 3 + hd x 3 ∧ d f ∧ d x 1 + hd x 3 ∧ dg ∧ d x 2
= 0, always.

Problem 3.24 Consider the 1-forms ω1 , ω2 , ω3 defined by

ω1 = hd x 1 − x 1 dh − x 2 d x 3 + x 3 d x 2
ω2 = hd x 2 − x 2 dh − x 3 d x 1 + x 1 d x 3
ω3 = hd x 3 − x 3 dh − x 1 d x 2 + x 2 d x 1

 terms of a local co-ordinate system (x , x , x ) h=


1 2 3
in where
1 − (x ) − (x ) − (x ) . Show that dω1 = −2ω2 ∧ ω3 .
1 2 2 2 3 2

Solution: Note that

dω1 = −2d x 1 ∧ dh − 2d x 2 ∧ d x 3 ,
x 1d x 1 + x 2d x 2 + x 3d x 3
dh = −  .
1 − (x 1 )2 − (x 2 )2 − (x 3 )2

Thus
x 2d x 1 ∧ d x 2 + x 3d x 1 ∧ d x 3
d x 1 ∧ dh = −  .
1 − (x 1 )2 − (x 2 )2 − (x 3 )2

Again

ω2 ∧ ω3 = h 2 d x 2 ∧ d x 3 − x 3 h d x 2 ∧ dh + x 2 h d x 2 ∧ d x 1 − x 2 hdh ∧ d x 3
+ x 1 x 2 dh ∧ d x 2 − (x 2 )2 dh ∧ d x 1 − x 3 h d x 1 ∧ d x 3
+ (x 3 )2 d x 1 ∧ dh + x 1 x 3 d x 1 ∧ d x 2 − x 1 x 3 d x 3 ∧ dh
− (x 1 )2 d x 3 ∧ d x 2 + x 1 x 2 d x 3 ∧ d x 1 .

It is to be noted that
3.3 Exterior Differentiation 153

h 2 d x 2 ∧ d x 3 = {1 − (x i )2 }d x 2 ∧ d x 3
i
−x h d x ∧ dh = −x x d x 1 ∧ d x 2 + (x 3 )2 d x 2 ∧ d x 3
3 2 1 3
 
x 2 h d x 2 ∧ d x 1 = −x 2 1 − (x i )2 d x 1 ∧ d x 2
i

−x h dh ∧ d x = −x x d x ∧ d x 1 + (x 2 )2 d x 2 ∧ d x 3
2 3 1 2 3

(x 1 )2 x 2 x1x2x3
x 1 x 2 dh ∧ d x 2 = −   d x 1
∧ d x 2
+   dx2 ∧ dx3
1− (x i )2 1− (x i )2
i i

(x ) 2 3
(x ) x 2 2 3
−(x 2 )2 dh ∧ d x 1 = −   dx1 ∧ dx2 +   dx3 ∧ dx1
1− (x i )2 1− (x i )2
i i
 
−x 3 h d x 1 ∧ d x 3 = x 3 1− (x i )2 d x 3 ∧ d x 1
i

x (x 3 )2
2
(x 3 )3
(x 3 )2 d x 1 ∧ dh = −   d x 1
∧ d x 2
+   dx3 ∧ dx1
1− (x i )2 1− (x i )2
i i

(x 1 )2 x 3 x1x2x3
−x 1 x 3 d x 3 ∧ dh =   dx3 ∧ dx1 −   dx2 ∧ dx3
1− (x )
i 2
1− (x )
i 2

i i

−(x ) d x ∧ d x = (x ) d x ∧ d x
1 2 3 2 1 2 2 3

2x 2 2x 3
Thus dω1 =   dx1 ∧ dx2 −   d x 3 ∧ d x 2 − 2 d x 2 ∧ d x 3.
1− (x )
i 2
1− (x )
i 2

i i

Substituting the above results, one gets

x2 x3
ω2 ∧ ω 3 = d x 2 ∧ d x 3 −   dx1 ∧ dx2 +   dx3 ∧ dx1
1− (x )
i 2
1− (x )
i 2

i i

2x 2 2x 3
∴ −2 ω2 ∧ ω3 =   dx1 ∧ dx2 −   dx3 ∧ dx1 − 2 dx2 ∧ dx3
1− (x )
i 2
1− (x )
i 2

i i

= dω1 .

Thus dω1 = −2 ω2 ∧ ω3 .
154 3 Differential Forms

Exercises

Exercise 3.20 Find the exterior derivative of the following:


(i) f = 2x yz 2 (ii) f = x 2 + y 2 − 3z 4 (iii) f = x 2 y + y 2 z − z 2 x.

Exercise 3.21 Find the exterior derivative of:


(i) ω1 = x 2 ydy − x y 2 d x.
(ii) ω2 = x 3 d x + yzdy − (x 2 + y 2 + z 2 )dz.
(iii) ω3 = cos(x y 2 )d x ∧ dz.
(iv) ω4 = xdy ∧ dz + ydz ∧ d x + zd x ∧ dy.
(v) ω1 ∧ ω2 .
(vi) ω1 ∧ ω3 .

Exercise 3.22 Verify (iv) of (3.21) for Exercise 3.21.

Exercise 3.23 If x = r cos θ, y = r sin θ , then compute d x ∧ dy, dr, dθ .

Exercise 3.24 From Exercise 3.11(ii), compute d x ∧ dy ∧ dz.

Exercise 3.25 Consider the following forms in R3 and verify the property (iii), (iv)
of (3.21) where
(i) ω = x 2 d x − z 2 dy and μ = yd x − xdz.
(ii) ω = x yd x + 3dy − yzdz and μ = xd x − yz 2 dy + 2xdz.

Answers

3.20. (i) 2yz 2 d x + 2x z 2 dy + 4x yz dz.


(ii) 2x d x + 2y dy − 12z 3 dz.
(iii) (2x y − z 2 )d x + (x 2 + 2yz)dy + (y 2 − 2zx)d x.
3.21. (i) 4x yd x ∧ dy (ii) −3ydy ∧ dz + 2xdz ∧ d x (iii) 2x y sin(x y 2 )d x ∧ dy ∧
dz.
(iv) 3d x ∧ dy ∧ dz (v) {−4x y(x 2 + y 2 + z 2 ) − 2x 3 y + 3x y 3 }d x ∧ dy ∧ dz
(vi) 0. 
xd x + ydy xdy − yd x y
3.23. r dr ∧ dθ, dr =  , dθ =  , r= x 2 + y 2 , θ = tan .
x 2 + y2 x 2 + y2 x
3.24. r 2 sin θ dr ∧ dθ ∧ dφ.

Remark 3.9 It is to be noted that if φ is a 0-form, then the 1-form dφ corresponds


to grad φ. The exterior derivative of a function f corresponds to the gradient vector
field.
Verification: In 3-dimension, for a function f ,
 
∂f ∂f ∂f ∂f ∂f ∂f
df = dx + dy + dz ↔ ∇ f = , , .
∂x ∂y ∂z ∂ x ∂ y ∂z
3.3 Exterior Differentiation 155

x 2 + y3
Example: Let φ = , then
z

2x 3y 2 x 2 + y3
dφ = dx + dy − dz
z z z2
2x  3y 2  x 2 + y 3 
grad φ = i+ j− k.
z z z2

Remark 3.10 If the 1-form ω corresponds to the vector field V , then the 2-form
dω corresponds to curl V .
Verification: In 3-dimension, let ω = f 1 d x 1 + f 2 d x 2 + f 3 d x 3 where f i , i =
1, 2, 3 are functions. Then

dω = d f 1 ∧ d x 1 + d f 2 ∧ d x 2 + d f 3 ∧ d x 3
3
∂ f1 3
∂ f2 3
∂ f3
= dxi ∧ dx1 + dxi ∧ dx2 + dxi ∧ dx3
∂x i ∂x i ∂xi
i=1 i=1 i=1
     
∂ f2 ∂ f1 ∂ f3 ∂ f2 ∂ f3 ∂ f1
= − d x 1
∧ d x 2
+ − d x 2
∧ d x 3
+ − dx1 ∧ dx3
∂x1 ∂x2 ∂x2 ∂x3 ∂x1 ∂x3
≡ ∇ X f (Vector Product)

which is basically the curl(area).

EXAMPLE:

Let ω = (x 2 + y 3 z)d x + (y 2 − 2x z)dy + (x 4 + y 3 − z 2 )dz in R3 . Then


dω = (−2z − 3y 2 z) d x ∧ dy + (3y 2 + 2x) dy ∧ dz + (y 3 − 4x 3 ) dz ∧ d x.
Again if V = (x 2 + y 3 z)i + (y 2 − 2x z) j + (x 4 + y 3 − z 2 )k is a vector field, then
curl V = (3y 2 + 2x)i + (y 3 − 4x 3 ) j + (−2z − 3y 2 z)k.


Remark 3.11 If the 2-form ω corresponds to the vector field V , then the 3–form
dω corresponds to the div V .
Verification: In 3-dimension, let ω be a 2-form, where

ω = f 12 d x 1 ∧ d x 2 + f 13 d x 1 ∧ d x 3 + f 23 d x 2 ∧ d x 3 , f 12 , f 13 , f 23 being smooth functions. Then



3
∂ f 12 i 
3
∂ f 13 i 
3
∂ f 23 i
dω = dx ∧ dx1 ∧ dx2 + dx ∧ dx1 ∧ dx3 + dx ∧ dx1 ∧ dx3
∂x i ∂x i ∂xi
i=1 i=1 i=1
∂ f 12 3 ∂f ∂f
= d x ∧ d x 1 ∧ d x 2 + 132 d x 2 ∧ d x 1 ∧ d x 3 + 231 d x 1 ∧ d x 2 ∧ d x 3
∂x3 ∂x ∂x
 
∂ f 12 ∂ f 13 ∂ f 23
= − + dx ∧ dx ∧ dx
1 2 3
∂x3 ∂x2 ∂x1

which is the divergence operator.


156 3 Differential Forms

EXAMPLE:

Let ω = (x 2 + y 3 + z 4 ) dy ∧ dz + x 2 y 3 z 4 dz ∧ d x + (x + 2y + 3z + 1) d x ∧ dy.
Therefore
dw = 2xd x ∧ dy ∧ dz + 3x 2 y 2 z 4 dy ∧ dz ∧ d x + 3dz ∧ d x ∧ dy
= (2x + 3x 2 y 2 z 4 + 3) d x ∧ dy ∧ dz.
If V = (x 2 + y 3 + z 4 )i + (x 2 y 3 z 4 ) j + (x + 2y + 3z + 1)k,
 then
div V = 2x + 3x 2 y 2 z 4 + 3.

A form ω is closed if
dω = 0. (3.23)

However, if ω is a r -form and


dμ = ω (3.24)

for some (r − 1)-form μ, then ω is said to be an exact form.

Problem 3.25 Test whether ω is closed or not, where


 
1 2
(i) ω = x yd x + x − y dy.
2
(ii) ω = e x cos yd x + e x sin ydy.

Solution:
1
(i) Here dω = xdy ∧ d x + 2xd x ∧ dy = −xd x ∧ dy + xd x ∧ dy = 0.
2
Thus ω is closed.
(ii) Note that dω = −e x sin y dy ∧ d x + e x sin y d x ∧ dy = 2e x sin y d x ∧ dy.
Thus ω is not closed.

Problem 3.26 The necessary and sufficient condition that a 1-form ω is a gradient
of a function f is that its curl vanishes.

Solution: If f is a function, then grad f = d f [refer to Remark 3.9]. Let us assume


that 1-form ω be grad f i.e. ω = grad f = d f . By virtue of Remark 3.10, we find

curl V = dω = d(d f ) = 0 [refer to (3.21)],

where the 1-form ω corresponds to the vector field V .


For the converse part, suppose for a 1-form ω, curl V = 0. Taking help of
Remark 3.10, we obtain dω = 0. Then d 2 f = 0 holds, for some function f . Thus
ω = grad f . This completes the proof.

Note that all exact forms are closed but the converse is not always true. The
following lemma ensures the converse.
3.3 Exterior Differentiation 157

Lemma 3.1 (Poincaré Lemma) Let ω be a k-form defined on a set

Brn = {(x 1 , . . . , x n ) ∈ Rn |(x 1 )2 + (x 2 )2 + . . . + (x n )2 ≤ r },

such that dω = 0. Thus ∃ a (k − 1)-form μ defined on Brn such that ω = dμ.

Problem 3.27 Let ω = yz d x + x z dy + x y dz. Find μ such that dμ = ω.

Solution: Set ω = dμ = μx d x + μ y dy + μz dz, then


 
μx d x = yzd x = yzx + C1 (y, z), where C1 is a function of y, z.

Then

x z = μy = (yzx + C1 (y, z)) = zx + C1 (y, z)
∂y
⇒ C1 (y, z) = 0
⇒ C1 (y, z) = constant = C2 (z)say.

Then μ = yzx + C2 (z). Now


x y = μz = (yzx + C2 (z)) = yx + C2 (z)
∂z
⇒ C2 (z) = 0
⇒ C2 (z) = constant = Csay.

Finally, μ = yzx + C.
Alternative
Set f (x) = yz d x = yzx + C = μ(say), where C being integration constant, so
that dμ = d(yzx + C) = zx dy + yx dz + yz d x = ω.
Alternative
Set g(y) = x z dy = x zy + C = θ (say), where C being integration constant, so
that dθ = ω.
Alternative
Set h(z) = x y dz = x yz + C = φ(say), where C being integration constant, so
that dφ = ω.

Problem 3.28 Let ω = (12x 2 y 3 + 2y)d x ∧ dy. Find μ such that dμ = ω.


158 3 Differential Forms

Solution: Set f (x, y) = (12x 2 y 3 + 2y)d x = 4x 3 y 3 + 2yx + C = μ(say), where


C being integration constant, so that

dμ = d(4x 3 y 3 + 2yx + C) ∧ dy, by (3.22)


= (12x 2 y 3 d x + 12x 3 y 2 dy + 2xdy + 2yd x) ∧ dy
= (12x 2 y 3 + 2y)d x ∧ dy, as dy ∧ dy = 0
= ω(say).

Problem 3.29 Compute the exterior derivative of the 2-form

1
ω= (x dy ∧ dz + y dz ∧ d x + z d x ∧ dy),
(x 2 + y 2 + z 2 )3/2

defined on R3 \ {(0, 0, 0)}, where (x, y, z) ∈ R3 .


Find the local expression of this form in terms of the spherical coordinates
(r, θ, φ), where x = r sin θ cos φ, y = r sin θ sin φ, z = r cos θ .

Solution: Note that


 
1
dω = d (x dy ∧ dz + y dz ∧ d x + z d x ∧ dy)
(x 2 + y 2 + z 2 )3/2
   
x y
=d ∧ dy ∧ dz + d ∧ dz ∧ d x
(x 2 + y 2 + z 2 )3/2 (x 2 + y 2 + z 2 )3/2
 
z
+d ∧ d x ∧ dy, by (3.22).
(x 2 + y 2 + z 2 )3/2

Now
 
x
d = d(x(x 2 + y 2 + z 2 )−3/2 )
(x 2 + y2+ z 2 )3/2
3
= d x(x 2 + y 2 + z 2 )−3/2 − x(x 2 + y 2 + z 2 )−5/2 (2x d x + 2y dy + 2z dz)
2
dx 3x
= − 2 .
(x 2 + y 2 + z 2 )3/2 (x + y 2 + z 2 )5/2

So, the first term is:

d x ∧ dy ∧ dz 3x 2 d x ∧ dy ∧ dz y 2 + z 2 − 2x 2
− = d x ∧ dy ∧ dz.
(x 2 + y 2 + z 2 )3/2 (x 2 + y 2 + z 2 )5/2 (x 2 + y 2 + z 2 )5/2
3.3 Exterior Differentiation 159

z 2 + x 2 − 2y 2
Similarly, the second term is d x ∧ dy ∧ dz and finally, the third
(x 2 + y 2 + z 2 )5/2
x + y − 2z
2 2 2
term is given by 2 d x ∧ dy ∧ dz. Hence dω = 0. Thus ω is closed.
(x + y 2 + z 2 )5/2
Again
x = r sin θ cos φ, y = r sin θ sin φ, z = r cos θ.

Therefore

d x = sin θ cos φ dr + r cos φ cos θ dθ − r sin θ sin φ dφ,


dy = sin θ sin φ dr + r sin φ cos θ dθ + r sin θ cos φ dφ,
dz = cos θ dr − r sin θ dθ.

After a few steps, one gets

xdy ∧ dz = −r 2 sin θ sin φ cos φ dr ∧ dθ + r 2 sin2 θ cos θ cos2 φ dφ ∧ dr


− r 2 sin3 θ cos2 φ dφ ∧ dθ
ydz ∧ d x = r 2 sin θ sin φ cos φ dr ∧ dθ − r 2 sin2 θ sin2 φ cos θ dr ∧ dφ
+ r 2 sin3 θ sin2 φ dθ ∧ dφ
zd x ∧ dy = r 2 sin2 θ cos θ dr ∧ dφ + r 3 sin θ cos2 θ dθ ∧ dφ,

where r 3 = (x 2 + y 2 + z 2 )3/2 . Thus the local expression of the given form is ω =


sin θ
dθ ∧ dφ.
r
Problem 3.30 Consider the vector fields on R2

∂ ∂ ∂
X = x1 + 2x2 , Y = x1 x2 ,
∂ x1 ∂ x2 ∂ x2

and let ω be a 1-form on R2 defined by

ω = (x1 + x22 ) d x1 + (x12 + x2 ) d x2 .

Show that ω satisfies the relation

dω(X, Y ) = X (ω(Y )) − Y (ω(X )) − ω([X, Y ]).

Solution: Here
160 3 Differential Forms

ω(X ) = x1 (x1 + x22 ) + 2x2 (x12 + x2 ) = x12 + x1 x22 + 2x12 x2 + 2x22 .


ω(Y ) = x1 x2 (x12 + x2 ) = x13 x2 + x1 x22 .
X (ω(Y )) = 5x13 x2 + 5x1 x22 .
Y (ω(X )) = 2x12 x22 + 2x13 x2 + 4x1 x22 .
 
∂ ∂ ∂
[X, Y ] = x1 + 2x2 , x1 x2 ,
∂ x1 ∂ x2 ∂ x2

= x1 x2 , after few steps.
∂ x2
Again ω([X, Y ]) = x13 x2 + x1 x22 .
∴ X (ω(Y )) − Y (ω(X )) − ω([X, Y ]) = 2x13 x2 − 2x12 x22 .
Now dω = 2x2 d x2 ∧ d x1 + 2x1 d x1 ∧ d x2 ,
= (2x1 − 2x2 ) d x1 ∧ d x2 .
 
∂ ∂ ∂
∴ dω(X, Y ) = (2x1 − 2x2 ) d x1 ∧ d x2 x1 + 2x2 , x1 x2
∂ x1 ∂ x2 ∂ x2
    
∂ ∂ ∂
= (2x1 − 2x2 ) d x1 x1 + 2x2 d x2 x1 x2
∂ x1 ∂ x2 ∂ x2
 
 ∂ ∂ ∂
− d x1 x1 x2 d x2 x1 + 2x2
∂ x2 ∂ x1 ∂ x2
= 2x13 x2 − 2x12 x22 .

Thus
dω(X, Y ) = X (ω(Y )) − Y (ω(X )) − ω([X, Y ]).

Exercises

Exercise 3.26 Test which of the following differential forms are closed and which
are exact:
(i) ω1 = x2 x3 d x1 + x1 x3 dy + x1 x2 d x3
(ii) ω2 = x2 d x2 + x12 x22 dy + x2 x3 d x3
(iii) ω3 = 2x y 2 d x ∧ dy + z dy ∧ dz
1
(iv) ω4 = 2 (−yd x + xdy) on R2 \ {(0, 0)}.
x + y2
Exercise 3.27 Show that the following forms are exact:
(i) ω = y 2 d x + 2x ydy
(ii) ω = (3y 2 − 4z 4 )d x + 6x ydy − 16x z 3 dz
In each case, find a function f such that d f = ω.

Exercise 3.28 Show that the following 2-forms are exact:


(i) ω = 24x 3 y 2 d x ∧ dy
(ii) ω = (6x 2 y − 3x y 2 ) d x ∧ dy.
In each case find a 1-form μ such that dμ = ω.
3.3 Exterior Differentiation 161

Exercise 3.29 Let ω = zd x ∧ dy − 2x f (x)dy ∧ dz + y f (x)dz ∧ d x. Find f


where
(i) dω = d x ∧ dy ∧ dz and
(ii) d(dω) = 0.

Answers

3.26. (i) closed and exact. (ii) not closed, not exact (iii) closed and exact (iv)
closed.
3.27. (i) x y 2 (ii) 3y 2 x − 4z 4 x
3.28. (i) 6x 4 y 2 dy (ii) x y 3 d x + 2x 3 y dy.
constant
3.29. f (x) = √ .
x
Remark 3.12 Given a closed k-form ω,  a (k − 1)-form μ such that ω = dμ i.e.
not every closed form is exact. However such an μ always exist locally and this
result is known as the Poincaré Lemma.
xdy − yd x
A well known example is given by the 1-form ω = on M = R2 \
x 2 + y2
{(0, 0)}, where dω = 0. If γ : [0, 2π ] → M is defined by γ (t) = (cos t, sin t), then

ω = 2π,
γ

which is different from zero and thus  a function defined on all of M whose differ-
ential coincides with ω.
Remark 3.13 The difference between closed and exact form is measured by de
Rham Cohomology. The precise definition or computation, of this, is far beyond
the scope of this book.
Theorem 3.4 If ω is a 1-form, then

1
dω(X, Y ) = {X (ω(Y )) − Y (ω(X )) − ω([X, Y ])}, ∀ X, Y ∈ χ (M).
2
Proof Without any loss of generality, one may take an 1-form as ω = f dg, ∀ f, g ∈
F(M). Therefore by virtue of (3.22), we obtain dω = d f ∧ dg. Hence

dω(X, Y ) = (d f ∧ dg)(X, Y )
1
= {d f (X )dg(Y ) − (d f )(Y )dg(X )} by (3.17)
2
1
= {(X f )(Y g) − (Y f )(X g)} by (3.5).
2
Using Leibnitz Product Rule, we see that X ( f (Y g)) = (X f )(Y g) + f (X (Y g)).
Thus
162 3 Differential Forms

1
dω(X, Y ) = {X ( f (Y g)) − f (X (Y g)) − Y ( f (X g)) + f (Y (X g))}.
2
Furthermore,

ω(X ) = ( f dg)(X ) = f dg(X ) = f (X g), by (3.5).

Thus the above expression reduces to,

1
dω(X, Y ) = {X (ω(Y )) − f (X (Y g) − Y (X g)) − Y (ω(X ))}
2
1
= {X (ω(Y )) − Y (ω(X )) − f ([X, Y ]g)}, by (2.27)
2
1
= {X (ω(Y )) − Y (ω(X )) − ω([X, Y ])}, from above.
2
This completes the proof.

Exercise

Exercise 3.30 Consider the vector fields X, Y and 1-form ω on R2 as follows:


∂ ∂ ∂
(i) X = x + y 2 , Y = y , ω = (x + y 2 )d x + (y + x 2 )dy;
∂x ∂y ∂y
∂ ∂ ∂
(ii) X = 2 − , Y = ex ω = e x cos yd x + e x sin ydy;
∂x ∂y ∂y
Show that the 1-form ω satisfies Theorem 3.4 in each case.

Theorem 3.5 If ω is a 1-form, then

dω( f X, Y ) = f dω(X, Y ), f ∈ F(M), ∀ X, Y ∈ χ (M).

Proof Using Theorem 3.4, we write

1
dω( f X, Y ) = {( f X )ω(Y ) − Y (ω( f X )) − ω([ f X, Y ])}.
2
By (2.28), we know that ( f X )ω(Y ) = f (X (ω(Y ))).

Y (ω( f X )) = Y ( f ω(X ))
= (Y f )ω(X ) + f (Y (ω(X ))), by Leibnitz Product rule
[ f X, Y ] = f [X, Y ] − (Y f )X.

Thus taking into consideration (3.1), we have


3.3 Exterior Differentiation 163

ω([ f X, Y ]) = ω( f [X, Y ]) − ω((Y f )X )


= f ω([X, Y ]) − (Y f )ω(X ), Y f ∈ F(M).

Consequently

1
dω( f X, Y ) = { f (X (ω(Y ))) − (Y f )ω(X ) − f (Y (ω(X ))) − f (ω([X, Y ])) + (Y f )ω(X )}
2
1
= f {2dω(X, Y )}, refer to Theorem 3.4.
2
∴ dω( f X, Y ) = f dω(X, Y ).

This completes the proof.

EXISTENCE AND UNIQUENESS of Exterior Differentiation:

Without any loss of generality, we may take an r -form as

ω = f i1 i2 ···ir d x i1 ∧ d x i2 ∧ · · · ∧ d x ir , where f i1 i2 ···ir ∈ F(M).

Let us define an R-linear map d : D → D as

dω = d f i1 i2 ···ir ∧ d x i1 ∧ d x i2 ∧ · · · ∧ d x ir .

Then
(i) d(Dr ) ⊂ Dr +1
(ii) if ω is a 0-form, then dω is the total differential of ω
(iii) let μ ∈ Ds . We consider μ = g j1 j2 ··· js d x j1 ∧ d x j2 ∧ · · · ∧ d x js , where
g j1 j2 ··· js ∈ F(M). Then we get

ω ∧ μ = f i1 i2 ···ir g j1 j2 ··· js d x i1 ∧ d x i2 ∧ · · · ∧ d x ir ∧ d x j1 ∧ d x j2 ∧ · · · ∧ d x js .

Using (3.22), we find

d(ω ∧ μ) = d( f i gj ) ∧ d x i1 ∧ d x i2 ∧ · · · ∧ d x ir ∧ d x j1 ∧ d x j2 ∧ · · · ∧ d x js
1 i 2 ···ir 1 j2 ··· js

= gj d fi ∧ d x i1 ∧ d x i2 ∧ · · · ∧ d x ir ∧ d x j1 ∧ d x j2 ∧ · · · ∧ d x js
1 j2 ··· js 1 i 2 ···ir

+ fi dg j ∧ d x i1 ∧ d x i2 ∧ · · · ∧ d x ir ∧ d x j1 ∧ d x j2 ∧ · · · ∧ d x js
1 i 2 ···ir 1 j2 ··· js

= d fi ∧ d x i 1 ∧ d x i 2 ∧ · · · ∧ d x ir ∧ g j d x j1 ∧ d x j2 ∧ · · · ∧ d x js +
1 i 2 ···ir 1 j2 ··· js

+ (−1)r f i d x i1 ∧ d x i2 ∧ · · · ∧ d x ir ∧ dg j ∧ d x j1 ∧ d x j2 ∧ · · · ∧ d x js
1 i 2 ···ir 1 j2 ··· js

= dω ∧ μ + (−1)r ω ∧ dμ.

(iv) Finally, from (3.22) we obtain


164 3 Differential Forms

 ∂f
dω = d x i s ∧ d x i 1 ∧ · · · ∧ d x ir
i
∂ x is
s

 ∂2 f
or, d(dω) = d x i k ∧ d x i s ∧ d x i 1 ∧ · · · ∧ d x ir
i k ,i s
∂ x ik ∂ x is

= 0.

Thus d satisfies (3.21) and hence the existence is established. It is easy to


establish the uniqueness of d. Consequently, there exists a unique exterior dif-
ferentiation on D.

Exercise 3.31 If ω is a 1-form on M, prove that

dω(X, f Y ) = f dω(X, Y ), f ∈ F(M), ∀ X, Y ∈ χ (M).

3.4 Pull-Back Differential Form

Let M be an n-dimensional and N be an m-dimensional manifold and f : M → N


be a differentiable mapping (Fig. 3.1).
Let Tp (M) be the tangent space at p ∈ M and Tp∗ (M) be its dual. Let T f ( p) (N )
and T f∗( p) (N ) be respectively the tangent space and dual space at f ( p) ∈ N .
Let ω be an 1-form on D1 (N ) ≡ T f∗( p) (N ). We define an 1-form on D1 (M) ≡
Tp (M) called the pull-back 1-form at p ∈ M, denoted by f ∗ ω as follows:

{ f ∗ (ω f ( p) )}(X p ) = ω f ( p) { f ∗ (X p )}, ∀ p ∈ M (3.25)

Fig. 3.1 Pull-Back Differential Form


3.4 Pull-Back Differential Form 165

where f ∗ , f ∗ are already defined in §2.12.


Let us write
f ∗ (ω f ( p) ) = ( f ∗ ω) p . (3.26)

Then (3.25) reduces to

( f ∗ ω) p X p = ω f ( p) ( f ∗ X ) f ( p) by (2.35). (3.27)

We write it as
( f ∗ ω)(X ) = ω( f ∗ X ). (3.28)

Problem 3.31 Let f : M n → N m be a C ∞ map between two manifolds. Let ω be a


C ∞ r -form on N . Show that f ∗ ω is also a C ∞ r -form on M.
Solution: Let (U, x 1 , . . . , x n ) and (V, y 1 , . . . , y m ) be two charts on M and N
respectively, where p ∈ U ⊂ M and f ( p) ∈ V ⊂ N . As ω is a C ∞ r -form on N ,
by Theorem 3.2

ω= g j1 j2 ... jm dy j1 ∧ dy j2 ∧ . . . ∧ dy jr ,
j1 , j2 ,..., jr
j1 < j2 <...< jr

where g j1 j2 ... jm are C ∞ functions on N such that g j1 j2 ... jm ∈ F(N ). Then from (3.33),
we have 
f ∗ω = (g j1 j2 ... jr ◦ f ) d f j1 ∧ d f j2 ∧ . . . ∧ d f jr .
j1 , j2 ,..., jr
j1 < j2 <...< jr

By virtue of Theorem 3.3, one gets

 ∂( f j1 , f j2 , . . . , f jr ) i
f ∗ω = (g j1 j2 ... jr ◦ f ) d x 1 ∧ d x i 2 ∧ . . . ∧ d x ir .
i 1 ,i 2 ,...,ir
∂(x i1 , x i2 , . . . , x ir )
i 1 <i 2 <...<ir

∂( f j1 , f j2 , . . . , f jr )
Now (g j1 j2 ... jr ◦ f ) and are all C ∞ functions. Thus, f ∗ ω is also
∂(x i1 , x i2 , . . . , x ir )
a C ∞ r -form on M.
Theorem 3.6 If f is a differentiable mapping from an n-dimensional manifold M
to an m-dimensional manifold N where (x 1 , x 2 , . . . , x n ) and (y 1 , y 2 , . . . , y m ) are
respectively the coordinates at p ∈ M and f ( p) ∈ N , then
n 
 
∂f j
f ∗ (dy j ) f ( p) = (d x i ) p , f j = y j ◦ f, j = 1, 2, 3, . . . , m.
i=1
∂xi p
166 3 Differential Forms

Proof We write

n
f ∗ (dy j ) f ( p) = aij (d x i ) p ,
i=1

where aij ’s are to be determined. Therefore


 

f ∗ (dy j ) f ( p) = akj , by (3.6)
∂xk p
   

or (dy j ) f ( p) f ∗ = akj , by (3.25)
∂xk p
m  s 
 
∂f ∂
or (dy j ) f ( p) = akj , by Theorem 2.6 of Sect. 2.12 and f s = y s ◦ f.
∂ x k p ∂ ys
s=1
∂f j
∴ = akj , by (3.6).
∂xk

Consequently, we have


n
∂f j
f ∗ (dy j ) f ( p) = (d x i ) p , j = 1, 2, 3, . . . , m.
i=1
∂xi

Remark 3.14 From Theorem 3.6, we observe

f ∗ (dy j ) f ( p) = (d f j ) p , j = 1, 2, 3, . . . , m. (3.29)

Remark 3.15 We can write

f ∗ (dy j ) f ( p) = d(y j ◦ f ) p , j = 1, 2, 3, . . . , m. (3.30)

Theorem 3.7 Let f : M n → N m be a C ∞ map. If ω, μ are 1-forms on N , then


(i) f ∗ (ω + μ) = f ∗ ω + f ∗ μ
(ii) f ∗ (gω) = f ∗ (g) f ∗ (ω)
(iii) f ∗ (λω) = λ f ∗ ω, ∀ g ∈ F(N ), λ ∈ R.

Proof (i) Since ω + μ is also a 1-form, in view of (3.25), one gets

{ f ∗ (ω + μ) f ( p) }(X p ) = (ω + μ) f ( p) { f ∗ (X p )}
= ω f ( p) { f ∗ (X p ) + μ f ( p) { f ∗ (X p )}.

Furthermore, applying (3.25) on the right hand side of the foregoing equation,
the result follows immediately.
(ii) Again for g ∈ F(N ), ω ∈ D1 (N ), gω ∈ D1 (N ) and hence from (3.25), we
have
3.4 Pull-Back Differential Form 167

{ f ∗ (gω) f ( p) }(X p ) = (gω) f ( p) { f ∗ (X p )}


= (g ◦ f )( p)ω f ( p) { f ∗ (X p )}, by (3.12)
or { f (gω)} p (X p ) = (g ◦ f )( p)( f ∗ ω) p (X p ), by (3.26), (3.27)

= ( f ∗ g)( p)( f ∗ ω) p (X p ), by (2.32)


f ∗ (gω) = f ∗ (g) f ∗ (ω), ∀ p ∈ M.

(iii) Left to the reader.


Let ω ∈ Dr (N ). Then a pull-back r -form on M, denoted by f ∗ ω, is defined as
follows:

{ f ∗ (ω) f ( p) }{(X 1 ) p , (X 2 ) p , . . . , (X r ) p } = ω f ( p) { f ∗ (X 1 ) p ), f ∗ (X 2 ) p ), . . . , f ∗ (X r ) p )}.


(3.31)
We also write it as

( f ∗ ω)(X 1 , X 2 , . . . , X r ) = ω( f ∗ X 1 , f ∗ X 2 , . . . , f ∗ X r ). (3.32)

For any ω ∈ Dr (N ), by (3.19) we can write



ω= gi1 ,i2 ,...,ir d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir , i 1 < i 2 < · · · < ir .
i 1 ,i 2 ,...,ir

Taking into consideration Theorem 3.7(ii), we have



f ∗ω = f ∗ (gi1 ,i2 ,...,ir ) f ∗ (d x i1 ) ∧ f ∗ (d x i2 ) ∧ · · · ∧ f ∗ (d x ir ). (3.33)

Combining (3.29) and (2.32), we find



f ∗ω = (gi1 ,i2 ,...,ir ◦ f ) d f i1 ∧ d f i2 ∧ d f i3 ∧ · · · ∧ d f ir , i 1 < i 2 < · · · < ir .
i 1 ,i 2 ,...,ir
(3.34)

Theorem 3.8 If f : M → N is a C n m
map and ω, μ are any forms, then
∗ ∗ ∗
(i) f (ω ∧ μ) = f ω ∧ f μ.
(ii) ( f ◦ h)∗ ω = h ∗ ( f ∗ ω), h being a C ∞ function.
Proof
(i) Let ω ∈ Dr (N ) and μ ∈ Dr (N ). Then ω ∧ μ ∈ Dr +s (N ) and hence by (3.32),
we get

{ f ∗ (ω ∧ μ)(X 1 , X 2 , . . . , X r , Y1 , Y2 , . . . , Ys )}
= (ω ∧ μ)( f ∗ X 1 , . . . , f ∗ X r ; f ∗ Y1 , . . . , f ∗ Ys )
1 
= (sgn σ )ω( f ∗ X σ (1) , . . . , f ∗ X σ (r ) )μ( f ∗ Yσ (1) , . . . , f ∗ Yσ (s) ), by (3.14).
(r + s)! σ
168 3 Differential Forms

The result follows immediately.


(ii) Using (3.30) and (2.32), we can write (3.33) as

f ∗ω = (gi ◦ f )d(x i 1 ◦ f ) ∧ d(x i 2 ◦ f ) ∧ · · · ∧ d(x ir ◦ f ).
1 ,i 2 ,...,ir
i

or h ∗ ( f ∗ ω) = {(gi ◦ f ) ◦ h}d((x i 1 ◦ f ) ◦ h) ∧ · · · · · · ∧ d((x ir ◦ f ) ◦ h).
1 ,i 2 ,...,ir
i

Using Theorem 3.7(ii) and (3.19), we obtain


! "
( f ◦ h)∗ gi ,i ,...,ir d x i1 ∧ d x i2 ∧ · · · ∧ d x ir
1 2

= ( f ◦ h)∗ gi ,i ,...,ir ( f ◦ h)∗ d x i1 ∧ ( f ◦ h)∗ d x i2 ∧ · · · ∧ ( f ◦ h)∗ d x ir
1 2
i

= (gi ◦ ( f ◦ h))d(x i1 ◦ ( f ◦ h)) ∧ d(x i2 ◦ ( f ◦ h)) ∧ · · · ∧ d(x ir ◦ ( f ◦ h)), by (3.30).
1 ,i 2 ,...,ir
i

Using the associativity of C ∞ functions, the result follows.

Theorem 3.9 For any form ω, d( f ∗ ω) = f ∗ (dω).

Proof The following cases will arise:


Case (I) ω is a 0-form:
Let ω = h, h being a C ∞ function.

{ f ∗ (dh)}(X ) = dh( f ∗ X ), by (3.25)


= ( f ∗ X )h, by (3.5)
= X (h ◦ f ), by (2.34)
= (d(h ◦ f ))X, by (3.5)
= d( f ∗ h)X, by (2.32).

Thus d( f ∗ ω) = f ∗ (dω) holds for every X .


Case (II) ω is a r -form:
Taking aid of Principle of Mathematical Induction, we assume that the result is true
for (r − 1)-form. Without any loss of generality, any r -form can be expressed as

ω = gi1 ,i2 ,...,ir d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir .

Using Theorem 3.8(i), we have

f ∗ ω = f ∗ (gi1 ,i2 ,...,ir d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 ) ∧ f ∗ (d x ir )


or d( f ∗ ω) = d{ f ∗ (gi1 ,i2 ,...,ir d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 ) ∧ f ∗ (d x ir )}.

In view of (3.21)(iii), we get


3.4 Pull-Back Differential Form 169

d( f ∗ ω) = d{ f ∗ (gi1 ,i2 ,...,ir d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 )} ∧ f ∗ (d x ir )+


+ (−1)r −1 f ∗ (gi1 ,i2 ,...,ir d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 )d( f ∗ (d x ir )).

Note that d x ir is a 1-form and as the result is true for (r − 1) form,

d( f ∗ (d x ir )) = f ∗ (d(d x ir )) = 0, by (3.21)(iv).

Thus

d( f ∗ ω) = d{ f ∗ (gi d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 )} ∧ f ∗ (d x ir )}
1 ,i 2 ,...,ir

= f ∗ {d(gi d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 )} ∧ f ∗ (d x ir ),
1 ,i 2 ,...,ir

as it is true for (r − 1) − form


= f ∗ {dgi ∧ d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 } ∧ f ∗ (d x ir ), by (3.22)
1 ,i 2 ,...,ir

= f ∗ {dgi ∧ d x i1 ∧ d x i2 ∧ d x i3 ∧ · · · ∧ d x ir −1 ∧ d x ir , refer to Theorem 3.8(i)


1 ,i 2 ,...,ir

= f ∗ (dω), by (3.22).

Thus the result is true for r -form. Hence, we can claim that for any form ω

d( f ∗ ω) = f ∗ (dω).

Problem 3.32 Let g = x 2 − y 2 , ω = x yd x + xdy. Let f : R → R2 be given by


f (t) = (t, t 2 ) = (x, y). Find A. (i) f ∗ g (ii) f ∗ ω. B. Show that f ∗ (dg) = d( f ∗ g).

Solution:
A. (i) f ∗ g = g ◦ f = t 2 − t 4 .
(ii) Here

f ∗ ω = f ∗ (x yd x + xdy)
= f ∗ (x yd x) + f ∗ (xdy), by linearity
= f ∗ (x y) f ∗ (d x) + f ∗ (x) f ∗ (dy), by Theorem 3.7(ii)
= (x y ◦ f )d( f ∗ x) + (x ◦ f )d( f ∗ y), by (2.32)
= (t ◦ t 2 )d(x ◦ f ) + td(y ◦ f ), by (3.30)
= t 3 d(t) + td(t 2 )
= t 3 dt + t (2tdt)
= (t 3 + 2t 2 )dt.

ALTERNATIVE METHOD: f ∗ ω = f ∗ (x yd x + xdy)=(t ◦ t 2 )dt + td(t 2 ) =


(t 3 + 2t 2 )dt.
170 3 Differential Forms

B. Note that

d( f ∗ g) = d(t 2 − t 4 ) = (2t − 4t 3 )dt,


dg = d(x 2 − y 2 ) = 2xd x − 2ydy
f ∗ (dg) = f ∗ (2xd x − 2ydy) = 2tdt − 2t 2 (2tdt) = (2t − 4t 3 )dt.
Thus f ∗ (dg) = d( f ∗ g).

Problem 3.33 Let U be the open set (0, ∞) × (0, 2π ) in the (ρ, θ )-plane R 2 . Define
f : U ⊂ R2 → R2 by f (ρ, θ ) = (ρ cos θ, ρ sin θ ). If x, y are the standard coordi-
nates on R2 , compute the pull-back f ∗ (d x ∧ dy).

Solution: We know that

f ∗ (d x ∧ dy) = f ∗ d x ∧ f ∗ dy, where


f ∗ d x = d(ρ cos θ ) = cos θ dρ − ρ sin θ dθ, and
f ∗ dy = d(ρ sin θ ) = sin θ dρ + ρ cos θ dθ.

∴ f (d x ∧ dy) = (cos θ dρ − ρ sin θ dθ ) ∧ (sin θ dρ + ρ cos θ dθ )
= ρ cos2 θ dρ ∧ dθ + ρ sin2 θ dρ ∧ dθ
= ρdρ ∧ dθ.

Problem 3.34 Consider a map f : U ⊂ R4 → R2 given by f (x1 , x2 , x3 , x4 ) =


(u, v) where

u = x12 + x22 + x32 + x42 − 1, v = x12 + x22 + x32 + x42 − 2x2 − 2x3 + 5.

∗ ∗
Calculate f (−1,5) (du + 2dv) ∈ T(−1,5) (R2 ), taking (0, 0, 0, 0) at f −1 (−1, 5).

Solution: Consider x1 = x2 = x3 = x4 = 0, u = −1, v = 5. Now

f ∗ (du + 2dv) = d(x12 + x22 + x32 + x42 − 1) + 2d(x12 + x22 + x32 + x42 − 2x2 − 2x3 + 5)
= 6x1 d x1 + 6x2 d x2 + 6x3 d x3 + 6x4 d x4 − 4d x2 − 4d x3 .

∴ f (−1,5) (du + 2dv) = (−4d x2 − 4d x3 )(0,0,0,0)
= −4(d x2 + d x3 )(0,0,0,0) .

Problem 3.35 Let M be a circle and N be R2 so that the map f : M → N be


1 1
defined by x 1 = r cos θ, x 2 = r sin θ . If ω = ad x 1 + bd x 2 and μ = d x 1 + d x 2 ,
a b
compute f ∗ (ω ∧ μ), where a, b are constants.
3.4 Pull-Back Differential Form 171

Solution: From Theorem 3.8(i), we have

f ∗ (ω ∧ μ) = f ∗ ω ∧ f ∗ μ.

Now

f ∗ ω = f ∗ (ad x 1 + bd x 2 ) = a f ∗ (d x 1 ) + b f ∗ (d x 2 )
= ad( f ∗ x 1 ) + bd( f ∗ x 2 ), by Theorem 3.8
= ad(x 1 ◦ f ) + bd(x 2 ◦ f )
= ad(r cos θ ) + b(r sin θ )
= a(cos θ dr − r sin θ dθ ) + b(sin θ dr + r cos θ dθ )
= (a cos θ + b sin θ ) dr + (br cos θ − ar sin θ ) dθ

Similarly
   
∗ 1 1 r r
f μ= cos θ + sin θ dr + cos θ − sin θ dθ.
a b b a

 
r r
∴ f ∗ (ω ∧ μ) = {(a cos θ + b sin θ ) cos θ − sin θ dr ∧ dθ }
b a
 
1 1
+ {(br cos θ − ar sin θ ) cos θ + sin θ } dθ ∧ dr
a b
a b
=r − dr ∧ dθ, after a few steps.
b a

Problem 3.36 Let U be the open set (0, ∞) × (0, π ) × (0, 2π ) in the (r, φ, θ )-
space R3 . Let f : U → R3 be defined by

f (r, φ, θ ) = (r sin φ cos θ, r sin φ sin θ, r cos φ).

If (x, y, z) are the standard coordinates of R3 , show that

f ∗ (d x ∧ dy ∧ dz) = r 2 sin φ dr ∧ dφ ∧ dθ.

Solution: It is known that

f ∗ (d x ∧ dy ∧ dz) = f ∗ d x ∧ f ∗ dy ∧ f ∗ dz.

Now
172 3 Differential Forms

f ∗ d x = d( f ∗ x) = d(x ◦ f ) = d f 1 = d(r sin φ cos θ )


= dr sin φ cos θ + r cos φ cos θ dφ − r sin φ sin θ dθ.

Similarly

f ∗ dy = sin φ sin θ dr + r cos φ sin θ dφ + r sin φ cos θ dθ,


f ∗ dz = cos φ dr − r sin φ dφ.

Therefore

f ∗ (d x ∧ d y ∧ dz) = {(sin φ cos φ sin θ cos θ) dr ∧ dφ + (r sin2 φ cos2 θ) dr ∧ dθ


+ (r sin φ cos φ sin θ cos θ) dφ ∧ dr + (r 2 sin φ cos φ cos2 θ) dφ ∧ dθ
− (r sin2 φ sin2 θ) dθ ∧ dr − (r 2 sin φ cos φ sin2 θ) dθ ∧ dφ}
∧ (cos φ dr − r sin φ dφ)
= (−r 2 sin3 φ cos2 θ) dr ∧ dθ ∧ dφ + (r 2 sin φ cos2 φ cos2 θ) dφ ∧ dθ ∧ dr
+ (r 2 sin3 φ sin2 θ) dθ ∧ dr ∧ dφ − (r 2 sin φ cos2 φ sin2 θ) dθ ∧ dφ ∧ dr
= r 2 sin φdr ∧ dφ ∧ dθ.

Exercises

Exercise 3.32 Let h = 2x y, μ = −2yd x + xdy, θ = (x 2 + y 2 )dy. Let f : R →


R2 be given by f (t) = (t, t 2 ) = (x, y).
A. Find (i) f ∗ h (ii) f ∗ μ (iii) f ∗ θ .
B. Show that f ∗ (dh) = d( f ∗ h).

Exercise 3.33 Let f : R2 → R2 be given by

f (u, v) = (u 2 + 1, uv) = (x, y), ω1 = (x y − y)d x ∧ dy, μ1 = (x + y 2 )d x ∧ dy.

Find (i) f ∗ h (ii) f ∗ μ (iii) f ∗ θ (iv) f ∗ ω1 (v) f ∗ μ1 , where h, μ, θ are defined in


Exercise 3.32.

Exercise 3.34 Let μ = −2yd x + xdy, θ = (x 2 + y 2 )dy. Let f : R2 → R2 be


given by f (u, v) = (u 2 + 1, uv) = (x, y). Show that
(i) f ∗ (μ ∧ θ ) = f ∗ μ ∧ f ∗ θ .
(ii) f ∗ (ω ∧ θ ) = f ∗ ω ∧ f ∗ θ , where ω = x yd x + xdy.

Exercise 3.35 If f : M → R3 be such that f (u, v) = (u cos v, u sin v, av), then


for a given 1-form ω = x 1 d x 1 − d x 2 + x 2 d x 3 on R3 , where x 1 = u cos v, x 2 =
u sin v, x 3 = av, compute f ∗ ω.
3.4 Pull-Back Differential Form 173

Exercise 3.36 If f : M → R3 be such that f (u, v) = (a cos u sin v, a sin u sin v,


a cos v), then for a given 1-form ω = d x 1 + ad x 2 + ad x 3 on R3 , determine f ∗ ω.
y x
Exercise 3.37 Let ω = − dx + 2 be the 1-form in R2 \ {(0, 0)}. Let
+y x22 x + y2
U = {r > 0 : 0 < θ < 2π } be the set in the plane (r, θ ) and f : U → R2 be the map
x = r cos θ
f (r, θ ) = . Compute f ∗ ω.
y = r sin θ

Exercise 3.38 Let S 1 be a unit circle and f : R → S 1 ⊂ R2 be given by f (θ ) =


(x1 , x2 ) = (cos θ, sin θ ). If ω is the 1-form −x2 d x1 + x1 d x2 on S 1 , compute f ∗ ω.

Exercise 3.39 Consider on R2 , θ = (2x y + x 2 + 1)d x + (x 2 − y)dy and f be the


map f : R3 → R2 given by (u, v, w) → (x, y) = (u − v, v 2 + w). Compute f ∗ θ .

Exercise 3.40 Let f : R2 → R2 be given by f (x1 , x2 ) = (u, v), where u = x12 +


x22 , v = x1 x2 . Calculate
(i) f ∗ (udu + vdv) and
∗ ∗
(ii) f (2,1) (udu + vdv) ∈ T(2,1) (R2 ), taking (1, 1) to f −1 (2, 1).

Answers

3.32. (i) 2t 3 (ii) 0 (iii) (2t 3 + 2t 5 )dt


3.33. (i) 2u v + 2uv
3
(ii) (−3u 2 v + v)du + (u 3 + u)dv
(iii) (u v + 2u v + v + u 2 v 3 )du + (u 5 + 2u 3 + u + u 3 v 2 )dv
4 2

(iv) 2u 5 v du ∧ dv (v) (2u 4 + 2u 2 + 2u 4 v 2 )du ∧ dv


3.35. (u cos v − sin v)du + (au sin v − u cos v − u 2 sin v cos v)dv
2

3.36. (−a sin u sin v + a 2 sin v cos u)du + (a cos u cos v + a 2 sin u cos v −
a sin v)dv
3.37. dθ 3.38. dθ
3.39. {2(u − v)(v 2 + ω) + (u − v)2 + 1}du + (u 2 − 2uv − ω)dω
+(−4uv − 2uω + 2v 3 − u 2 + 2uv − v 2 − 1 + 2u 2 v)dv.
3.40. (i) (2x13 + 3x1 x22 )d x + (2x23 + 3x12 x2 )d x2 . (ii) 5(d x1 + d x2 )(1,1) .
Chapter 4
Lie Group

4.1 Lie Group, Left and Right Translation

A Lie group is a group (in algebraic sense), which is also a differentiable manifold,
with the property that the group operations are compatible with the smooth structure.
Thus, a Lie group G consists of two structures on the same set G, namely, it is a
differentiable manifold and has also a group structure. We now state the formal
definition as follows:
Let G be a non-empty set. If


⎪ (i) G is a group (whose operation is denoted by multiplication),


⎨ (ii) G is an n − dimensional smooth manifold and
(iii) the inverse map τ : G → G such that τ (x) = x −1 and



⎪ the multiplication map φ : G × G → G such that φ(x, y) = x y, ∀ x, y

are smooth maps, then G is called an n-dimensional Lie Group.
(4.1)

Remark 4.1 The group is called “Lie Group”, after the Norwegian mathematician
Sophus Lie (1842–1899).

Remark 4.2 (a) The product of two second countable and Hausdorff spaces is
respectively the second countable and Hausdorff space. [for details refer to any
standard textbook of topology]
(b) If M and N are C ∞ manifolds, then M × N with its product topology is Haus-
dorff and second countable. To show that M × N is a smooth manifold, it remains
to find a C ∞ atlas. If we have a chart (U, φU ) and (V, φV ) respectively on M and
N , then U × V ⊆ M × N an open subset of the product space. Let us define

φU ×V = φU × φV : U × V → Rm × Rn = Rm+n .

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 175
M. Majumdar and A. Bhattacharyya, An Introduction to Smooth Manifolds,
https://doi.org/10.1007/978-981-99-0565-2_4
176 4 Lie Group

If (Ũ , φŨ ) and (Ṽ , φṼ ) are another pair of charts for M and N , respectively, then
we can set the transition function

φŨ ×Ṽ ◦ φU−1


×V
= (φŨ × φṼ ) ◦ (φU × φV )−1 = (φŨ ◦ φU−1 ) × (φṼ ◦ φV−1 ).

Since each is a transition function from one of the two smooth atlases as already
known on M and N , therefore, each of these factors is smooth. Since smoothness
is determined componentwise, it follows that the product mapping is smooth as
well. So we have an atlas making M × N a smooth manifold. It should also be
clear that its dimension is m + n, as asserted.

Remark 4.3 Let G be a group, which is also a differentiable manifold of dimension


n. It is easy to check that the map (x y) → x y −1 is a smooth map from 2n-dimensional
manifold G × G to n-dimensional smooth manifold G. Thus, G is a Lie group.

Remark 4.4 Note that τ 2 (x) = τ (τ (x)) = τ (x −1 ) = x. Therefore, τ 2 = I is


differentiable. Thus, G possesses τ 2 = I as diffeomorphism. Moreover, it possesses
two other diffeomorphisms, viz. “Left Translation”and “Right Translation”.

Example 4.1 Note that the n-dimensional space Rn is a differentiable manifold of


n-dimension and is a group with respect to addition, defined as

x + y = (x 1 + y 1 , . . . , x n + y n )
−x = (−x 1 , −x 2 , . . . , −x n )
where, x = (x 1 , x 2 , . . . , x n ), y = (y 1 , y 2 , . . . , y n ).

Furthermore, the operations φ(x + y) = x + y, τ (x) = −x are C ∞ -functions.


Hence, Rn is a Lie group of dimension n.

Problem 4.1 G L(n, R) is a Lie group.

Solution : In reference to Problem 2.17, we have already proved that G L(n, R)(⊂
M(n, R) forms a smooth manifold of dimension n 2 . Note that G L(n, R) forms a
group under usual matrix multiplication. It only remains to prove that G L(n, R) is
a Lie group.
For that, we define φ : G L(n, R) × G L(n, R) → G L(n, R) by φ(A, B) = AB,
so that
φ(A, B) = AB = [φ ij (AB)]i, j=1,2,...,n ,

where φ ij (AB) is the i j-th element of the matrix AB. Then, φ ij (AB) =
n
k=1 x k (A)x j (B), where x j (A) denotes the i j-th coordinate function on G L(n, R).
i k i

Thus, for all i, j; φ ij is a smooth function. Moreover, the inverse map τ : G L(n, R) →
G L(n, R) defined by τ (A) = A−1 is also smooth. Hence G L(n, R) is a Lie Group
of dimension n 2 .
4.1 Lie Group, Left and Right Translation 177

For every a ∈ G, a mapping L a : G → G defined by

L a (x) = ax, ∀ x ∈ G (4.2)

is called a left translation on G. Similarly, a mapping Ra : G → G defined by

Ra (x) = xa, ∀ x ∈ G (4.3)

is called a right translation on G.

Exercises

Exercise 4.1 Show that

L a L b = L ab , Ra Rb = Rba , L a Rb = Rb L a . (4.4)
L a L b = L b L a , unless G is commutative. (4.5)
L a −1 = (L a )−1 , Ra −1 = (Ra )−1 . (4.6)

It is to be noted that each L a , Ra are C ∞ maps, as each L a , L a −1 , Ra , Ra −1 is


homeomorphism and differentiable from G onto G.

Examples

Example 4.2 The left translation L a : G L(n, R) → G L(n, R) is

x = (x ij ) → ax = (aki x kj )

and the right translation Ra : G L(n, R) → G L(n, R) is

x = (x ij ) → xa = (xki a kj ).

Example 4.3 The Lie group R n is a commutative group. Hence, for every a ∈ Rn ,
L a = Ra . Also, the group operation is addition and the identity element is 0. So left
translation is actually the parallel translation x → x + a i.e.L a x = x + a = a + x.
⎧⎛ ⎞ ⎫
⎨ 1x y ⎬
Problem 4.2 Let H = ⎝ 0 1 z ⎠ : x, y, z ∈ R . Show that H admits a Lie group
⎩ ⎭
001
structure with usual matrix multiplication. Such H is called Heisenberg Group.
⎛ ⎞
1x y
Solution : Let us define the map φ : H → R3 as ⎝ 0 1 z ⎠ → (x, y, z). Note that
001
the map φ is homeomorphic. Thus, {(H, φ)} forms an C ∞ -atlas for H . If
178 4 Lie Group
⎛ ⎞ ⎛ ⎞
1x y 1x y
A = ⎝ 0 1 z ⎠ , B = ⎝ 0 1 z ⎠ , then AB ∈ H.
001 0 0 1
⎛ ⎞
1 −x −y
Now A−1 = ⎝ 0 1 −z ⎠ ∈ H, ∀ A ∈ H . Again H is a group with respect to
0 0 1
matrix multiplication. Moreover, let us define the maps

 : H × H → H by (A, B) → AB,

and the map


ψ : H → H by A → A−1 .

Then the map φ ◦  ◦ (φ −1 × φ −1 ) : R3 × R3 → R3 defined by

(φ ◦  ◦ (φ −1 × φ −1 ))((x, y, z), (a, b, c)) = (a + x, b + y, c + z)

and the map φ ◦ ψ ◦ φ −1 : R3 → R3 given by

(φ ◦ ψ ◦ φ −1 )(x, y, z) = (−x, −y, −z)

are C ∞ . Thus H is a Lie group.

Problem 4.3 Let φ : G 1 → G 2 be a homomorphism of a Lie group G 1 to another Lie


group G 2 . Show that (i) φ ◦ L a = L φ(a) ◦ φ (ii) φ ◦ Rb = Rφ(b) ◦ φ for all a, b ∈ G 1 .

Solution : (i) From the definition of group homomorphism φ : G 1 → G 2 given by

φ(ab) = φ(a)φ(b), ∀ a, b ∈ G 1 .

Now

(φ ◦ L a )(x) = φ(L a (x)) = φ(ax)


= φ(a)φ(x)
= (L φ(a) ◦ φ)(x).
∴ φ ◦ L a = L φ(a) ◦ φ, ∀ x ∈ G 1 .

(ii) Note that


(φ ◦ Rb )(x) = φ(xb) = φ(x)φ(b) = (Rφ(b) ◦ φ)(x).

Therefore, φ ◦ Rb = Rφ(b) ◦ φ, ∀ x ∈ G1 .

Problem 4.4 Let φ be a non-identity 1 − 1 map from G onto G. If φ ◦ L g = L g ◦ φ


holds for all g ∈ G, then there exists h ∈ G such that φ = Rh .
4.2 Invariant Vector Field 179

Solution : As φ is a 1 − 1 map, for every a ∈ G, there exists an unique element


b ∈ G such that φ(a) = b. Again e ∈ G, φ is a 1 − 1 map, there exists an unique
element, say h ∈ G, such that φ(e) = h, where φ(e) = e, as φ is not an identity
mapping. Now

g = ge, ∀ g ∈ G
∴ φ(g) = φ(ge) = φ(L g (e)) = (φ ◦ L g )(e) = (L g ◦ φ)(e), as given
= L g (φ(e))
= L g (h) = gh = Rh g
⇒ φ = Rh .

Exercises

Exercise 4.2 Show that the set of all left(right) translations on a Lie group form a
group.
  
α0
Exercise 4.3 Let G = : α > 0, β ∈ R . Prove that G admits a Lie group
β1
structure with matrix multiplication.
  
x y
Exercise 4.4 Let G = : x, y ∈ R, x = 0 . Prove that G admits a Lie
01
group structure, with matrix multiplication.
⎧⎛ ⎞ ⎫
⎨ x1 0 x2 ⎬
Exercise 4.5 Let G = ⎝ 0 x1 x3 ⎠ : x1 , x2 , x3 ∈ R, x1 > 0 . Prove that
⎩ ⎭
0 0 1
G admits a Lie group structure with respect to usual matrix multiplication.

Exercise 4.6 Let ψ be the diffeomorphism of G defined by ψ(x) = x −1 . Show that


ψ ◦ L g = Rg−1 ◦ ψ, ψ ◦ Rg = L g−1 ◦ ψ.

Exercise 4.7 Prove that R is an abelian Lie group where the smooth maps φ :
R × R → R is defined by
φ(a, b) = a + b

and τ : R → R is defined by
τ (a) = −a.

4.2 Invariant Vector Field

We have already defined an invariant vector field under a transformation in §2.14.


As each left translation and right translation on a Lie group G are transformations,
we can similarly define the invariant vector fields on G.
180 4 Lie Group

A vector field X on a Lie group G is called a left invariant vector field on G if

(L a )∗ X p = X L a ( p) = X ap , ∀ p ∈ G, (4.7)

where (L a )∗ is the differential of the left translation L a , for some fixed a in G.


Using (2.35), we write

((L a )∗ X ) L a ( p) = X L a ( p) i.e.
(L a )∗ X = X. (4.8)

Similarly, for a right invariant vector field

(Ra )∗ X = X. (4.9)

Again, for every f ∈ F(G), by virtue of (2.34), we have

{(L a )∗ X p } f = X p ( f ◦ L a )
 
i.e. (L a )∗ X L a ( p) f = X p ( f ◦ L a ) by (2.35).

If L a ( p) = q, then p = a −1 q (refer to (4.6)). Thus

{(L a )∗ X }q f = X a −1 q ( f ◦ L a ). (4.10)

Theorem 4.1 A vector field X on a Lie group is left invariant if and only if

(X f ) ◦ L a = X ( f ◦ L a ), ∀ f ∈ F(G) (4.11)

where for some fixed a ∈ G, L a is the left translation of G.

Proof Let X be a left invariant vector field of a Lie group G. Then from (4.7), we
find

{(L a )∗ X p } f = X L a ( p) f, ∀ f ∈ F(G)
= (X f )L a ( p), by (2.23)
or X p ( f ◦ L a ) = (X f )L a ( p), by (2.24)
 
or, {X ( f ◦ L a )}( p) = (X f ) ◦ L a ( p), by (2.23).
∴ X ( f ◦ L a ) = (X f ) ◦ L a , ∀ p ∈ G.

The converse follows immediately.

Problem 4.5 Find the left invariant vector fields on Rn .


4.2 Invariant Vector Field 181

Solution : Let p ∈ Rn and p = (x 1 , x 2 , . . . , x n ). Any vector field X on the manifold


Rn can be expressed uniquely as
 ∂
X= ξi , where ξ i ∈ F(Rn ), i = 1, 2, 3, . . . , n.
∂xi
Furthermore
 ∂f 
(X f ) ◦ L a = ξi ◦ L a , a ∈ Rn
i
∂xi
  ∂f  
∴ {(X f ) ◦ L a }( p) = ξi ◦ L a ( p)
i
∂x i

 ∂f 
= ξi L a ( p)
i
∂xi
 ∂f 
= ξi (a + p), see Example 4.3
i
∂xi
 ∂f
or {(X f ) ◦ L a }( p) = ξ i (a + p) (a + p).
i
∂xi

Also
   i ∂ 
X ( f ◦ L a ) ( p) = ξ ( f ◦ L a ) ( p)
i
∂x i

 ∂
= ξ i ( p) i ( f ◦ L a )( p)
i
∂x
 ∂f
= ξ i ( p) i (L a ( p))
i
∂x
   ∂f
or X ( f ◦ L a ) ( p) = ξ i ( p) i (a + p).
i
∂ x

In view of (4.11), we find


 ∂f  ∂f
ξ i (a + p) (a + p) = ξ i ( p) i (a + p)
i
∂x i
i
∂x
or ξ i (a + p) = ξ i ( p), i = 1, 2, 3, . . . , n.

Thus, the functions ξ i ’s are constants and hence all the left invariant vector fields on
Rn are of the form

ξ i i , ξ i ∈ R, i = 1, 2, 3, . . . , n,
∂x
182 4 Lie Group


i.e.constant multiple of or the left invariant vector fields on Rn are constant
∂xi
vector fields.

Problem 4.6 If X, Y are left invariant vector fields on a Lie group G, so is [X, Y ].

Solution : From (4.8), we have (L a )∗ X = X and (L a )∗ Y = Y . Now

(L a )∗ [X, Y ] = [(L a )∗ X, (L a )∗ Y ] (see Exercise 2.30)


= [X, Y ], from above.

Thus, [X, Y ] is also left invariant vector field on G.


ALTERNATIVE METHOD: Taking into consideration (2.34), for every vector field
[X, Y ] in χ (G), we have
 
(L a )∗ [X, Y ] f = [X, Y ]( f ◦ L a )
= X (Y ( f ◦ L a )) − Y (X ( f ◦ L a )), by (2.27)
     
= X (L a )∗ Y f − Y (L a )∗ X f
= X {Y f } − Y {X f }, use (2.34)
= [X, Y ] f
∴ (L a )∗ [X, Y ] = [X, Y ], ∀ f ∈ F(G).

We say (G, [ , ]) is a Lie algebra over R if




⎪ (i) G is a vector space over R

(ii) [, ] : G × G → G is a bilinear map
(4.12)

⎪ (iii) [X, Y ] = −[Y, X ] : anti-commutative

(iv) [X, [Y, Z ]] + [Y, [Z , X ]] + [Z , [X, Y ]] = θ : Jacobi Identity.

Problem 4.7 Show that the vector space R3 with the operation cross product of
vectors is a Lie algebra.

Solution : Let x = (x 1 , x 2 , x 3 ) and y = (y 1 , y 2 , y 3 ) be any two vectors of R3 . We


define the cross product of x and y as follows:

x × y = (x 2 y 3 − x 3 y 2 , x 3 y 1 − x 1 y 3 , x 1 y 2 − x 2 y 1 ).

Then, for all λ, μ ∈ R, it can be shown that



(i) (λ x + μy ) × z = λ
x × z + μy × z
: Bilinearit y
(ii) z × (λ
x + μy ) = λz × x + μz × y

(iii) x × y = −y × x
(iv) (x × y) × z + (y × z ) × x + (z × x) × y = 0.
4.2 Invariant Vector Field 183

Thus, the real vector space R3 with the operation:Bilinearity of the cross product of
vectors is a Lie algebra.

Remark 4.5 Note that the set of all C ∞ -vector fields, denoted by χ (M), of the
manifold M, forms a Lie algebra under the Lie bracket operation on vector fields.

Let g be the set of all left invariant vector fields on G. Then for every X, Y in g,

(L a )∗ (cX + dY ) = c(L a )∗ X + d(L a )∗ Y = cX + dY, ∀ c, d ∈ R.

Therefore, cX + dY ∈ g. Hence, g is a linear space.


Again, (L a )∗ [X, Y ] = [(L a )∗ X, (L a )∗ Y ] = [X, Y ] (refer to Exercise 2.30). Thus,
[X, Y ] ∈ g where [X, Y ] = −[Y, X ]. Further, it can be shown that

[X, [Y, Z ]] + [Y, [Z , X ]] + [Z , [X, Y ]] = θ.

Thus, the set of all left invariant vector fields, denoted by g, on a Lie group G, is
a Lie algebra.
Now a Lie subalgebra h 1 of a Lie algebra h 2 is a vector subspace h 1 ⊂ h 2 , that is
closed under the bracket [, ].
Obviously, g is a Lie subalgebra of the Lie algebra χ (G) of all C ∞ vector
fields of the Lie group G i.e.g ⊂ χ (G)

Remark 4.6 If g∗ denotes the set of all right invariant vector fields on a Lie group
G, it can be shown that g∗ is also a Lie algebra.

The behaviour of a Lie group is determined largely by its behaviour in the


neighbourhood of the identity element e. The behaviour can be represented by an
algebraic structure on the tangent space at e.
Theorem 4.2 As a vector space, the Lie subalgebra g of the Lie group G is iso-
morphic to the tangent space Te (G) at the identity element e of G i.e.g ∼
= Te (G)
(Fig. 4.1).

Fig. 4.1 g ∼
= Te (G)
184 4 Lie Group

Proof Note that two vector spaces U and V are said to be isomorphic if a mapping
f : U → V is linear and has an inverse f −1 . So, let us define a mapping φ : g →
Te (G) by
φ(X ) = X e . (4.13)

Clearly, φ is linear. Let us define φ −1 : Te (G) → g by

φ −1 (Ye ) = X. (4.14)

Now for every a ∈ G, L a : G → G is left translation and

(L a )∗ : Te (G) → TL (a) (e)=a (G)

is a differential mapping such that

(L a )∗ Ye = X a . (4.15)

Now for any s ∈ G, we have

(L s )∗ X s −1 a = (L s )∗ (L s −1 a )∗ Ye , by (4.15)
= (L s ◦ L s −1 a )∗ Ye , by Problem 2.51
= (L a )∗ Ye , by (4.4)
 
or, (L s )∗ X a = X a , by (2.35), (4.15)
or (L s )∗ X = X, ∀ a ∈ G.

Therefore, X ∈ g. Hence the mapping φ −1 is well-defined. Finally

(φφ −1 )Ye = X e , by (4.13) & (4.14)


= (L e )∗ Ye , by (4.15)
= Ye , where (L e )∗ is the identity differential.
 
(φ φ)(X ) = φ −1 (L e )∗ Ye = X, by (4.13), (4.14) & (4.15).
−1

Thus, an inverse mapping φ −1 exists. Consequently,

g∼
= Te (G).

Corollary 4.1 If a Lie group G is of dimension n, then the dimension of Lie subal-
gebra g of the Lie group G is also n.
Proof Left to the reader.
Problem 4.8 If φ : g → Te (G) is an isomorphism, g being the set of all left invariant
vector fields of a Lie group G, then for X̃ = φ(X ) = X e , X ∈ g, show that [ X̃ , Ỹ ] =

[X, Y ], ∀ Y ∈ g.
4.2 Invariant Vector Field 185

Solution : From the definition

[ X̃ , Ỹ ] = X̃ (Ỹ ) − Ỹ ( X̃ )
= X e (Ye ) − Ye (X e ), as defined
= [X e , Ye ].

Further


[X, Y ] = {X (Y
) − Y (X )}
=X (Y ) − Y (X )
= X e (Ye ) − Ye (X e )
= [X e , Ye ].


Thus, [ X̃ , Ỹ ] = [X, Y ].

Problem 4.9 If Cikj (i, j, k = 1, 2, 3, . . . , n) are structure constants of a Lie group


G with respect to the basis {X 1 , X 2 , . . . , X n } of g, show that

(i) Cikj = −C kji , where [X i , X j ] = Cikj X k , each Cikj ∈ R.
(ii) Cikj Cks
t
+ C kjs Ckit + Csik Ckt j = 0.

Solution : It is given that Cikj (i, j, k = 1, 2, 3, . . . , n) are structure constants of a


Lie group G with respect to the basis {X 1 , X 2 , . . . , X n } of g, where g is the set of all
left invariant vector fields on G, such that


n
[X i , X j ] = Cikj X k , Cikj ∈ R. (4.16)
k=1

Note that [X i , X j ] = θ , when i = j. So let i = j. As [X i , X j ] = −[X j , X i ], by virtue


of (4.16), we get


n 
n
Cikj X k = − C kji X k
k=1 k=1

or (Cikj + C kji ) = 0,

as {X 1 , X 2 , . . . , X n } is a basis and hence linearly independent. Thus, Cikj = −C kji .


Also, from the Jacobi identity, we have

[[X i , X j ], X s ] + [[X j , X s ], X i ] + [[X s , X i ], X j ] = θ.

Using (4.16), we get


186 4 Lie Group


n 
n 
n
[ Cikj X k , X s ] + [ C kjs X k , X i ] + [ Csik X k , X j ] = θ
k=1 k=1 k=1

n 
n 
n
or Cikj [X k , X s ] + C kjs [X k , X i ] + Csik [X k , X j ] = θ
k=1 k=1 k=1
  
Cikj Cks
t
Xt + C kjs Ckit X t + or Csik Ckt j X t = θ, by (4.16).
t k t k t k

As {X t : t = 1, 2, 3, . . . , n} is a basis of g and hence linearly independent and thus

Cikj Cks
t
+ C kjs Ckit + Csik Ckt j = 0.

Problem 4.10 Consider the product T 1 × R+ of the one-dimensional torus by the


multiplicative group of positive numbers. Let (α̃, x1 ) denote the local coordinates.
∂ ∂
Prove that the vector field X = + x1 is left invariant.
∂ α̃ ∂ x1

Solution : For a fixed (θ̃ , p) ∈ T 1 × R+ , the left translation of the product mani-
fold, denoted by L (θ̃ , p) , is by definition

L (θ̃ , p) (α̃, x1 ) = (θ̃ + α̃, px1 )


⎛ ⎞
∂(θ̃ +α̃) ∂(θ̃ +α̃)  
10
∴ (L (θ̃ , p) )∗ = ⎝ ∂ α̃ ∂ x1
∂( px1 ) ∂( px1 )
⎠=
0 p
∂ α̃ ∂ x1

Given vector field X ∈ T 1 × R+ is left invariant if

(L (θ̃ , p) )∗ X (0,1) = X L (0,1) = X (θ̃ , p) .


(θ̃ , p)

 
∂ ∂ ∂ ∂  ∂ ∂ 
Now X = + x1 , so X (0,1) = + and X (θ̃ , p) = +p .
∂ α̃ ∂ x1 ∂ α̃ ∂ x1 (0,1) ∂ α̃ ∂ x1 (θ̃ , p)
Again
     
10 1 1 ∂ ∂ 
(L (θ̃ , p) )∗ X (0,1) = = = +p = X (θ̃ , p) .
0 p 1 p ∂ α̃ ∂ x1 (θ̃ , p)

Thus, X is a left invariant vector field.

Problem 4.11 Consider the Lie group G defined in Exercise 4.5. Show that
 ∂ ∂ ∂ 
(i) X = x1 , Y = x1 , Z = x1 is a basis of left invariant vector fields
∂ x1 ∂ x2 ∂ x3
of G.
4.2 Invariant Vector Field 187

(ii) Find the structure constants, as defined by (4.16), of G with respect to the basis
{X, Y, Z } defined in (i) above.

Solution : (i) For a fixed (a1 , a2 , a3 ) ∈ G. the left translation denoted by L (a1 ,a2 ,a3 )
is given by
⎛ ⎞⎛ ⎞ ⎛ ⎞
a1 0 a2 x1 0 x2 a1 x 1 0 a1 x 2 + a2
L (a1 ,a2 ,a3 ) (x1 , x2 , x3 ) = ⎝ 0 a1 a3 ⎠ ⎝ 0 x1 x3 ⎠ = ⎝ 0 a1 x1 a1 x3 + a3 ⎠
0 0 1 0 0 1 0 0 1
= (a1 x1 , a1 x2 + a2 , a1 x3 + a3 )
⎛ ⎞
a1 0 0
∴ (L (a1 ,a2 ,a3 ) )∗ = ⎝ 0 a1 0 ⎠ .
0 0 a1


Let e = (1, 0, 0) denote the identity element of G. Then X = x1 is given by
∂ x1


X e = X (1,0,0) = .
∂ x1

 ∂   ∂   ∂   ∂ 
Similarly, Ye = x1 = and Z e = x1 = . Now
∂ x2 e ∂ x2 e ∂ x3 e ∂ x3 e

∂ 
X L (a = X (a1 ,a2 ,a3 ) = x1 .
∂ x1 (a
(1,0,0)
1 ,a2 ,a3 )
1 ,a2 ,a3 )

Similarly

∂  ∂
YL (a = Y(a1 ,a2 ,a3 ) = x1 = a1 , and
∂ x2 (a ,b,a )
(1,0,0)
1 ,a2 ,a3 ) ∂ x2
 1 3

∂  ∂
Z L (a = Y(a1 ,a2 ,a3 ) = x1 = a1 .
∂ x3 (a ,b,a )
(1,0,0)
1 ,a2 ,a3 ) ∂ x3
1 3

To show X, Y, Z are left invariant vector fields, we have to show

(L (a1 ,a2 ,a3 ) )∗ X e = X L (a (1,0,0) = X (a1 ·1,a1 ·0+a2 ,a1 ·0+a3 ) = X (a1 ,a2 ,a3 ) .
1 ,a2 ,a3 )

Similarly

(L (a1 ,a2 ,a3 ) )∗ Ye = Y(a1 ,a2 ,a3 ) and (L (a1 ,a2 ,a3 ) )∗ Z e = Z (a1 ,a2 ,a3 ) .
188 4 Lie Group

Now
⎛ ⎞⎛ ⎞ ⎛ ⎞
a1 0 0 1 a1

(L (a1 ,a2 ,a3 ) )∗ X e = ⎝ 0 a1 0 ⎠ ⎝ 0 ⎠ = ⎝ 0 ⎠ = a1 = X (a1 ,a2 ,a3 ) .
0 0 a1 0 0 ∂ x1

Thus, X is a left invariant vector field. Again


⎛ ⎞⎛ ⎞ ⎛ ⎞
a1 0 0 0 0

(L (a1 ,a2 ,a3 ) )∗ Ye = ⎝ 0 a1 0 ⎠ ⎝ 1 ⎠ = ⎝ a1 ⎠ = a1 = Y(a1 ,a2 ,a3 ) ,
0 0 a1 0 0 ∂ x2
⎛ ⎞⎛ ⎞ ⎛ ⎞
a1 0 0 0 0

(L (a1 ,a2 ,a3 ) )∗ Z e = ⎝ 0 a1 0 ⎠ ⎝ 0 ⎠ = ⎝ 0 ⎠ = a1 = Z (a1 ,a2 ,a3 ) .
0 0 a1 1 a1 ∂ x3

Thus, Y and Z are left invariant vector fields and {X, Y, Z } are linearly indepen-
dent at e. Thus, {X, Y, Z } is a basis of g, where g is a left invariant vector fields
of G.
∂ ∂ ∂
(ii) If we denote X = x1 by X 1 , Y = x1 by X 2 and Z = x1 by X 3 , then
∂ x1 ∂ x2 ∂ x3
by (4.13), we see that


3
[X i , X j ] = Cikj X k , i = j.
k=1

Now
   
∂ ∂ ∂ ∂ ∂ ∂
[X 1 , X 2 ] = [x1 ,x ] = x1 x − x1 x
∂ x1 1 ∂ x2 ∂ x1 1 ∂ x2 ∂ x2 1 ∂ x1
∂ ∂2 ∂2
= x1 + x12 − x12
∂ x2 ∂ x1 ∂ x2 ∂ x1 ∂ x2

= x1 .
∂ x2
∴ 2
C12 X 2 = X 2 , by above

2
i.e. C12 = 1, as X 2 ≡ Y is linearly independent. By Problem 4.9(i), C21 2
=
−C12 = −1. Similarly, it can be shown that [X 1 , X 3 ] = X 3 , i.e.C13 = 1 and
2 3
3
C31 = −1 and [X 2 , X 3 ] = 0. Thus, with respect to the basis {X, Y, Z }, the struc-
ture given by
2
C12 = −C21 2
= C133
= −C31 3
= 1.
4.3 Invariant Differential Form 189

Exercises
Exercise 4.8 Prove the converse of Theorem 4.1.
Exercise 4.9 Find the left invariant vector fields on R.
Exercise 4.10 If e is the identity element of a Lie group G and Te (G) is the tangent
space to G at e, show that
(L a )∗ X e = X a ,

where X is a left invariant vector field.


Exercise 4.11 Consider the Problem 4.2. Show that
∂ ∂ ∂ ∂
(i) {X = ,Y = ,Z =x + } is a basis of left invariant fields of H .
∂x ∂y ∂y ∂z
(ii) Find the structure constants of H with respect to the basis {X, Y, Z }.
Answer
4.11. C132 = −C312 = 1.

4.3 Invariant Differential Form

A differential form ω on a Lie group G is said to be left invariant if


 
L a∗ ωL a ( p) = ω p , ∀ p ∈ G (4.17)

where L a∗ is the pull-back of L a defined in §3.4.


We write it as
L a∗ ω = ω. (4.18)

Similarly, a differential form ω on a Lie group G is said to be right invariant if

Ra∗ ω = ω. (4.19)

A differential form, which is both left and right invariant is said to be bi-invariant
differential form.
Problem 4.12 Show that if ω is a left invariant form, then dω is also so.
Solution : From Theorem 3.9, we see that

d(L a∗ ω) = L a∗ (dω), ∀ ω
or, dω = L a∗ (dω), by (4.18).

Hence, dω is also left invariant by (4.18).


190 4 Lie Group

Problem 4.13 Prove that a 1-form ω on a Lie group G is left invariant if and only
if for every left invariant vector field X on G, ω(X ) is a constant function on G.

Solution : For every given 1-form ω on G, L a∗ ω will be pull-back 1-form. Hence,


by (3.27), we find
   
L a∗ (ωL a ( p) ) (X p ) = ωL a ( p) (L a )∗ X p , ∀ p ∈ G, ∀ X ∈ g,

g being the set of all invariant vector fields, (L a )∗ X p being the differential of X .
Consequently, by (4.7), we write
 
L a∗ (ωL a ( p) ) (X p ) = ωL a ( p) (X L a ( p) ). (4.20)

Let us now consider ω to be left invariant. Then in view of (4.17), one gets from above

ω p (X p ) = ωap (X ap ).

Taking p = e ∈ G, one gets the desired result, i.e.ω(X ) is a constant function.


For the converse part, let ω(X ) be a constant function on G. Then for fixed a ∈ G
and arbitrary p ∈ G, one must have

ω p (X p ) = ωap (X ap )
= ωL a ( p) (X L a ( p) )
  
= L a∗ ω L a ( p) (X p ), see (4.20)
∴ L a∗ (ωL a ( p) ) = ω p , ∀ X p ∈ g.

Thus, by (4.17), the 1-form ω is left invariant.

Problem 4.14 Prove that the set of all left invariant forms on a Lie group G forms
an algebra over R.

Solution : Let A be the set of all left invariant forms on a Lie group G. We wish to
show that:
(i)A is a linear space over R.
(ii)the mapping A × A → A, defined as (ω, μ) → ω ∧ μ is bilinear.
(iii)the
 operation ‘∧’ is skew-symmetric.
ω ∧ (μ + γ ) = ω ∧ μ + ω ∧ γ
(iv)
(ω + μ) ∧ γ = ω ∧ γ + μ ∧ γ .
Note that

L a∗ (cω + dμ) = c L a∗ ω + d L a∗ μ, as L a∗ is linear, c, d ∈ R


= cω + dμ, where ω, μ are left invariant.

Thus, A is a linear space over R.


4.3 Invariant Differential Form 191

Further, it can be shown that

bω ∧ μ = b(ω ∧ μ) = ω ∧ bμ, ∀ b ∈ R.

Thus, the mapping is bilinear. Also, ω ∧ μ = −μ ∧ ω. Moreover, the set A satisfies


(iv). Thus, A is an algebra over R.
Problem 4.15 Let L a : S 1 → S 1 be given by

L a (x, y) = {(cos t)x − (sin t)y, (sin t)x + (cos t)y}

where a = (cos t, sin t) ∈ S 1 ⊂ R2 . If ω = −yd x + xdy on S 1 , show that (L a )∗ ω =


ω,
S 1 being the unit circle.
Solution : Here

L a∗ ω = L a∗ (−yd x + xdy)
= L a∗ (−yd x) + L a∗ (xdy)
= −{(sin t)x + (cos t)y}d{(cos t)x − (sin t)y}
+ {(cos t)x − (sin t)y}d{(sin t)x + (cos t)y}
= −{(sin t)x + (cos t)y}{(cos t)d x − (sin t)dy}
+ {(cos t)x − (sin t)y}{(sin t)d x + (cos t)dy}
= xdy − yd x
= ω.

Thus, ω is left invariant 1-form on S 1 .


Theorem 4.3 If g is Lie subalgebra of a Lie group G and g∗ denotes the set of all
left invariant forms on G, then

1
dω(X, Y ) = − ω([X, Y ]), ∀ X, Y ∈ g, ω ∈ g∗ .
2
Proof From Theorem 3.4, we see that if ω is a 1-form, then

1
dω(X, Y ) = {X (ω(Y )) − Y (ω(X )) − ω([X, Y ])}, ∀ X, Y ∈ χ (G).
2
Now if X, Y ∈ g, ω ∈ g∗ , then by Problem 4.13, ω(X ), ω(Y ) are constant functions
on G. Taking help of Exercise 2.30, we see that

X (ω(Y )) = 0 = Y (ω(X )).

1
Thus dω(X, Y ) = − ω([X, Y ]).
2
192 4 Lie Group

Remark 4.7 Such an equation is called Maurer–Cartan Equation.


1 i
Problem 4.16 Show that dωi = − C ω j ∧ ωk , where C ijk ’s are defined in
2 j,k jk
Exercise 4.9(i).

Solution : If {X 1 , X 2 , . . . , X n } is a basis of g and {ω1 , ω2 , . . . , ωn } is the dual basis


of g∗ , then
ωi (X j ) = δ ij . (4.21)

Hence, by virtue of the last theorem, we obtain

1
dωi (X j , X k ) = − ωi ([X j , X k ])
2
1  m
= − ωi { C jk X m }, see (4.16)
2
1 
=− C m ωi (X m ), as C mjk ∈ R
2 m jk
1
dωi (X j , X k ) = − C ijk . (4.22)
2
Now
 1 i
i
Cmn (ωm ∧ ωn )(X j , X k ) = C {ωm (X j )ωn (X k ) − ωm (X k )ωn (X j )}, by (3.17)
m,n
2 m,n mn
1 i
= C (δ m δ n − δkm δ nj ), by (4.21)
2 m,n mn j k
1 i
= (C − Cki j )
2 jk
= C ijk , see Problem 4.9(i)

Thus, (4.22) reduces to

1 i
dωi (X j , X k ) = − C (ωm ∧ ωn )(X j , X k )
2 m,n mn
1 i
or dωi = − C (ωm ∧ ωn ), ∀ X j , X k
2 m,n mn
1 i
i.e. dωi = − C (ω j ∧ ωk ).
2 j,k jk
4.3 Invariant Differential Form 193


i
Problem 4.17 Prove that dωi = C ijk ωk ∧ ω j .
j,k
j<k

Solution : Let us consider j, k = 1, 2, 3. Then



C ijk ω j ∧ ωk = C12
i
ω1 ∧ ω2 + C13
i
ω1 ∧ ω3 + C21
i
ω2 ∧ ω1
j,k=1,2,3

+ C23
i
ω2 ∧ ω3 + C31
i
ω3 ∧ ω1 + C32
i
ω3 ∧ ω2
= 2C12
i
ω1 ∧ ω2 + 2C13
i
ω1 ∧ ω3 + 2C23
i
ω2 ∧ ω3 ,

as ωi ∧ ωi = 0, ωi ∧ ω j = −ω j ∧ ωi , C ijk = −Cki j . Thus

 
i
C ijk ω j ∧ ωk = 2 C ijk ω j ∧ ωk .
j,k=1,2,3 j,k
j<k

Hence, from Problem 4.16, we have

1  i
dωi = − × 2 C ijk ω j ∧ ωk
2 j,k
j<k


i
or dωi = C ijk ωk ∧ ω j .
j,k
j<k

  
x1 x2 
Problem 4.18 Consider G = x1 , x2 ∈ R, x1 = 0 which is a Lie sub-
0 1
group of G L(2, R).
 
x1 x2
(i) Show that ω = a −1 da is a left invariant 1-form, where a = .
0 1
dx dx
(ii) Show that {ω1 = 1 , ω2 = 2 } is a basis of left invariant 1-form g ∗ and find
x1 x1
the structure constants of G with respect to {ω1 , ω2 }.
(iii) Show that dω + ω ∧ ω = 0.
     
x1 x2 1 −x2 −1 1 1 −x2
Solution : (i) Here a = , |a|=x1 , adj a = ,a = .
0 1 0 x1 x1 0 x1
194 4 Lie Group
    
d x1 d x2 1 1 −x2 d x1 d x2
Also, da = . Thus, ω = a −1 da =
0 0 x1 0 x1 0 0
 
1 d x1 d x2
= .
x1 0 0
   
10 r s
Let us choose e = and arbitrary q = . Thus,
01 01
      
r s x1 x2 r x1 r x2 + s r 0
Lq a = = , (L q )∗ = .
01 0 1 0 1 0r
   
1 d x1 d x2 d x1 d x2
Now, ω p = , ωe = . We are to show (L p )∗ ωL p e = ωe ,
p 0 0 0 0
i.e. (L p )∗ ω p = ωe . Now
  1            
p 0 p 0 0 1 0 d x1 d x2
(L p )∗ ω p = p , 1 = , ≡ = ωe .
0 p 0 0 p p 0 1 0 0

Thus, ω = a −1 da is a left invariant 1-form.


(ii) Note that

d x1 1
dω1 = d( )=d ∧ d x1 = 0
x1 x1
dx 1 1
dω2 = d( 2 ) = d ∧ d x2 − − 2 d x1 ∧ d x2
x1 x1 x1
= −ω1 ∧ ω2 .

Taking advantage of Maurer–Cartan Equation, we obtain



dωi = − C ijk ω j ∧ ωk ,
j,k

for i = 2, dω2= −C12 ω ∧ ω2 . Comparing, we find that C12


2 1 2
= 1 = −C21
2
.
1 d x1 d x2
(iii) Here ω = . Therefore
x1 0 0
     
1 d x1 d x2 1 d x1 d x2 1 0 d x1 ∧ d x2
ω∧ω = ∧ = .
x1 0 0 x1 0 0 x12 0 0
4.4 Automorphism 195

Further
 
1 d x1 d x2
dω = d ∧
x1 0 0
 
1 d x1 d x2
= − 2 d x1 ∧
x1 0 0
 
1 0 d x1 ∧ d x2
=− 2
x1 0 0
= −ω ∧ ω, from above.

Thus, dω + ω ∧ ω = 0.
Exercises
Exercise 4.12 If ω1 , ω2 are left invariant differential forms, prove that ω1 ∧ ω2 is
also so.
Exercise 4.13 Prove that a r -form ω on a Lie group G is left invariant if and only
if for every left invariant vector fields X i ’s(1 ≤ i ≤ r ) on G, ω(X 1 , X 2 , . . . , X r ) is
a constant function on G.
Exercise 4.14 Let φ : G → G be such that φ(a) = a −1 , ∀ a ∈ G. Show that a form
ω is left invariant if and only if φ ∗ ω is right invariant.
Exercise 4.15 If g∗ denotes the dual space of g, prove that A ∼
= g∗ , where the set A
is defined in the solution of Problem 4.14.

4.4 Automorphism

Let G 1 and G 2 be Lie group. A map f : G 1 → G 2 is said to be a Lie group homo-


morphism if f is a C ∞ map and for all h, x in G 1 ,

f (hx) = f (h) f (x). (4.23)

We can also write it as


f ◦ L h = L f (h) ◦ f, ∀ x ∈ G 1 . (4.24)

Let eG 1 and eG 2 be the identity elements of G 1 and G 2 , respectively. Taking h, x in


(4.23) to be the identity eG 1 , it follows that

f (eG 1 ) = eG 2 . (4.25)

So a group homomorphism always maps the identity to identity.


If moreover, f is bijective, then f is said to be a Lie group isomorphism.
196 4 Lie Group

For every a ∈ G, a mapping σa : G → G, defined by

σa (x) = axa −1 (4.26)

is said to be an inner automorphism if



(i) σa is bijective
(4.27)
(ii) σa (x y) = σa (x)σa (y)

Such σa is written as ada.


An inner automorphism of a Lie group G is defined by

(ada)(x) = axa −1 , ∀ x ∈ G. (4.28)

  
x y
Problem 4.19 If the Lie group G is defined by G = : x, y ∈ R, x > 0 ,
0x
 
x y
verify whether the map f : G → R3 defined by f = (x, y, x − y) is a Lie
0x
group homomorphism or not.
      
x y x y x y x x x y + yx
Solution : For ∈ G, we have = . From
0 x 0x 0 x 0 xx
the hypothesis,
  
x y x y
f = (x x , x y + yx , x x − x y − yx ).
0x 0 x

To show that f is a Lie group homomorphism, our claim is

f (AB) = f (A) + f (B), A, B ∈ G, where G is a lie group.

Now       
x y x y x y x y
f + f = f .
0x 0 x 0x 0 x

Thus, f is not a Lie group homomorphism.

Exercises

Exercise 4.16 Show that if G is a Lie group, then the map Ih : G → G for every
h ∈ G defined by Ih (x) = hxh −1 , x ∈ G is an automorphism.

Exercise 4.17 Show that

ada = L a Ra −1 = Ra −1 L a . (4.29)
4.4 Automorphism 197
⎧⎛ ⎞ ⎫
⎨ 1x y ⎬
Exercise 4.18 Let H = ⎝ 0 1 z ⎠ : x, y, z ∈ R be the Lie group and the map
⎩ ⎭
001
f : H → R defined by A → f (A) = x + y + z. Is it a Lie group homomorphism?

Answer

4.18. No.

Let g1 and g2 be Lie subalgebra of the Lie group G. A mapping f : g1 → g2 is said


to be a Lie algebra homomorphism if f is a linear mapping and

f [X, Y ] = [ f X, f Y ]. (4.30)

Moreover, if f is bijective then f is said to be Lie algebra isomorphism.

Problem 4.20 Let G 1 = G L(n, R) and G 2 = G L(1, R). Then the map f given by
f (A) = det A, A ∈ G 1 is a homomorphism.

Solution : Clearly

f (c A + B) = c f (A) + f (B), ∀ c ∈ R, A, B ∈ G 1 ,

which implies f is linear. Furthermore, f (AB) = det(AB) = det A det B =


f (A) f (B) implies f is a homomorphism.

Problem 4.21 If f : G 1 → G 2 is a Lie group homomorphism and X is a left invari-


ant vector field on G 1 , prove that the left invariant vector field f ∗ X on G 2 is f -related
to the left invariant vector field X .

Solution : For h ∈ G 1 , f (h) ∈ G 2 . For every X h ∈ g1 , f ∗ (X h ) ∈ g2 . To show that


f ∗ X is f -related to X , we need to show f ∗ (X h ) = ( f ∗ X ) f (h) , by (2.38). Note that

(L h )∗ X e = X L h (e) = X h , X ∈ g1 , g1 ∼
= Te1 (G 1 ) .

Thus

f ∗ (X h ) = f ∗ ((L h )∗ X e ) = ( f ◦ L h )∗ X e ,
= (L f (h) ◦ f )∗ (X e ), by (4.24)
= (L f (h) )∗ { f ∗ (X e )}
= { f ∗ (X )}L f (h) (e) , by (4.7).
∴ f ∗ (X h ) = { f ∗ (X )} f (h) .

Hence, the left invariant vector field f ∗ X on G 2 is f -related to the left invariant
vector field X on G 1 .
198 4 Lie Group

Theorem 4.4 Let f : G 1 → G 2 be a Lie group homomorphism. Then the induced


map f ∗ : Te (G 1 ) → Te (G 2 ) is a homomorphism between the Lie algebras of the Lie
group, where e, e are respectively the identity elements of G 1 and G 2 .
Proof For the identity element e ∈ G 1 ,

f (e) = e ,

e being the identity element of G 2 . In view of Theorem 4.2, we can write

g1 ∼
= Te (G 1 ), and g2 ∼
= Te (G 2 ),

where g1 , g2 are respectively the Lie algebras of G 1 and G 2 .


Now, let X e ∈ Te (G 1 ) be such that

f ∗ (X e ) = Ye ∈ Te (G 2 ).

Then for any a ∈ G 1 , (L a )∗ X e = X L a (e) = X a . Therefore

f ∗ (X a ) = f ∗ {(L a )∗ X e }
= ( f ◦ L a )∗ X e , as ( f g)∗ = f ∗ ◦ g∗
= (L f (a) ◦ f )∗ X e , by (4.24)
= (L f (a) )∗ Ye
= Y f (a) , refer to the definition of left invariant vector field

Thus, the image of a left invariant vector field on G 1 under f ∗ is a left invariant vector
field on G 2 .
Again, we know that

f ∗ [X 1 , X 2 ] = [ f ∗ X 1 , f ∗ X 2 ]
= [Y1 , Y2 ],

where f ∗ X i = Yi , i = 1, 2. Thus, f ∗ is a homomorphism between the Lie algebras


of two spaces and this completes the proof.
Theorem 4.5 Every inner automorphism of a Lie group G induces an automorphism
of the Lie algebra g of G.
Proof For every a ∈ G, let us denote the inner automorphism on G by

(ada)(x) = axa −1 , ∀ x ∈ G.

Now every ada : G → G induces a differential mapping

(ada)∗ : Te (G) → Tada(e) (G) ≡ Te (G), see (4.28).


4.4 Automorphism 199

Such a mapping is a linear mapping, and from Theorem 4.2, we have g ∼


= Te (G).
Thus, to prove the theorem, we need to prove the following:
(i) (ada)∗ : g → g is a well-defined mapping.
(ii) (ada)∗ is bijective.
(iii) (ada)∗ is homomorphism.
By (4.29), we get
(ada)∗ = (Ra −1 )∗ , ∀ Y ∈ g. (4.31)

Now

(L p )∗ {(Ra −1 )∗ Y } = (L p ◦ Ra −1 )∗ Y, p ∈ G
= (Ra −1 )∗ Y, see (4.29).

Thus, (Ra −1 )∗ Y ∈ g. Consequently, (ada)∗ is a well-defined mapping that proves (i).


Also, (ada)∗ is a linear mapping and by Exercise 2.30, the (iii) follows immediately.
Finally, let (ada)∗ X = (ada)∗ Y . Then from (4.31), it follows that X = Y and
hence (ada)∗ is injective. For surjectivity, let us set (ada −1 )∗ X = Y , as for every
a, a −1 ∈ G. Now

(L s )∗ Y = (L s )∗ ((ada −1 )∗ X )
 
= (L s )∗ ◦ (L a −1 ◦ Ra )∗ X , by (4.29)
= (L s )∗ ◦ ((Ra )∗ X ), by (4.4)
= (Ra )∗ X, by (4.4) and Problem 2.51
= Y, by (4.31) and as assumed.

Thus, Y ∈ g. Also

(ada)∗ Y = (L a ◦ Ra −1 )∗ Y, by (4.29)
= (L a ◦ Ra −1 )∗ (ada −1 )∗ X, as set
= (L a ◦ Ra −1 )∗ (Ra ◦ L a−1 )∗ X, by (4.29)
= X, by Problem 2.51

which proves (ada)∗ is surjective and consequently, (ada)∗ is a bijective mapping.


Thus, the induced map (ada)∗ : g → g is a Lie algebra automorphism.

Remark 4.8 For every a ∈ G, we write

(ada)∗ ≡ Ada, i.e. a → Ada (4.32)

and is called the Adjoint Representation of G.


200 4 Lie Group

Problem 4.22 Is Ada invertible?

Solution : By virtue of the last theorem, we find

Ada ≡ (ada)∗ : g → g

is a Lie algebra homomorphism. Further g ∼ = Te (G), where Te (G) is a finite-


dimensional vector space. Thus, Ada is a linear transformation from Te (G) →
Te (G). Again, for every a ∈ G, we obtain

(Ada) ◦ (Ada −1 ) = (ada)∗ ◦ (ada −1 )∗


= (L a ◦ Ra −1 ) ◦ (Ra ◦ L a −1 )∗ , by (4.29)
= Ade

Similarly, it can be shown that (Ada −1 ) ◦ (Ada) = Ade. Hence, Ada−1 = (Ada)−1 .
This completes the proof.

4.5 One-Parameter Subgroup of a Lie Group

Let G be a Lie group and a mapping a : R → G denoted by a : t → a(t) be a


differentiable curve on G. If for all t, s in R

a(t + s) = a(t)a(s), (4.33)

then the family {a(t)| t ∈ R} is called a one-parameter subgroup of G.


Exercises

Exercise 4.19 Let H = {a(t)|t ∈ R} be a one-parameter subgroup of a Lie group


G. Show that H is a commutative subgroup of G.

Exercise 4.20 If X is a left invariant vector field on a Lie Group G, prove that X is
complete.

Theorem 4.6 Let X be the generator generated by one-parameter group of transfor-


mations Rat and let Y be that of L at . Then X is left invariant and Y is right invariant
and X e = Ye = a (0) hold, where a (t) denotes the tangent vector to the curve a at
a(t).

Proof For h ∈ G, let L h ( p) = q. Therefore, p = h −1 q. Now from (2.34)


4.5 One-Parameter Subgroup of a Lie Group 201

{(L h )∗ X p } f = X p ( f ◦ L h ), ∀ f ∈ F(G)
= X h−1 q ( f ◦ L h )
1
= lim [( f ◦ L h ){Rat (h −1 q)} − ( f ◦ L h )(h −1 q)], by (2.47)
t→0 t

1
= lim {( f ◦ Rat )(q) − f (q)}, by (4.2), (4.4)
t→0 t

= X q f, by (2.47)
∴ (L h )∗ X p = X q , ∀ f
or (L h )∗ X p = X L h ( p) ,

which shows that X is a left invariant vector field on Lie group G (refer to (4.8)).
Similarly, for h ∈ G, let Rh ( p) = q, therefore, p = h −1 q. In view of (2.34), we have

{(Rh )∗ Y p } f = Y p ( f ◦ Rh )∀ f ∈ F(G)
= Yh−1 q ( f ◦ Rh )
1
= lim [( f ◦ Rh ){L at (qh −1 )} − ( f ◦ Rh )(qh −1 )]
t→0 t

1
= lim {( f ◦ L at )(q) − f (q)}, by (4.3), (4.4)
t→0 t

= Yq f
∴ (Rh )∗ Y p = Yq , ∀ f
= Y Rh ( p) .

Thus, by (4.9), we can say that Y is right invariant. Since Rat (e) = at = a(t), a(t)
is an integral curve of X and hence X a(t) = a (t) holds. In particular, X e = a (0). By
similar manner, Ye = a (0).

Theorem 4.7 Let {φt |t ∈ R} be a one-parameter group of transformations on a Lie


group G generated by a left invariant vector field and

φt (e) = a(t) = at .

If for every s ∈ G, φt ◦ L s = L s ◦ φt , for all s ∈ G, then the set {a(t)| t ∈ R} is a


one-parameter subgroup of G and φt = Rat holds for all t ∈ R. If φt ◦ Rs = Rs ◦ φt
holds for all s ∈ G, then the set {a(t)| t ∈ R} is a one-parameter subgroup of G and
φt = L at holds for all t ∈ R

Proof As defined
202 4 Lie Group

a(s + t) = φs+t (e) = φt+s (e), ∀ s, t ∈ R


= φt (φs (e)), as {φt |t ∈ R} is one-parameter group of transformations
= φt a(s)
= φt (L a(s) e)
= (L a(s) ◦ φt )(e), from the hypothesis, as a(s) ∈ G
= L a(s) (a(t))
= a(s)a(t).

So from (4.33), {a(t)| t ∈ R} is a one-parameter subgroup of G. Also

φt (s) = φt (se) = (φt ◦ L s )(e) = (L s ◦ φt )(e) = L s (at ) = sat = Rat (s).

Therefore, φt = Rat . Similarly, it can be shown that φt = L at when φt ◦ Rs = Rs ◦ φt


holds. This completes the proof.

For each X ∈ g, we set

φt (e) = Rat (e) = at = a(t) = exp(t X ), (4.34)

where {φt |t ∈ R} is the one-parameter group of transformations on G generated by


X . We call {at = a(t) = φt (e)} the one-parameter subgroup of G, generated by
X.
The map X → exp X , is a map from g to G and is said to be the exponential
map.
Problem 4.23 Let G be a Lie group. For every X ∈ g, let Y be the generator induced
by the one-parameter group of transformations {φt |t ∈ R} defined by

φ : G → G, φ(t, x) ≡ φt (x) = exp(t X )x, ∀ x ∈ G.

Prove that Y is right invariant.

Solution : We have to prove that, for every a ∈ G, (Ra )∗ Y = Y . In view of Exer-


cise 2.48, we are to prove that (Ra ◦ φt ) = (φt ◦ Ra ). Now

(Ra ◦ φt )(x) = Ra (exp(t X )x), by hypothesis


= exp(t X )Ra x
= exp(t X )xa, by (4.3).

Also, (φt ◦ Ra )(x) = φt (xa) = exp(t X )xa. Thus, (Ra ◦ φt ) = (φt ◦ Ra ) i.e.(Ra )∗ Y =
Y . Hence, the vector field Y is right invariant.

Problem 4.24 Let G be an Abelian Lie group. Prove that [X, Y ] = 0, where X, Y
are left invariant vector fields on G.
4.5 One-Parameter Subgroup of a Lie Group 203

Solution : From Problem 4.23, for X, Y ∈ g, let

φt (x) = exp(t X )x,


ψs (x) = exp(sY )x, t, s ∈ R.

By virtue of Theorem 2.8, we know that if {φt |t ∈ R} is the one-parameter group of


transformations generated by X , then for every vector field Y ,

1
[X, Y ] = lim {Y − (φt )∗ Y }
t→0 t

1
or [X, Y ]q = lim {Yq − ((φt )∗ Y )q }, q ∈ G.
t→0 t

Let us write φt ( p) = q, p ∈ G, then p = φ−t (q) and hence

((φt )∗ Y )q = ((φt )∗ Y )φt ( p) = (φt )∗ Y p [r e f er to(2.35)] = (φt )∗ Yφ−t (q) .

Thus, [X, Y ] = 0 if ((φt )∗ Y )q = Yq i.e.(φt )∗ Yφ−t (q) = Yq , i.e.we have to show that Y
is invariant under φt . Hence, from Exercise 2.48, we wish to prove φt ◦ ψs = ψs ◦ φt .
Now

(φt ◦ ψs )(x) = φt (exp(sY )x)


= exp(sY ) exp(t X )x
= exp(t X ) exp(sY )x, as G is abelian
= ψs (exp(t X )x)
= (ψs ◦ φt )(x).

Therefore, φt ◦ ψs = ψs ◦ φt , for all x ∈ G. This completes the proof.


1
Theorem 4.8 If X, Y ∈ g, then [Y, X ] = lim {(Adat−1 )Y − Y }.
t→0 t

Proof If {φt |t ∈ R} is the one-parameter group of transformations on a Lie group


generated by the left invariant vector field X , then

1
[Y, X ] = lim {(φt )∗ Y − Y }, for every vector field Y ∈ χ (G).
t→0 t

Now Adat−1 = (adat−1 )∗ = (Rat ◦ L at−1 )∗ , see (4.29).


If Y ∈ g, then from above after a few steps,

(Adat−1 )Y = (Rat )∗ Y = (φt )∗ Y, from Theorem 4.7.


204 4 Lie Group

Thus
1
[Y, X ] = lim {(Adat−1 )Y − Y }.
t→0 t

This completes the proof.

4.6 Lie Transformation Group (Action of a Lie Group


on a Manifold)

A Lie Group G is a Lie Transformation Group on a manifold M or G is said to act


differentiably on M if the following conditions hold:

⎨ (i) (a, p) : G × M → pa(∈ M) is a differentiable map;
(ii) Each a ∈ G induces a transformation on M, denoted by p → pa (4.35)

(iii) p(ab) = ( pa)b, ∀ a, b ∈ G.

We say that G acts on M on the right as (i) and (iii) can be written as
(i) Ra p = pa.
(ii) Rab p = p(ab) = ( pa)b.
Similarly, the action of G on M on the left can be defined.
Exercises

Exercise 4.21 Let G = G L 2 (R), M = R and θ : G × M → M be a differentiable


mapping defined by
  
ab
θ , p = ap + b, a > 0, a, b ∈ R.
01

Show that θ is an action on M.

If G acts on M on the right such that pa = p, ∀ p ∈ M implies that a = e, then


G is said to act effectively on M.
However, if G acts on M on the right such that pa = p, ∀ p ∈ M implies that
a = e, for some p ∈ M, then G is said to act freely on M.
Theorem 4.9 If G acts on M, then the mapping σ : g → χ (M) denoted by
σ : A → σ (A) = A∗ is a Lie algebra homomorphism. It is to be noted that σ (A) =
A∗ is called the fundamental vector field on M, corresponding to A ∈ g (Fig. 4.2).

Proof For every p ∈ M, let σ p : G → M be a mapping such that

σ p (a) = pa, ∀ a ∈ G. (4.36)


4.6 Lie Transformation Group (Action of a Lie Group on a Manifold) 205

Fig. 4.2 Lie Algebra Homomorphism

Note that both g, χ (M) are algebra, hence we have to show


(i) σ is linear.
(ii) σ ([A, B]) = [σ (A), σ (B)].
Note that every A ∈ g induces {φt (e)| t ∈ R} as its one-parameter group of
transformations on G such that

a(t) = at = φt (e).

The map (σ p )∗ : Te (G) → Tσ p (e) (M) = T p (M) is a linear map.


Now by hypothesis, we have

(σ p )∗ Ae = {σ (A)}σ p (e) = {σ (A)} p = A∗p . (4.37)

For every A, B ∈ g, we have A + B ∈ g and

{σ (A + B)} p = (σ p )∗ (A + B)e = (σ p )∗ (Ae + Be ) = {σ (A)} p + {σ (B)} p


{σ (b A)} p = (σ p )∗ Ae = b{σ (A)} p , ∀ b ∈ R.

This proves that σ is linear.


Again, Ae is the tangent vector to the curve a(t) ≡ at at a(0) = e. Then (σ p )∗ Ae
is the tangent vector to the curve

σ p (at ) = pat , by (4.36)


= Rat ( p), by (4.3)

at σ p (a0 ) = σ p (e) = p. Thus, A∗p induces {Rat ( p)} as its one-parameter group of
transformations on M. Now
206 4 Lie Group

[σ (A), σ (B)] p = [A∗ , B ∗ ] p


1   
= lim B ∗p − (Rat )∗ B ∗ p , by Theorem 2.8
t→0 t
1
= lim {(σ p )∗ Be − (Rat )∗ Bq∗ }, say, where Rat (q) = p,
t→0 t
i.e. q = pat−1 .
∴ (Rat )∗ Bq∗ = (Rat )∗ B ∗pa −1 = (Rat ◦ σ pat−1 )∗ Be , by (4.37).
t

Now, Rat ◦ σ pat−1 : G → M and hence for b ∈ G, we have

(Rat ◦ σ pat−1 )(b) = pat−1 bat = σ p ((adat−1 )(b)) = (σ p ◦ adat−1 )(b).


∴ Rat ◦ σ pat−1 = σ p ◦ adat−1 .

Hence
   
(Rat )∗ Bq∗ = (σ p )∗ (adat−1 )∗ Be = (σ p )∗ (Adat−1 )Be , by (4.32).

Thus
1  
[σ (A), σ (B)] p = lim (σ p )∗ Be − (σ p )∗ (Adat−1 )Be
t→0 t
= (σ p )∗ [A, B]e , as (σ p )∗ is linear and by Theorem 4.8
 
= σ ([A, B]) p , by (4.37)
i.e. σ ([A, B]) = [σ (A), σ (B)].

Hence, the mapping σ : g → χ (M) is a linear algebra homomorphism. This com-


pletes the proof.
Theorem 4.10 If G acts effectively on M, then the map σ : g → χ (M) defined by
A → σ (A) = A∗ is an isomorphism.
Proof In view of Theorem 4.9, we can say that σ is a Lie algebra homomorphism.
We are left to prove that σ is bijective. Let σ (A) = σ (B) hold for every A, B ∈ g.
Then,
σ (A − B) = θ ⇒ (A − B)∗ = θ.

Now for every A, B in g, A − B(∈ g) will generate {ψt (e)| t ∈ R}(say) as its one-
parameter group of transformations on G such that (A − B)e is the tangent vector to
the curve, given by
bt = ψt (e) at ψ0 (e) = b0 = e. (4.38)

Consequently, (σ p )∗ (A − B)e is the tangent vector to the curve

σ p (bt ) = pbt = Rbt ( p), at p b0 = p.


4.6 Lie Transformation Group (Action of a Lie Group on a Manifold) 207

Thus, (A − B)∗ ≡ (σ p )∗ (A − B)e will generate {Rbt ( p)| t ∈ R} as its one-parameter


group of transformations on M. But the integral curve of the null-vector (A − B)∗
will reduce to the initial point itself, refer to Problem 2.42. Hence,

σ p (bt ) = p i.e. Rbt ( p) = p i.e. p bt = p.

As G acts effectively on M, from the foregoing equation, we obtain bt = e, ∀ p ∈ M.


Again, (L p )∗ (A − B) = A − B. Thus, using Exercise 2.48, we have L p ◦ ψt = ψt ◦
L p . Hence

ψt (q) = ψt (qe) = (ψt ◦ L q )(e) = L q (ψt (e))


= L q (bt ), refer to (4.38)
= q, from above.

From the definition, we have

1
(A − B)q f = lim { f (ψt (q)) − f (q)} = 0.
t→0 t

Therefore, A − B is a null vector. So from σ (A) = σ (B), we must have A = B.


Clearly, σ is surjective. Hence, σ is injective. Thus, σ is bijective and consequently,
σ is an isomorphism.

Exercises

Exercise 4.22 If G acts freely on M, the proof that for every non-null left invariant
vector field A, the fundamental vector field A∗ can never vanish.

Exercise 4.23 Prove that the map θ : R+ × R → R defined by (a, p) → ax is an


action of R+ on R. Is it free?

Answer/Hint

4.22. Use Theorems 4.9 and 4.10.


4.23. No
References

1. Abraham, R., Marsden, J.E., Ratiu, T.: Manifolds, Tensor Analysis and Applications, 2nd edn.
Springer (1983)
2. Bishop, R.L., Crittenden, R.J.: Geometry of Manifolds. AMS Chelsea Publishing, Providence
(2001)
3. Boothby, W.M.: An Introduction to Differentiable Manifolds and Riemannian Geometry, 2nd
edn. Academic Press, New York (2002)
4. Castillo, G.F.: Differentiable Manifolds. Birkhäuser (2012)
5. Curtis, W.D., Miller, F.R.: Differential Manifolds and Theoretical Physics. Academic Press
(1985)
6. Fitzpatrick, P.M.: Advanced Calculus. Pure and Applied, Undergraduate Texts, 2nd edn. Amer-
ican Mathematical Society, Providence, Rhode Island
7. Gadea, P.M., Masqué, J.M., Mykytyuk, I.V.: Analysis and Algebra on Differentiable Manifolds,
2nd edn. Springer (2013)
8. Gallot, S., Hulin, D., Lafontaine, J.: Riemannian Geometry. Springer, Berlin (2004)
9. Guillemin, V., Pollack, A.: Differential Topology. Prentice-Hall, New Jersey (1974)
10. Helgason, S.: Differential Geometry, Lie Groups and Symmetric Spaces. Graduate Studies in
Mathematics, vol. 34. Amc. Math. Soc, Providence (2012)
11. Hicks, N.J.: Notes on Differential Geometry. Von Nostrand Reinhold, London (1965)
12. Hoffman, K., Kunze, R.: Linear Algebra, 2nd edn. Pearson (2015)
13. Kobayashi, S., Nomizu, K.: Foundations of Differential Geometry, vol. I, II. Wiley, New York
(1996)
14. Kumaresan, S.: A Course in Differential Geometry and Lie Groups. Hindustan Book Agency
(2002)
15. Lee, J.M.: Introduction to Smooth Manifolds, 2nd edn. Springer, New York (2012)
16. Matsushima, Y.: Differentiable Manifold. Mercel Dekker INC, New York (1972)
17. Millman, R.S., Parker, G.D.: Elements of Differential Geometry. Pearson (1977)
18. Morita, S.: Geometry of Differential Forms. A.M.S. (2001)
19. Mukherjee, A.: Differential Topology. Hindustan Book Agency (2015)
20. Munkres, J.: Analysis on Manifolds. Addison-Wesley, CA (1991)
21. Munkres, J.: Topology, 2nd edn. Pearson (2000)

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer 209
Nature Singapore Pte Ltd. 2023
M. Majumdar and A. Bhattacharyya, An Introduction to Smooth Manifolds,
https://doi.org/10.1007/978-981-99-0565-2
210 References

22. Prakash, N.: Differential Geometry, An Integrated Approach. Tata-McGraw-Hill Pub. Com
Ltd, New-Delhi (1981)
23. Pressley, A.: Elementary Differential Geometry, 2nd edn. Springer (2010)
24. Rudin, W.: Principles of Mathematical Analysis, 3rd edn. Mc.Graw-Hill, Inc
25. Schutz, B.: Geometrical Methods of Mathematical Physics. Cambridge Univ. Press (1980)
26. Sen, S.: A Short Course on Differentiable Manifolds. University of Calcutta (2011)
27. Shirali, S., Vasudeva, H.L.: Multivariable Analysis. Springer (2011)
28. Sinha, B.B.: An Introduction to Modern Differential Geometry. Kalyani Publisher, New-Delhi
(1982)
29. Spivak, M.: Calculus on Manifolds. Benjamin, New York (1965)
30. Spivak, M.: Differential Geometry, vol. 1–5, 3rd edn. Publish or Perish Wilmington (1999)
31. Sinha, R.: Smooth Manifolds. Springer (2014)
32. Tu, L.W.: An Introduction to Smooth Manifolds, 2nd edn. Springer, Berlin (2008)
33. Warner, F.W.: Foundations of Differentiable Manifolds and Lie Groups. Springer, Berlin (2010)
34. Weintraub, S.: Differential Forms, Theory and Practice, 2nd edn. Academic Press (2014)
35. Yano, K., Kon, M.: Structures on Manifolds. Worlds Scientific (1984)

You might also like