(University Texts in The Mathematical Sciences) Manjusha Majumdar, Arindam Bhattacharyya - An Introduction To Smooth Manifolds-Springer (2023)
(University Texts in The Mathematical Sciences) Manjusha Majumdar, Arindam Bhattacharyya - An Introduction To Smooth Manifolds-Springer (2023)
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
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.
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
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
About the Authors
xiii
List of Figures
xv
Chapter 1
Calculus on Rn
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.
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
f (x) =
0, x = 0.
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
24u
f (0) = 2 lim .
u→∞ 4ueu 2 + 8ueu 2 + 8u 3 eu 2
f n (0) = 0, n = 1, 2, 3, . . . . . . .
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
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
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.
Answers
1.2 No.
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)
∂x1 ∂x2
⎜ ∂f 2 ∂f 2 2 ⎟
⎜
J =⎜ · · · ∂∂ xf n ⎟ ,
∂x1 ∂x2 ⎟
⎝ ··· ··· ··· ··· ⎠
∂f ∂f
· · · ∂∂ fx n
m m m
∂x1 ∂x2
Solution: Here
Solution:
(i) Here
f (0) = lim h 2 = 0
h→0
f (x) = 3x 2 .
f (x) − f (y) = x 3 − y 3
⇒ f (x) = f (y) if and only if x = y.
f −1 (x) = x 3 .
1
Exercises
φ(x 1 , x 2 , x 3 ) = (x 1 + 1, x 2 + 2, x 3 + 3).
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
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
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.
Prove that φ is a diffeomorphism from the unit sphere S 2 onto itself.
S2
1.2 Tangent Vector 9
Answers
1.11 diffeomorphism. 1.12 (ii) No, f is not one-to-one.
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 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.
∂ 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
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
∂ ∂
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
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
∂ ∂
+ .
∂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.
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.
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
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
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 1.20 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).
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
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
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
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
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
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
∂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
∂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 .
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
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
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
which implies
|g (t)| ≤ f (γ (t))||b − a|| ≤ M||b − a||, ∀ t ∈ [0, 1].
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
1
|ψ y (x)| < , x ∈ V. (1.15)
2
Hence, by virtue of Lemma 1.1, it follows that
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.
y ∈ Rn , ||y − w◦ || < ⇒ y ∈ W.
2|F (x∗ )−1 |
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)
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||
||h|| ≤ |F (F −1 (y))| ||k|| + ||h|| |F (F −1 (y))| ||R(h)||,
||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.
and
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
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
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.
Let us define
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
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
3x1 + x2 − x3 − u 3 = 0
x1 − x2 + 2x3 + u = 0
2x1 + 2x2 − 3x3 + 2u = 0,
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.
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.
F1 (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ) = 0
..
.
Fi (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ) = 0
..
.
Fn (x1 , x2 , . . . , xm , y1 , y2 , . . . , yn ) = 0.
(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 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,
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.
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 .
Exercises
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.
(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
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).
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
Example 2.1 Rn is a topological manifold covered by a single chart (Rn , IRn ) where
I is the identity map.
Problem 2.2 Consider the open subsets U and V of the unit circle S 1 of R2 given
by
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.
Problem 2.4 Find the functional relation between the two local coordinate systems
defined in the overlap region of a topological manifold.
Then
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.
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
Br (0, 0, 0, . . . , 0) \ (0, 0, 0, . . . , 0)
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
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}.
Answers
2.1. Yes. 2.2. Not on the y-axis. 2.3. Yes. 2.5. No.
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 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
( f ◦ φ−1 ) : φ(Dom f ∩ U ) → R
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.
(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).
where
[γ] = {γi ∈ p (M) : γ ≺ γi }.
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
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),
and
t [γ1 ] = φ−1
∗
(t (φ∗ (γ1 ))), ∀ t ∈ R.
2.2 Smooth Germs on a Topological Manifold 45
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 ]),
where
[ f ] = {g : g ∈ C ∞
p (M), g f }
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 2.12 F p (M) forms a real linear space. Also, F p (M) forms an algebra.
( 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
( f ◦ γ)(0 + t) − ( f ◦ γ)(0)
lim
t→0 t
( f ◦ γ)(0 + t) − ( f ◦ γ)(t)
< γ, [ f ] >≡ lim ,
t→0 t
i.e.
d( f ◦ γ)(t)
< γ, [ f ] >≡ .
dt t=0
(( 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] > .
d( f (γ(t)) · g(γ(t)))
=
dt t=0
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
are of class C ∞ .
Note that
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
Example 2.5 Rn is a smooth manifold with respect to the atlas {(U, φ)} where
U = Rn and φ is the identity map.
is of class C ∞ .
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.
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 ∞
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
such that
α, if α ∈ (0, π)
(ψ ◦ φ−1 )(α) = ψ(cos α, sin α) =
α − 2π, if α ∈ (π, 2π)
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
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 .
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
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
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
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
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
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.
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
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
N = lim Nφ = (−1, 0, 0)
v→0
N = lim Nφ = (1, 0, 0).
v→2π
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
Exercise
f ◦ φ−1 : φ(U ) ⊂ Rn → R
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 ).
f ◦ ψ −1 = ( f ◦ φ−1 ) ◦ (φ ◦ ψ −1 )
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 .
Proposition 2.5 The notion of smoothness of a map between two smooth manifolds
is independent of the choice of a coordinate chart.
is C ∞ . This proves that the C ∞ structure of a map does not depend on any coordinate
chart chosen.
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.
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)
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,
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
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
Often, we write it as
x i (t) = σ i (t). (2.14)
d d
= a f (σ(t))t=t + b g(σ(t))t=t , by (1.8)
dt 0 dt 0
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
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
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).
∂ 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: 31 Here
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
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
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.
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.
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)
[X, Y ] f = X (Y f ) − Y (X f ). (2.27)
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.
[X + Y, X + Y ] = θ
i.e. [X, Y ] + [Y, X ] = θ,
∴ [X, Y ] = −[Y, X ].
Thus
( f X )h = f (X h), ∀ p ∈ M. (2.28)
∂ ∂ ∂
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)
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
∂
{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
∂ ∂ ∂
X= + + .
∂x ∂y ∂z
∂ ∂ ∂
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
∂ 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 ).
∂ ∂
{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.
x 2 − y 2 = f (x − y) + gx 2 and x 2 + y 2 = − f + g y.
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 2.21 (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
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.
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).
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
d x i (t)
= ξ i (x 1 (t), x 2 (t), . . . , x n (t)). (2.29)
dt
84 2 Manifold Theory
∂ ∂ ∂
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 ,
∂ ∂
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
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.
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
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 ),
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
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 ),
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 ,
∂ ∂
[X, Y ] = −x +y .
∂x ∂y
88 2 Manifold Theory
dx dy
= −x, = y,
dt dt
which on solving, one gets
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 .
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?
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
f ∗ : F( f ( p)) → F( p), by
{ f ∗ (X p )}g = X p ( f ∗ g) = X p (g ◦ f ) (2.34)
f ∗ (X p ) = ( f ∗ X ) f ( p) . (2.35)
Also
{ f ∗ (X p )}(λh) = λX p (h ◦ f ) = λ{ f ∗ (X p )h}, λ ∈ R.
{(I∗ ) p X p }g = X p (g ◦ I ) = X p g,
∴ (I∗ ) p X p = X p , ∀ g.
(g ◦ f )∗ = g∗ ◦ f ∗ .
((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).
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.
{ 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
f ( p) = q ⇒ p = f −1 (q), ∀ p, q ∈ M.
{ 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 ].
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).
∂ 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
∂ k j k
f∗ y = ai δ j = aik .
∂x i p j
∂ 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.
∂ ∂
φ∗ = ,
∂x i p ∂u i φ( p)
(I N )∗ = ( f −1 ◦ f )∗ = f ∗−1 ◦ f ∗ ,
∂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
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).
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
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) .
∂ ∂
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.
cos θ − sin θ −x2
( f ∗ )(X ) p =
sin θ cos θ x1 f ( p)= f (x ,x )
1 2
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 )}.
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)
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).
∂ ∂
A +B
∂x1 ∂x2 g(0,0)
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
rank f | p = dim M = n.
Here
rank f ∗, p ≤ min{2, 3} = 2 ⇒ rank f ∗, p = 1 or 2.
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
(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
∂2 f
det ( p) = 0.
∂x ∂x
i j
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.
(x, y, z) → (x z, y).
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
f∗ f∗ f∗ f∗ f∗ f∗ f∗ f∗
Injective Surjective Injective Surjective Injective Surjective Injective Surjective
2.13 Submanifolds
ζi : Mi → M1 × M2 × · · · × Ms ,
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
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).
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.
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
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.
{ f ∗ (X p )}g = Y f ( p) g, ∀ g ∈ F( f ( p)).
If f is a transformation on M and
f ∗ (X p ) = X f ( p)
i.e. ( f ∗ X ) f ( p) = X f ( p) ,
f ∗ X = X. (2.40)
∂ ∂ ∂ ∂
X̄ = a +b +c +t , (2.41)
∂u ∂v ∂ω ∂t
{ f ∗ X }u = X̄ u = a.
X (u ◦ f ) = a.
∂ ∂
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
∂ ∂ ∂ ∂
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
∂ ∂ ∂ ∂
Ȳ = a + b + c + d , (2.42)
∂u ∂v ∂ω ∂t
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)
X (g ◦ f ) = (Y g) ◦ f
or X ( f ∗ g) = f ∗ (Y g). by (2.32)
X ( f ∗ g) = f ∗ (( f ∗ X )g).
X p = X σ(t0 ) ,
d
Xp f = f (σ(t))t=t .
dt 0
2.14 f -Related Vector Fields 113
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
Y f ( p) = Y f (σ(t0 )) ,
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
Answer
∂ ∂ ∂ ∂ ∂
2.41. uw + (v − 1) + (ω − 1)2 ; u + (ω − 1) .
∂u ∂v ∂ω ∂u ∂ω
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).
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
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)
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)
∂ ∂ ∂
+ 2 + 3 where p = (x 1 , x 2 , x 3 ) ∈ R3 .
∂x 1 ∂x ∂x
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
φt (φ−t (x , y )) = (x , y ) and φ−t (φt (x, y)) = (x, y),
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
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
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 )
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).
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.
dx dy dz
=1 =0= , where (x, y, z) ∈ R3 .
dt dt dt
On solving, we get
x = t + A, y = B, z = C
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
φ( p) = (0, 0, 0, . . . , 0) ∈ Rn .
∂
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
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)
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)),
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
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,
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
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
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.
d d
Xp = ψ(t)t=0 = (φt ( p))t=0 .
dt 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.
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
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
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
1
g(0, p) = lim { f (φt ( p)) − f (φ◦ ( p))} = X p f.
t→0 t
Therefore,
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
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
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
(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.
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
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)
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
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
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
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)].
Solution: Here
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
∂ ∂ ∂
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 3.7 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
∂ ∂
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
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
∂ ∂ ∂ ∂ ∂ ∂
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
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
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
∂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
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
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 ).
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
1 y
e2 = dy − dz
(1 + y 2 ) (1 + y 2 )
1
e3 = − dz.
(1 + y 2 )
Exercises
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
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
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
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
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
(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
f ∧ g = f g; ∀ f, g ∈ F(M). (3.15)
142 3 Differential Forms
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 .
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 .
α × β = (a1 b2 − a2 b1 , a2 b3 − a3 b2 , a3 b1 − a1 b3 ).
which is a 3-form.
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 )
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 .
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 )
∂ ∂ ∂ 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
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
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
Answers
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
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
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
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
k
μi = Ai j ω j with Ai j = A ji , i = 1, 2, 3, 4, . . . , k.
j=1
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
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
k
μi = Aim ω j , with Ai j = A ji .
j=1
3.3 Exterior Differentiation 149
Exercise
n
μi = ai j ω j .
j=1
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.
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.
(iv) Similarly
dω4 = (2x yz − 2x z + 3x yz 2 ) d x ∧ dy ∧ dz.
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.
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
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
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 3.25 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
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)
EXAMPLE:
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
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)
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.
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
Then
∂
x z = μy = (yzx + C1 (y, z)) = zx + C1 (y, z)
∂y
⇒ C1 (y, z) = 0
⇒ C1 (y, z) = constant = C2 (z)say.
∂
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.
1
ω= (x dy ∧ dz + y dz ∧ d x + z d x ∧ dy),
(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
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
∂ ∂ ∂
X = x1 + 2x2 , Y = x1 x2 ,
∂ x1 ∂ x2 ∂ x2
Solution: Here
160 3 Differential Forms
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.
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,
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
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.
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 ).
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 .
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.
∂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.
( 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)
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
∂( 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
Consequently, we have
n
∂f j
f ∗ (dy j ) f ( p) = (d x i ) p , j = 1, 2, 3, . . . , m.
i=1
∂xi
f ∗ (dy j ) f ( p) = (d f j ) p , j = 1, 2, 3, . . . , m. (3.29)
{ 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 ∗ ω)(X 1 , X 2 , . . . , X r ) = ω( f ∗ X 1 , f ∗ X 2 , . . . , f ∗ X r ). (3.32)
{ 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
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
= 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).
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.
B. Note that
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).
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).
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) .
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 ∗ (d x ∧ dy ∧ dz) = f ∗ d x ∧ f ∗ dy ∧ f ∗ dz.
Now
172 3 Differential Forms
Similarly
Therefore
Exercises
Answers
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
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
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.
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 ).
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
Exercises
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)
Examples
x = (x ij ) → ax = (aki x kj )
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,
φ(ab) = φ(a)φ(b), ∀ a, b ∈ G 1 .
Now
Therefore, φ ◦ Rb = Rφ(b) ◦ φ, ∀ x ∈ G1 .
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 4.7 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.
(L a )∗ X p = X L a ( p) = X ap , ∀ p ∈ G, (4.7)
((L a )∗ X ) L a ( p) = X L a ( p) i.e.
(L a )∗ X = X. (4.8)
(Ra )∗ X = X. (4.9)
{(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).
{(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)
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.
∂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
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 ].
Problem 4.7 Show that the vector space R3 with the operation cross product of
vectors is a Lie algebra.
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 ).
(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,
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.
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)
φ −1 (Ye ) = X. (4.14)
(L a )∗ Ye = X a . (4.15)
(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.
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
[ 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 ].
n
[X i , X j ] = Cikj X k , Cikj ∈ R. (4.16)
k=1
n
n
Cikj X k = − C kji X k
k=1 k=1
or (Cikj + C kji ) = 0,
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
Cikj Cks
t
+ C kjs Ckit + Csik Ckt j = 0.
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
∂ ∂ ∂ ∂ ∂ ∂
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)
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
(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, 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 ,
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).
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.
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 ).
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.
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
bω ∧ μ = b(ω ∧ μ) = ω ∧ bμ, ∀ b ∈ R.
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.
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
1
Thus dω(X, Y ) = − ω([X, Y ]).
2
192 4 Lie Group
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)
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
+ C23
i
ω2 ∧ ω3 + C31
i
ω3 ∧ ω1 + C32
i
ω3 ∧ ω2
= 2C12
i
ω1 ∧ ω2 + 2C13
i
ω1 ∧ ω3 + 2C23
i
ω2 ∧ ω3 ,
i
C ijk ω j ∧ ωk = 2 C ijk ω j ∧ ωk .
j,k=1,2,3 j,k
j<k
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
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 .
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
f (eG 1 ) = eG 2 . (4.25)
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
Now
x y x y x y x y
f + f = f .
0x 0 x 0x 0 x
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.
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.
f [X, Y ] = [ f X, f Y ]. (4.30)
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 ,
(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
f (e) = e ,
g1 ∼
= Te (G 1 ), and g2 ∼
= Te (G 2 ),
f ∗ (X e ) = Ye ∈ Te (G 2 ).
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 ],
(ada)(x) = axa −1 , ∀ x ∈ G.
Now
(L p )∗ {(Ra −1 )∗ Y } = (L p ◦ Ra −1 )∗ Y, p ∈ G
= (Ra −1 )∗ Y, see (4.29).
(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
Ada ≡ (ada)∗ : g → g
Similarly, it can be shown that (Ada −1 ) ◦ (Ada) = Ade. Hence, Ada−1 = (Ada)−1 .
This completes the proof.
Exercise 4.20 If X is a left invariant vector field on a Lie Group G, prove that X is
complete.
{(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).
φt (e) = a(t) = at .
Proof As defined
202 4 Lie Group
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
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
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
1
[Y, X ] = lim {(φt )∗ Y − Y }, for every vector field Y ∈ χ (G).
t→0 t
Thus
1
[Y, X ] = lim {(Adat−1 )Y − Y }.
t→0 t
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
a(t) = at = φt (e).
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
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)].
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)
1
(A − B)q f = lim { f (ψt (q)) − f (q)} = 0.
t→0 t
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.
Answer/Hint
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)