INTRODUCTION TO

DIFFERENTIABLE
MANIFOLDS
spring 2012
Course by Prof. dr. R.C.A.M. van der Vorst
Solution manual Dr. G.J. Ridderbos
1
2
Introduction to differentiable manifolds
Lecture notes version 2.1, November 5, 2012
This is a self contained set of lecture notes. The notes were written by Rob van
der Vorst. The solution manual is written by Guit-Jan Ridderbos. We follow the
book Introduction to Smooth Manifolds by John M. Lee as a reference text [1].
Additional reading and exercises are take from An introduction to manifolds by
Loring W. Tu [2].
This document was produced in L
A
T
E
X and the pdf-le of these notes is available
on the following website
www.few.vu.nl/vdvorst
They are meant to be freely available for non-commercial use, in the sense that
free software is free. More precisely:
This work is licensed under the Creative Commons
Attribution-NonCommercial-ShareAlike License.
To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-sa/2.0/
or send a letter to Creative Commons, 559 Nathan Abbott Way, Stanford, California
94305, USA.
CONTENTS
References 5
I. Manifolds 6
1. Topological manifolds 6
2. Differentiable manifolds and differentiable structures 13
3. Immersions, submersions and embeddings 20
II. Tangent and cotangent spaces 32
4. Tangent spaces 32
5. Cotangent spaces 38
6. Vector bundles 41
6.1. The tangent bundle and vector elds 44
6.2. The cotangent bundle and differential 1-forms 46
III. Tensors and differential forms 50
7. Tensors and tensor products 50
8. Symmetric and alternating tensors 55
3
9. Tensor bundles and tensor elds 61
10. Differential forms 66
11. Orientations 69
IV. Integration on manifolds 75
12. Integrating m-forms on R
m
75
13. Partitions of unity 76
14. Integration on of m-forms on m-dimensional manifolds. 81
15. The exterior derivative 87
16. Stokes Theorem 92
V. De Rham cohomology 97
17. Denition of De Rham cohomology 97
18. Homotopy invariance of cohomology 98
VI. Exercises 102
Manifolds
Topological Manifolds 102
Differentiable manifolds 103
Immersion, submersion and embeddings 104
Tangent and cotangent spaces
Tangent spaces 106
Cotangent spaces 107
Vector bundles 108
Tensors
Tensors and tensor products 109
Symmetric and alternating tensors 109
Tensor bundles and tensor elds 110
Differential forms 111
Orientations 111
Integration on manifolds
Integrating m-forms on R
m
112
Partitions of unity 112
Integration of m-forms on m-dimensional manifolds 112
The exterior derivative 113
Stokes Theorem 114
4
Extras 115
De Rham cohomology
Denition of De Rham cohomology 117
VII. Solutions 118
5
References
[1] John M. Lee. Introduction to smooth manifolds, volume 218 of Graduate Texts in Mathematics.
Springer-Verlag, New York, 2003.
[2] Loring W. Tu. An introduction to manifolds. Universitext. Springer, New York, second edition,
2011.
6
I. Manifolds
1. Topological manifolds
Basically an m-dimensional (topological) manifold is a topological space M
which is locally homeomorphic to R
m
. A more precise denition is:
Denition 1.1.
1
A topological space M is called an m-dimensional (topological)
manifold, if the following conditions hold:
(i) M is a Hausdorff space,
(ii) for any p M there exists a neighborhood
2
U of p which is homeomorphic
to an open subset V R
m
, and
(iii) M has a countable basis of open sets.
Axiom (ii) is equivalent to saying that p M has a open neighborhood U 3 p
homeomorphic to the open disc D
m
in R
m
. We say M is locally homeomorphic to
R
m
. Axiom (iii) says that M can be covered by countably many of such neighbor-
hoods.

FIGURE 1. Coordinate charts (U, ).

1
See Lee, pg. 3.
2
Usually an open neighborhood U of a point p M is an open set containing p. A neighborhood
of p is a set containing an open neighborhood containing p. Here we will always assume that a
neighborhood is an open set.
7

FIGURE 2. The transition maps
i j
.
Recall some notions from topology: A topological space M is called Hausdorff
if for any pair p, q M, there exist (open) neighborhoods U 3 p, and U
0
3 q such
that U U
0
= . For a topological space M with topology , a collection is
a basis if and only if each U can be written as union of sets in . A basis is
called a countable basis if is a countable set.
Figure 1 displays coordinate charts (U, ), where U are coordinate neighbor-
hoods, or charts, and are (coordinate) homeomorphisms. Transitions between
different choices of coordinates are called transitions maps
i j
=
j

1
i
, which
are again homeomorphisms by denition. We usually write x = (p), : U
V R
n
, as coordinates for U, see Figure 2, and p =
1
(x),
1
: V U M, as
a parametrization of U. A collection A = {(
i
,U
i
)}
iI
of coordinate charts with
M =
i
U
i
, is called an atlas for M.
The following Theorem gives a number of useful characteristics of topological
manifolds.
Theorem 1.4.
3
A manifold is locally connected, locally compact, and the union of
countably many compact subsets. Moreover, a manifold is normal and metrizable.
J1.5 Example. M =R
m
; the axioms (i) and (iii) are obviously satised. As for (ii)
we take U =R
m
, and the identity map. I
3
Lee, 1.6, 1.8, and Boothby.
8
J1.6 Example. M = S
1
= {p = (p
1
, p
2
) R
2
| p
2
1
+p
2
2
= 1}; as before (i) and (iii)
are satised. As for (ii) we can choose many different atlases.
(a): Consider the sets
U
1
= {p S
1
| p
2
> 0}, U
2
= {p S
1
| p
2
< 0},
U
3
= {p S
1
| p
1
< 0}, U
4
= {p S
1
| p
1
> 0}.
The associated coordinate maps are
1
(p) = p
1
,
2
(p) = p
1
,
3
(p) = p
2
, and

4
(p) = p
2
. For instance

1
1
(x) =

x,
_
1x
2

,
and the domain is V
1
= (1, 1). It is clear that
i
and
1
i
are continuous, and
therefore the maps
i
are homeomorphisms. With these choices we have found an
atlas for S
1
consisting of four charts.
(b): (Stereographic projection) Consider the two charts
U
1
= S
1
\{(0, 1)}, and U
2
= S
1
\{(0, 1)}.
The coordinate mappings are given by

1
(p) =
2p
1
1p
2
, and
2
(p) =
2p
1
1+ p
2
,
which are continuous maps fromU
i
to R. For example

1
1
(x) =

4x
x
2
+4
,
x
2
4
x
2
+4

,
which is continuous from R to U
1
.

FIGURE 3. The stereographic projection describing

1
.
(c): (Polar coordinates) Consider the two charts U
1
= S
1
\{(1, 0)}, and U
2
=
S
1
\{(1, 0)}. The homeomorphism are
1
(p) = (0, 2) (polar angle counter
9
clockwise rotation), and
2
(p) = (, ) (complex argument). For example

1
1
() = (cos(), sin()). I
J 1.8 Example. M = S
n
= {p = (p
1
, p
n+1
) R
n+1
| |p|
2
= 1}. Obvious exten-
sion of the choices for S
1
. I
J 1.9 Example. (see Lee) Let U R
n
be an open set and g : U R
m
a continuous
function. Dene
M =(g) = {(x, y) R
n
R
m
: x U, y = g(x)},
endowed with the subspace topology, see 1.17. This topological space is an n-
dimensional manifold. Indeed, dene : R
n
R
m
R
n
as (x, y) = x, and
= |
(g)
, which is continuous onto U. The inverse
1
(x) = (x, g(x)) is also
continuous. Therefore {((g), )} is an appropriate atlas. The manifold (g) is
homeomorphic to U. I
J 1.10 Example. M = PR
n
, the real projective spaces. Consider the following
equivalence relation on points on R
n+1
\{0}: For any two x, y R
n+1
\{0} dene
x y if there exists a 6= 0, such that x =y.
Dene PR
n
= {[x] : x R
n+1
\{0}} as the set of equivalence classes. One can
think of PR
n
as the set of lines through the origin in R
n+1
. Consider the natural
map : R
n+1
\{0} PR
n
, via (x) = [x]. A set U PR
n
is open if
1
(U) is open
in R
n+1
\{0}. This makes a quotient map and is continuous. The equivalence
relation is open (see [2, Sect. 7.6]) and therefore PR
n
Hausdorff and second
countable, see 1.17. Compactness of PR
n
can be proves by showing that there
exists a homeomorphism from PR
n
to S
n
/ , see [2, Sect. 7.6]. In order to verify
that we are dealing with an n-dimensional manifold we need to describe an atlas
for PR
n
. For i = 1, n+1, dene V
i
R
n+1
\{0} as the set of points x for which
x
i
6= 0, and dene U
i
=(V
i
). Furthermore, for any [x] U
i
dene

i
([x]) =

x
1
x
i
, ,
x
i1
x
i
,
x
i+1
x
i
, ,
x
n+1
x
i

,
which is continuous. This follows from the fact that is continuous. The
continuous inverse is given by

1
i
(z
1
, , z
n
) =

(z
1
, , z
i1
, 1, z
i
, , z
n
)

.
These charts U
i
cover PR
n
. In dimension n = 1 we have that PR

= S
1
, and in the
dimension n = 2 we obtain an immersed surface PR
2
as shown in Figure 4. I
The examples 1.6 and 1.8 above are special in the sense that they are subsets of
some R
m
, which, as topological spaces, are given a manifold structure.
10
FIGURE 4. Identication of the curves indicated above yields an
immersion of PR
2
into R
3
.
Dene
H
m
= {(x
1
, , x
m
) | x
m
0},
as the standard Euclidean half-space.
Denition 1.12. A topological space M is called an m-dimensional (topological)
manifold with boundary M M, if the following conditions hold:
(i) M is a Hausdorff space,
(ii) for any point p M there exists a neighborhood U of p, which is homeo-
morphic to an open subset V H
m
, and
(iii) M has a countable basis of open sets.
Axiom (ii) can be rephrased as follows, any point p M is contained in a neigh-
borhood U, which either homeomorphic to D
m
, or to D
m
H
m
. The set M is locally
homeomorphic to R
m
, or H
m
. The boundary M M is a subset of M which con-
sists of points p for which any neighborhood cannot be homeomorphic to an open
subset of int (H
m
). In other words point p M are points that lie in the inverse
image of V H
m
for some chart (U, ). Points p M are mapped to points on
H
m
and M is an (m1)-dimensional topological manifold.
J 1.14 Example. Consider the bounded cone
M =C = {p = (p
1
, p
2
, p
3
) R
3
| p
2
1
+ p
2
2
= p
2
3
, 0 p
3
1},
with boundary C = {p C | p
3
= 1}. We can describe the cone via an atlas
consisting of three charts;
U
1
= {p C | p
3
< 1},
with x = (x
1
, x
2
) =
1
(p) = (p
1
, p
2
+1), and

1
1
(x) =

x
1
, x
2
1,
_
x
2
1
+(x
2
1)
2

,
11
FIGURE 5. Coordinate maps for boundary points.
The other charts are given by
U
2
= {p C |
1
2
< p
3
1, (p
1
, p
2
) 6= (0, p
3
)},
U
3
= {p C |
1
2
< p
3
1, (p
1
, p
2
) 6= (0, p
3
)}.
For instance
2
can be constructed as follows. Dene
q =(p) =

p
1
p
3
,
p
2
p
3
, p
3

, (q) =

2q
1
1q
2
, 1q
3

,
and x =
2
(p) = ( )(p),
2
(U
2
) = R[0,
1
2
) H
2
. The map
3
is dened
similarly. The boundary is given by C=
1
2
(R{0})
1
3
(R{0}), see Figure
6. I
J 1.16 Example. The open cone M = C = {p = (p
1
, p
2
, p
3
) | p
2
1
+ p
2
2
= p
2
3
, 0
p
3
< 1}, can be described by one coordinate chart, see U
1
above, and is therefore a
manifold without boundary (non-compact), and C is homeomorphic to D
2
, or R
2
,
see denition below. I
So far we have seen manifolds and manifolds with boundary. A manifold can
be either compact or non-compact, which we refer to as closed, or open manifolds
respectively. Manifolds with boundary are also either compact, or non-compact. In
both cases the boundary can be compact. Open subsets of a topological manifold
are called open submanifolds, and can be given a manifold structure again.
Let N and M be manifolds, and let f : N M be a continuous mapping. A
mapping f is called a homeomorphism between N and M if f is continuous and
has a continuous inverse f
1
: MN. In this case the manifolds N and M are said
12
FIGURE 6. Coordinate maps for the cone C.
to be homeomorphic. Using charts (U, ), and (V, ) for N and M respectively,
we can give a coordinate expression for f , i.e.

f = f
1
.
A (topological) embedding is a continuous injective mapping f : X Y, which
is a homeomorphism onto its image f (X) Y with respect to the subspace topol-
ogy. Let f : N M be an embedding, then its image f (N) M is called a subman-
ifold of M. Notice that an open submanifold is the special case when f =i : U ,M
is an inclusion mapping.
J 1.17 Facts. Recall the subspace topology. Let X be a topological space and let
S X be any subset, then the subspace, or relative topology on S (induced by the
topology on X) is dened as follows. A subset U S is open if there exists an open
set V X such that U =V S. In this case S is called a (topological) subspace of
X.
For a subjective map : X Y we topologize Y via: U Y is open if and
only if
1
(U) is open in X. In this case is continuous and is called a quotient
map. For a given topological space Z the map f : Y Z is continuous if and
only if f : X Z is continuous. For an equivalence relation , the mapping
: X X/ is a quotient map, see [1].
If is a open equivalence relation, i.e. the image of an open set under : X
X/ is open, then X/ is Hausdorff if and only if the set {(x, x
0
) X X | x x
0
}
is closed in X X. If X is second countable (countable basis of open sets), then
also X/ is second countable, see [2]. I
13
2. Differentiable manifolds and differentiable structures
A topological manifold M for which the transition maps
i j
=
j

1
i
for all
pairs
i
,
j
in the atlas are diffeomorphisms is called a differentiable, or smooth
manifold. The transition maps are mappings between open subsets of R
m
. Dif-
feomorphisms between open subsets of R
m
are C

-maps, whose inverses are also

C

-maps. For two charts U

i
and U
j
the transitions maps are mappings:

i j
=
j

1
i
:
i
(U
i
U
j
)
j
(U
i
U
j
),
and as such are homeomorphisms between these open subsets of R
m
.
Denition 2.1. AC

-atlas is a set of charts A= {(U,

i
)}
iI
such that
(i) M =
iI
U
i
,
(ii) the transition maps
i j
are diffeomorphisms between
i
(U
i
U
j
) and

j
(U
i
U
j
), for all i 6= j (see Figure 2 ).
The charts in a C

-atlas are said to be C

-compatible. Two C

-atlases Aand A
0
are equivalent if AA
0
is again a C

-atlas, which denes a equivalence relation on

C

-atlases. An equivalence class of this equivalence relation is called a differen-

tiable structure D on M. The collection of all atlases associated with D, denoted
A
D
, is called the maximal atlas for the differentiable structure. Figure 7 shows
why compatibility of atlases denes an equivalence relation.
Denition 2.3. Let M be a topological manifold, and let D be a differentiable
structure on M with maximal atlas A
D
. Then the pair (M, A
D
) is called a (C

-
)differentiable manifold.
Basically, a manifold M with a C

-atlas, denes a differentiable structure on

M. The notion of a differentiable manifold is intuitively a difcult concept. In
dimensions 1 through 3 all topological manifolds allow a differentiable structure
(only one up to diffeomorphisms). Dimension 4 is the rst occurrence of manifolds
without a differentiable structure. Also in higher dimensions uniqueness of differ-
entiable structures is no longer the case as the famous example by Milnor shows;
S
7
has 28 different (non-diffeomorphic) differentiable structures. The rst example
of a manifold that allows many non-diffeomorphic differentiable structures occurs
in dimension 4; exotic R
4
s. One can also consider C
r
-differentiable structures and
manifolds. Smoothness will be used here for the C

-case.
Theorem 2.4.
4
Let M be a topological manifold with a C

-atlas A. Then there

exists a unique differentiable structure D containing A, i.e. AA
D
.
4
Lee, Lemma 1.10.
14
FIGURE 7. Differentiability of
00

1
is achieved via the map-
pings
00
(
0
)
1
, and
0

1
, which are diffeomorphisms since
AA
0
, and A
0
A
00
by assumption. This establishes the equiva-
lence AA
00
.
Proof: Let A be the collection of charts that are C

-compatible with A. By the

same reasoning as in Figure 7 we prove that all charts in Aare C

-compatible with
eachother proving that A is a smooth altas. Now any chart that is C

-compatible
with any chart in A is C

-compatible with any chart in A and is thus in A and A is

a maximal atlas. Clearly any other maximal altas is contained in A and therefore
A=A
D
.
J 2.5 Remark. In our denition of of n-dimensional differentiable manifold we
use the local model over standard R
n
, i.e. on the level of coordinates we express
differentiability with respect to the standard differentiable structure on R
n
, or dif-
feomorphic to the standard differentiable structure on R
n
. I
J 2.6 Example. The cone M =C R
3
as in the previous section is a differentiable
manifold whose differentiable structure is generated by the one-chart atlas (C, )
as described in the previous section. As we will see later on the cone is not a
smoothly embedded submanifold. I
J 2.7 Example. M = S
1
(or more generally M = S
n
). As in Section 1 consider
S
1
with two charts via stereographic projection. We have the overlap U
1
U
2
=
S
1
\{(0, 1)}, and the transition map
12
=
2

1
1
given by y =
12
(x) =
4
x
,
15
x 6= 0, , and y 6= 0, . Clearly,
12
and its inverse are differentiable functions
from
1
(U
1
U
2
) =R\{0} to
2
(U
1
U
2
) =R\{0}. I

FIGURE 8. The transition maps for the stereographic projection.

J 2.9 Example. The real projective spaces PR
n
, see exercises Chapter VI. I
J2.10 Example. The generalizations of projective spaces PR
n
, the so-called (k, n)-
Grassmannians G
k
R
n
are examples of smooth manifolds. I
Theorem 2.11.
5
Given a set M, a collection {U

}
A
of subsets, and injective
mappings

: U

R
m
, such that the following conditions are satised:
(i)

(U

) R
m
is open for all ;
(ii)

(U

) and

(U

) are open in R
m
for any pair , A;
(iii) for U

6=, the mappings

(U

(U

) are
diffeomorphisms for any pair , A;
(iv) countably many sets U

cover M.
(v) for any pair p 6= q M, either p, q U

, or there are disjoint sets U

,U

such that p U

and q U

.
Then M has a unique differentiable manifold structure and (U

) are smooth
charts.
Proof: Let us give a sketch of the proof. Let sets
1

(V), V R
n
open, form a
basis for a topology on M. Indeed, if p
1

(V)
1

(W) then the latter is again

of the same form by (ii) and (iii). Combining this with (i) and (iv) we establish
a topological manifold over R
m
. Finally from (iii) we derive that {(U

)} is a
smooth atlas.
Let N and M be smooth manifolds (dimensions n and m respectively). Let f :
N M be a mapping from N to M.
5
See Lee, Lemma 1.23.
16
Denition 2.12. A mapping f : N M is said to be C

, or smooth if for every

p N there exist charts (U, ) of p and (V, ) of f (p), with f (U) V, such that

f = f
1
: (U) (V) is a C

-mapping (from R
n
to R
m
).
FIGURE 9. Coordinate representation for f , with f (U) V.
The above denition also holds true for mappings dened on open subsets of
N, i.e. let W N is an open subset, and f : W N M, then smoothness on W
is dened as above by restricting to pionts p W. With this denition coordinate
maps : U R
m
are smooth maps.
The denition of smooth mappings allows one to introduce the notion of dif-
ferentiable homeomorphism, of diffeomorphism between manifolds. Let N and M
be smooth manifolds. A C

-mapping f : N M, is called a diffeomorphism if

it is a homeomorphism and also f
1
is a smooth mapping, in which case N and
M are said to be diffeomorphic. The associated differentiable structures are also
called diffeomorphic. Diffeomorphic manifolds dene an equivalence relation. In
the denition of diffeomorphism is sufces require that f is a differentiable bijec-
tive mapping with smooth inverse (see Theorem 2.15). A mapping f : N M is
called a local diffeomorphism if for every p N there exists a neighborhood U,
with f (U) open in M, such that f : U f (U) is a diffeomorphism. A mapping
f : N M is a diffeomorphism if and only if it is a bijective local diffeomorphism.
J 2.14 Example. Consider N = R with atlas (R, (p) = p), and M = R with at-
las (R, (q) = q
3
). Clearly these dene different differentiable structures (non-
compatible charts). Between N and M we consider the mapping f (p) = p
1/3
,
17
which is a homeomorphism between N and M. The claim is that f is also a diffeo-
morphism. Take U =V =R, then f
1
(p) = (p
1/3
)
3
= p is the identity and
thus C

on R, and the same for f

1

1
(q) = (q
3
)
1/3
=q. The associated dif-
ferentiable structures are diffeomorphic. In fact the above described differentiable
structures correspond to dening the differential quotient via lim
h0
f
3
(p+h)f
3
(p)
h
.
I
Theorem 2.15.
6
Let N, M be smooth manifolds with atlases A
N
and A
M
respec-
tively. The following properties hold:
(i) Given smooth maps f

: U

M, for all U

A
N
, with f

|
U

=
f

|
U

for all , . Then there exists a unique smooth map f : N M

such that f |
U

= f

.
(ii) A smooth map f : N M between smooth manifolds is continuous.
(iii) Let f : N M be continuous, and suppose that the maps

f

f
1

,
for charts (U

) A
N
, and (V

) A
M
, are smooth on their domains
for all , . Then f is smooth.
Proof: Dene f by f |
U

= f

, which is well-dened by the overlap conditioins.

Given p M, there exists a chart U

3 p and

f =

f

which is smooth by denition,

and thus f is smooth.
For any p U (chart) and choose f (p) V (chart), then

f is a smooth map and
f |
U
=
1

f : U V is continuous. Continuity holds for each neighborhood
of p M.
Let p U

and f (p) V

and set U = f
1
(V

) U

which is open by
continuity of f . Now

f =

f

on the charts (U,

|
U
) and (V

) which proves
the differentiability of f .
With this theorem at hand we can verify differentiability locally. In order to
show that f is a diffeomorphism we need that f is a homeomorphism, or a bijection
that satises the above local smoothness conditions, as well as for the inverse.
J 2.17 Example. Let N = M = PR
1
and points in PR
1
are equivalence classes
[x] = [(x
1
, x
2
)]. Dene the mapping f : PR
1
PR
1
as [(x
1
, x
2
)] 7[(x
2
1
, x
2
2
)]. Con-
sider the charts U
1
= {x
1
6= 0}, and U
2
= {x
2
6= 0}, with for example
1
(x) =
tan
1
(x
2
/x
1
), then
1
(
1
) =
_
[cos(
1
), sin(
1
)]
_
, with
1
V
1
= (/2, /2). In
local coordinates on V
1
we have

f (
1
) = tan
1
_
sin
2
(
1
)/cos
2
(
1
)
_
,
6
See Lee Lemmas 2.1, 2.2 and 2.3.
18

FIGURE 10. Coordinate diffeomorphisms and their local repre-

sentation as smooth transition maps.
and a similar expression for V
2
. Clearly, f is continuous and the local expressions
also prove that f is a differentiable mapping using Theorem 2.15. I
The coordinate maps in a chart (U, ) are diffeomorphisms. Indeed, : U
MR
n
, then if (V, ) is any other chart with UV 6=, then by the denition the
transition maps
1
and
1
are smooth. For we then have =
1
and
1
=
1
, which proves that is a diffeomorphism, see Figure 10. This
justies the terminology smooth charts.
J 2.18 Remark. For arbitrary subsets U R
m
and V R
n
a map f : U R
m
V R
n
is said to be smooth at p U, if there exists an open set U

R
m
containing p, and a smooth map f

: U

R
m
, such that f and f

coincide on
U U

. The latter is called an extension of f . A mapping f : U V between

arbitrary subsets U R
m
and V R
n
is called a diffeomorphism if f maps U
homeomorphically onto V and both f and f
1
are is smooth (as just described
above). In this case U and V are said to be diffeomorphic. I
J 2.19 Example. An important example of a class of differentiable manifolds are
appropriately chosen subsets of Euclidean space that can be given a smooth mani-
fold structure. Let M R
`
be a subset such that every p M has a neighborhood
U 3 p in M (open in the subspace topology, see Section 3) which is diffeomorphic
to an open subset V R
m
(or, equivalently an open disc D
m
R
m
). In this case the
set M is a smooth m-dimensional manifold. Its topology is the subsapce topology
and the smooth structure is inherited from the standard smooth structure on R
`
,
which can be described as follows. By denition a coordinate map is a diffeo-
19

FIGURE 11. The transitions maps

1
and
1
smooth
mappings establishing the smooth structure on M.
morphisms which means that : U V =(U) is a smooth mapping (smoothness
via a smooth map

), and for which also

1
is a smooth map. This then directly
implies that the transition maps
1
and
1
are smooth mappings, see
Figure 11. In some books this construction is used as the denition of a smooth
manifold. I
J 2.21 Example. Let us consider the cone M =C described in Example 1 (see also
Example 2 in Section 1). We already established that C is manifold homeomorphic
to R
2
, and moreover C is a differentiable manifold, whose smooth structure is
dened via a one-chart atlas. However, C is not a smooth manifold with respect to
the induced smooth structure as subset of R
3
. Indeed, following the denition in the
above remark, we have U =C, and coordinate homeomorphism(p) = (p
1
, p
2
) =
x. By the denition of smooth maps it easily follows that is smooth. The inverse
is given by
1
(x) =
_
x
1
, x
2
,
_
x
2
1
+x
2
2
_
, which is clearly not differentiable at the
cone-top (0, 0, 0). The cone C is not a smoothly embedded submainfold of R
3
(topological embedding). I
Let U,V andW be open subsets of R
n
, R
k
and R
m
respectively, and let f : U V
and g : V W be smooth maps with y = f (x), and z = g(y). Then the Jacobians
are
J f |
x
=
_
_
_
_
f
1
x
1

f
1
x
n
.
.
.
.
.
.
.
.
.
f
k
x
1

f
k
x
n
_
_
_
_
, Jg|
y
=
_
_
_
_
g
1
y
1

g
1
y
k
.
.
.
.
.
.
.
.
.
g
m
y
1

g
m
y
k
_
_
_
_
,
and J(g f )|
x
= Jg|
y=f (x)
J f |
x
(chain-rule). The commutative diagram for the
maps f , g and g f yields a commutative diagram for the Jacobians:
20
V
U W
A
A
A
A
A
AU
g

f
-
g f
R
k
R
n
R
m
A
A
A
A
A
AU
Jg|
y

J f |
x
-
J(g f )|
x
For diffeomorphisms between open sets in R
n
we have a number of important
properties. Let U R
n
and V R
m
be open sets, and let f : U V is a diffeomor-
phism, then n = m, and J f |
x
is invertible for any x U. Using the above described
commutative diagrams we have that f
1
f yields that J( f
1
)|
y=f (x)
J f |
x
is the
identity on R
n
, and f f
1
yields that J f |
x
J( f
1
)|
y=f (x)
is the identity on R
m
.
Thus J f
x
has an inverse and consequently n = m. Conversely, if f : U R
n
,
U R
n
, open, then we have the Inverse Function Theorem;
Theorem 2.22. If J f |
x
: R
n
R
n
is invertible, then f is a diffeomorphism between
sufciently small neighborhoods U
0
and f (U
0
) of x and y respectively.
3. Immersions, submersions and embeddings
Let N and M be smooth manifolds of dimensions n and m respectively, and let
f : N M be a smooth mapping. In local coordinates

f = f
1
: (U)
(V), with respects to charts (U, ) and (V, ). The rank of f at p N is dened
as the rank of

f at (p), i.e. rk( f )|
p
= rk(J

f )|
(p)
, where J

f |
(p)
is the Jacobian
of f at p:
J

f |
x=(p)
=
_
_
_
_

f
1
x
1



f
1
x
n
.
.
.
.
.
.
.
.
.

f
m
x
1



f
m
x
n
_
_
_
_
This denition is independent of the chosen charts, see Figure 12. Via the commu-
tative diagram in Figure 12 we see that

f = (
0

1
)

f (
0

1
)
1
, and by the
chain rule J

f |
x
0 =J(
0

1
)|
y
J

f |
x
J(
0

1
)
1
|
x
0 . Since
0

1
and
0

1
are diffeomorphisms it easily follows that rk(J

f )|
x
= rk(J

f )|
x
0 , which shows that
our notion of rank is well-dened. If a map has constant rank for all p N we sim-
ply write rk( f ). These are called constant rank mappings. Let us now consider
the various types of constant rank mappings between manifolds.
21
FIGURE 12. Representations of f via different coordinate charts.
Denition 3.2. A mapping f : N M is called an immersion if rk( f ) = n, and a
submersion if rk( f ) = m. An immersion that is injective,
7
or 1-1, and is a home-
omorphism onto (surjective mapping
8
) its image f (N) M, with respect to the
subspace topology, is called a smooth embedding.
A smooth embedding is an (injective) immersion that is a topological embed-
ding.
Theorem 3.3.
9
Let f : N M be smooth with constant rank rk( f ) = k. Then for
each p N, and f (p) M, there exist coordinates (U, ) for p and (V, ) for f (p),
with f (U) V, such that
( f
1
)(x
1
, x
k
, x
k+1
, x
n
) = (x
1
, x
k
, 0, , 0).
J 3.4 Example. Let N = (

2
,
3
2
), and M = R
2
, and the mapping f is given by
f (t) = (sin(2t), cos(t)). In Figure 13 we displayed the image of f . The Jacobian is
given by
J f |
t
=
_
2cos(2t)
sin(t)
_
Clearly, rk(J f |
t
) = 1 for all t N, and f is an injective immersion. Since N is
an open manifold and f (N) M is a compact set with respect to the subspace
7
A mapping f : N M is injective if for all p, p
0
N for which f (p) = f (p
0
) it holds that p = p
0
.
8
A mapping f : N M is surjective if f (N) = M, or equivalently for every q M there exists a
p N such that q = f (p).
9
See Lee, Theorem 7.8 and 7.13.
22
topology, it follows f is not a homeomorphism onto f (N), and therefore is not an
embedding. I
FIGURE 13. Injective parametrization of the gure eight.
J 3.6 Example. Let N = S
1
be dened via the atlas A= {(U
i
,
i
)}, with
1
1
(t) =
((cos(t), sin(t)), and
1
2
(t) = ((sin(t), cos(t)), and t (

2
,
3
2
). Furthermore, let
M = R
2
, and the mapping f : N M is given in local coordinates; in U
1
as in
Example 1. Restricted to S
1
R
2
the map f can also be described by
f (x, y) = (2xy, x).
This then yields for U
2
that

f (t) = (sin(2t), sin(t)). As before rk( f ) = 1, which
shows that f is an immersion of S
1
. However, this immersion is not injective at
the origin in R
2
, indicating the subtle differences between these two examples, see
Figures 13 and 14. I
FIGURE 14. Non-injective immersion of the circle.
J 3.8 Example. Let N =R, M =R
2
, and f : N M dened by f (t) = (t
2
, t
3
). We
can see in Figure 15 that the origin is a special point. Indeed, rk(J f )|
t
= 1 for all
t 6= 0, and rk(J f )|
t
= 0 for t = 0, and therefore f is not an immersion. I
J 3.10 Example. Consider M = PR
n
. We established PR
n
as smooth manifolds.
For n = 1 we can construct an embedding f : PR
1
R
2
as depicted in Figure 16.
For n = 2 we nd an immersion f : PR
2
R
3
as depicted in Figure 17. I
23

FIGURE 15. The map f fails to be an immersion at the origin.

FIGURE 16. PR is diffeomorphic to S
1
.
FIGURE 17. Identications for PR
2
giving an immersed non-
orientable surface in R
3
.
J 3.13 Example. Let N =R
2
and M =R, then the projection mapping f (x, y) = x
is a submersion. Indeed, J f |
(x,y)
= (1 0), and rk(J f |
(x,y)
) = 1. I
24
J3.14 Example. Let N =M =R
2
and consider the mapping f (x, y) = (x
2
, y)
t
. The
Jacobian is
J f |
(x,y)
=
_
2x 0
0 1
_
,
and rk(J f )|
(x,y)
= 2 for x 6= 0, and rk(J f )|
(x,y)
= 1 for x = 0. See Figure 18. I
FIGURE 18. The projection (submersion), and the folding of the
plane (not a submersion).
J 3.16 Example. We alter Example 2 slightly, i.e. N = S
1
R
2
, and M =R
2
, and
again use the atlas A. We now consider a different map f , which, for instance on
U
1
, is given by

f (t) = (2cos(t), sin(t)) (or globally by f (x, y) = (2x, y)). It is clear
that f is an injective immersion, and since S
1
is compact it follows from Lemma
3.18 below that S
1
is homeomorphic to it image f (S
1
), which show that f is a
smooth embedding (see also Appendix in Lee). I
J 3.17 Example. Let N = R, M = R
2
, and consider the mapping f (t) =
(2cos(t), sin(t)). As in the previous example f is an immersion, not injective how-
ever. Also f (R) = f (S
1
) in the previous example. The manifold N is the universal
covering of S
1
and the immersion f descends to a smooth embedding of S
1
. I
Lemma 3.18.
10
Let f : N M be an injective immersion. If
(i) N is compact, or if
(ii) f is a proper map,
11
then f is a smooth embedding.
Proof: For (i) we argue as follows. Since N is compact any closed subset X N
is compact, and therefore f (X) M is compact and thus closed; f is a closed
10
Lee, Prop. 7.4, and pg. 47.
11
A(ny) map f : N M is called proper if for any compact K M, it holds that also f
1
(K) N
is compact. In particular, when N is compact, continuous maps are proper.
25
mapping.
12
We are done if we show that f is a topological embedding. By the
assumption of injectivity, f : N f (N) is a bijection. Let X N be closed, then
( f
1
)
1
(X) = f (X) f (N) is closed with respect to the subspace topology, and
thus f
1
: f (N) N is continuous, which proves that f is a homeomorphism onto
its image and thus a topological embedding.
A straightforward limiting argument shows that proper continuous mappings
bewteen manifolds are closed mappings (see Exercises).
Let us start with dening the notion of embedded submanifolds.
Denition 3.19. A subset N M is called a smooth embedded n-dimensional
submanifold in M if for every p N, there exists a chart (U, ) for M, with p U,
such that
(U N) =(U)
_
R
n
{0}
_
= {x (U) : x
n+1
= = x
m
= 0}.
The co-dimension of N is dened as codimN = dimMdimN.
The set W = U N in N is called a n-dimensional slice, or n-slice of U, see
Figure 19, and (U, ) a slice chart. The associated coordinates x = (x
1
, , x
n
) are
called slice coordinates. Embedded submanifolds can be characterized in terms of
embeddings.
FIGURE 19. Take a k-slice W. On the left the image (W) corre-
sponds to the Eucliden subspace R
k
R
m
.
Theorem 3.21.
13
Let N M be a smooth embedded n-submanifold. Endowed
with the subspace topology, N is a n-dimensional manifold with a unique (induced)
smooth structure such that the inclusion map i : N ,M is an embedding (smooth).
12
A mapping f : N M is called closed if f (X) is closed in M for any closed set X N.
Similarly, f is called open if f (X) is open in M for every open set X N.
13
See Lee, Thms 8.2
26
To get an idea let us show that N is a topological manifold. Axioms (i) and
(iii) are of course satised. Consider the projection : R
m
R
n
, and an inclusion
j : R
n
R
m
dened by
(x
1
, , x
n
, x
n+1
, , x
m
) = (x
1
, , x
n
),
j(x
1
, , x
n
) = (x
1
, , x
n
, 0, , 0).
Now set Z = ( )(W) R
n
, and = ( )|
W
, then
1
= (
1
j)|
Z
, and
: W Z is a homeomorphism. Therefore, pairs (W, ) are charts for N, which
form an atlas for N. The inclusion i : N ,M is a topological embedding.
Given slice charts (U, ) and (U
0
,
0
) and associated charts (W, ) and (W
0
,
0
)
for N. For the transitions maps it holds that
1
=
0

1
j, which are
diffeomorphisms, which denes a smooth atlas for N. The inclusion i : N ,M can
be expressed in local coordinates;

i =i
1
, (x
1
, , x
n
) 7(x
1
, , x
n
, 0, , 0),
which is an injective immersion. Since i is also a topological embedding it is thus
a smooth embedding. It remains to prove that the smooth structure is unique, see
Lee Theorem 8.2.
Theorem 3.22.
14
The image of an embedding is a smooth embedded submanifold.
Proof: By assumption rk( f ) =n and thus for any p N it follows from Theorem
3.3 that

f (x
1
, , x
n
) = (x
1
, , x
n
, 0, , 0),
for appropriate coordinates (U, ) for p and (V, ) for f (p), with f (U) V. Con-
sequently, f (U) is a slice in V, since ( f (U) satises Denition 3.19. By assump-
tion f (U) is open in f (N) and thus f (U) = A f (N) for some open set A M. By
replacing V by V
0
= AV and restricting to V
0
, (V
0
, ) is a slice chart with slice
V
0
f (N) =V
0
f (U).
Summarizing we conclude that embedded submanifolds are the images of
smooth embeddings.
A famous result by Whitney says that considering embeddings into R
m
is not
not really a restriction for dening smooth manifolds.
Theorem 3.23.
15
Any smooth n-dimensional manifold M can be (smoothly) em-
bedded into R
2n+1
.
14
See Lee, Thms 8.3
15
See Lee, Ch. 10.
27
A subset N M is called an immersed submanifold if N is a smooth n-
dimensional manifold, and the mapping i : N ,M is a (smooth) immersion. This
means that we can endow N with an appropriate manifold topology and smooth
structure, such that the natural inclusion of N into M is an immersion. If f : N M
is an injective immersion we can endow f (N) with a topology and unique smooth
structure; a set U f (N) is open if and only if f
1
(U) N is open, and the
(smooth) coordinate maps are taken to be f
1
, where s are coordinate maps
for N. This way f : N f (N) is a diffeomorphism, and i : f (N) ,M an injective
immersion via the composition f (N) N M. This proves:
Theorem 3.24.
16
Immersed submanifolds are exactly the images of injective im-
mersions.
We should point out that embedded submanifolds are examples of immersed
submanifolds, but not the other way around. For any immersion f : N M, the
image f (N) is called an immersed manifold in M.
In this setting Whitney established some improvements of Theorem 3.23.
Namely, for dimension n > 0, any smooth n-dimensional manifold can be em-
bedded into R
2n
(e.g. the embedding of curves in R
2
). Also, for n > 1 any smooth
n-dimensional manifold can be immersed into R
2n1
(e.g. the Klein bottle). In
this course we will often think of smooth manifolds as embedded submanifolds of
R
m
. An important tool thereby is the general version of Inverse Function Theorem,
FIGURE 20. An embedding of R [left], and an immersion of S
2
[right] called the Klein bottle.
which can easily be derived from the Euclidean version 2.22.
Theorem 3.26.
17
Let N, M be smooth manifolds, and f : N M is a smooth
mapping. If, at some point p N, it holds that J

f |
(p)=x
is an invertible matrix,
16
See Lee, Theorem 8.16.
17
See Lee, Thms 7.6 and 7.10.
28
then there exist sufciently small neighborhoods U
0
3 p, and V
0
f (p) such that
f |
U
0
: U
0
V
0
is a diffeomorphism.
As a direct consequence of this result we have that if f : N M, with dimN =
dimM, is an immersion, or submersion, then f is a local diffeomorphism. If f is a
bijection, then f is a (global) diffeomorphism.
Theorem 3.27. Let f : N M be a constant rank mapping with rk( f ) = k. Then
for each q f (N), the level set S = f
1
(q) is an embedded submanifold in N with
co-dimension equal to k.
Proof: Clearly, by continuity S is closed in N. By Theorem 3.3 there are coor-
dinates (U, ) of p S and (V, ) of f (p) = q, such that

f (x
1
, , x
k
, x
k+1
, , x
n
) = (x
1
, , x
k
, 0, , 0) =(q) = 0.
The points in SU are characterized by (SU) = {x | (x
1
, , x
k
) = 0}. There-
fore, SU is a (nk)-slice, and thus S is a smooth submanifold in M of codimen-
sion k.
In particular, when f : N M is a submersion, then for each q f (N), the
level set S = f
1
(q) is an embedded submanifold of co-dimension codimS = m =
dimM. In the case of maximal rank this statement can be restricted to just one
level. A point p N is called a regular point if rk( f )|
p
= m = dimM, otherwise
a point is called a critical point. A value q f (N) is called a regular value if all
points p f
1
(q) are regular points, otherwise a value is called a critical value. If
q is a regular value, then f
1
(q) is called a regular level set.
Theorem 3.28. Let f : N M be a smooth map. If q f (N) is a regular value,
then S = f
1
(q) is an embedded submanifold of co-dimension equal to dimM.
Proof: Let us illustrate the last result for the important case N =R
n
and M =R
m
.
For any p S = f
1
(q) the Jacobian J f |
p
is surjective by assumption. Denote the
kernel of J f |
p
by ker J f |
p
R
n
, which has dimension nm. Dene
g : N =R
n
R
nm
R
m

=R
n
,
by g() = (L, f () q)
t
, where L : N = R
n
R
nm
is any linear map which
is invertible on the subspace ker J f |
p
R
n
. Clearly, Jg|
p
= L J f |
p
, which, by
construction, is an invertible (linear) map on R
n
. Applying Theorem 2.22 (Inverse
Function Theorem) to g we conclude that a sufciently small neighborhood of U
of p maps diffeomorphically onto a neighborhood V of (L(p), 0). Since g is a
diffeomorphism it holds that g
1
maps
_
R
nm
{0}
_
V onto f
1
(q)U (the 0
R
m
corresponds to q). This exactly says that every point p S allows an (nm)-
slice and is therefore an (n m)-dimensional submanifold in N = R
n
(codimS =
m).
29
FIGURE 21. The map g yields 1-slices for the set S.
J 3.30 Example. Let us start with an explicit illustration of the above proof. Let
N = R
2
, M = R, and f (p
1
, p
2
) = p
2
1
+ p
2
2
. Consider the regular value q = 2, then
J f |
p
= (2p
1
2p
2
), and f
1
(2) = {p : p
2
1
+ p
2
2
= 2}, the circle with radius

2.
We have ker J f |
p
=span{(p
1
, p
2
)
t
}, and is always isomorphic to R. For example
x the point (1, 1) S, then ker J f |
p
= span{(1, 1)
t
} and dene
g() =
_
L()
f () 2
_
=
_

1

2
1
+
2
2
2
_
, Jg|
p
=
_
1 1
2 2
_
where the linear map is L = (1 1). The map g is a local diffeomorphism and
on SU this map is given by
g(
1
,
_
2
2
1
) =
_

1

_
2
2
1
0
_
,
with
1
(1, 1+). The rst component has derivative 1+

1

2
2
1
, and there-
fore SU is mapped onto a set of the form
_
R{0}
_
V. This procedure can be
carried out for any point p S, see Figure 21. I
J3.31 Example. Let N =R
2
\{(0, 0)}R=R
3
\{(0, 0, )}, M = (1, )(0, ),
and
f (x, y, z) =
_
x
2
+y
2
1
1
_
, with J f |
(x,y,z)
=
_
2x 2y 0
0 0 0
_
We see immediately that rk( f ) = 1 on N. This map is not a submersion, but is of
constant rank, and therefore for any value q f (N) M it holds that S = f
1
(q)
is a embedded submanifold, see Figure 22. I
J 3.33 Example. Let N, M as before, but now take
f (x, y, z) =
_
x
2
+y
2
1
z
_
, with J f |
(x,y,z)
=
_
2x 2y 0
0 0 1
_
30
FIGURE 22. An embedding of a cylinder via a constant rank map-
ping.
Now rk( f ) = 2, and f is a submersion. For every q M, the set S = f
1
(q) is a
embedded submanifold, see Figure 23. I
FIGURE 23. An embedding circle of an submersion.
J3.35 Example. Let N =R
2
, M =R, and f (x, y) =
1
4
(x
2
1)
2
+
1
2
y
2
. The Jacobian
is J f |
(x,y)
= (x(x
2
1) y). This is not a constant rank, nor a submersion. However,
any q > 0, q 6=
1
4
, is a regular value since then the rank is equal to 1. Figure 24
shows the level set f
1
(0) (not an embedded manifold), and f
1
(1) (an embedded
circle) I
Theorem 3.37.
18
Let S M be a subset of a smooth m-dimensional manifold M.
Then S is a k-dimensional smooth embedded submanifold if and if for every p S
there exists a neighborhood U 3 p such that U S is the level set of a submersion
f : U R
mk
.
Proof: Assume that S M is a smooth embedded manifold. Then for each
p S there exists a chart (U, ), p U, such that (S U) = {x : x
k+1
= =
18
See Lee, Prop.8.12.
31
FIGURE 24. A regular and critical level set.
x
m
= 0}. Clearly, S U is the sublevel set of f : U R
mk
at 0, given by

f (x) =
(x
k+1
, , x
m
), which is a submersion.
Conversely, if S U = f
1
(0), for some submersion f : U R
mk
, then by
Theorem 3.28, S U is an embedded submanifold of U. This clearly shows that S
is an embedded submanifold in M.
32
II. Tangent and cotangent
spaces
4. Tangent spaces
For a(n) (embedded) manifold M R
`
the tangent space T
p
M at a point p
M can be pictured as a hyperplane tangent to M. In Figure 25 we consider the
parametrizations x+te
i
in R
m
. These parametrizations yield curves
i
(t) =
1
(x+
te
i
) on M whose velocity vectors are given by

0
(0) =
d
dt

1
(x +te
i
)

t=0
= J
1
|
x
(e
i
).
The vectors p+
0
(0) are tangent to M at p and span an m-dimensional afne linear
subspace M
p
of R
`
. Since the vectors J
1
|
x
(e
i
) span T
p
M the afne subspace is
given by
M
p
:= p+T
p
M R
`
,
which is tangent to M at p.
FIGURE 25. Velocity vectors of curves of M span the tangent
space.
33
The considerations are rather intuitive in the sense that we consider only em-
bedded manifolds, and so the tangent spaces are tangent m-dimensional afne sub-
spaces of R
`
. One can also dene the notion of tangent space for abstract smooth
manifolds. There are many ways to do this. Let us describe one possible way (see
e.g. Lee, or Abraham, Marsden and Ratiu) which is based on the above considera-
tions.
Let a < 0 < b and consider a smooth mapping : I = (a, b) RM, such that
(0) = p. This mapping is called a (smooth) curve on M, and is parametrized by
t I. If the mapping (between the manifolds N = I, and M) is an immersion,
then is called an immersed curve. For such curves the velocity vector J |
t
=
()
0
(t) in R
m
is nowhere zero. We build the concept of tangent spaces in order
to dene the notion velocity vector to a curve .
Let (U, ) be a chart at p. Then, two curves and

are equivalent,

, if

(0) =(0) = p, and (

)
0
(0) = ()
0
(0).
The equivalence class of a curve through p M is denoted by [].
Denition 4.2.
19
At a p M dene the tangent space T
p
M as the space of all
equivalence classes [] of curves through p. A tangent vector X
p
, as the equiva-
lence class of curves, is given by
X
p
:= [] =
_

(0) =(0) = p, (

)
0
(0) = ()
0
(0)
_
,
which is an element of T
p
M.

FIGURE 26. Immersed curves and velocity vectors in R

m
.
19
Lee, Ch. 3.
34
The above denition does not depend on the choice of charts at p M. Let
(U
0
,
0
) be another chart at p M. Then, using that (

)
0
(0) = ( )
0
(0), for
(
0
)
0
(0) we have
(
0
)
0
(0) =
_
(
0

1
) ()
_
0
(0)
= J(
0

1
)

x
()
0
(0)
= J(
0

1
)

x
(

)
0
(0)
=
_
(
0

1
) (

)
_
0
(0) = (
0

)
0
(0),
which proves that the equivalence relation does not depend on the particular choice
of charts at p M.
One can prove that T
p
M

= R
m
. Indeed, T
p
M can be given a linear structure as
follows; given two equivalence classes [
1
] and [
2
], then
[
1
] +[
2
] :=
_
: ()
0
(0) = (
1
)
0
(0) +(
2
)
0
(0)
_
,
[
1
] :=
_
: ()
0
(0) =(
1
)
0
(0)
_
.
The above argument shows that these operation are well-dened, i.e. indepen-
dent of the chosen chart at p M, and the operations yield non-empty equivalence
classes. The mapping

: T
p
M R
m
,

_
[]
_
= ()
0
(0),
is a linear isomorphism and

0 = J(
0

1
)|
x

. Indeed, by considering curves

i
(x) =
1
(x +te
i
), i = 1, ..., m, it follows that [
i
] 6= [
j
], i 6= j, since
(
i
)
0
(0) = e
i
6= e
j
= (
j
)
0
(0).
This proves the surjectivity of

. As for injectivity one argues as follows. Suppose,

()
0
(0) = (

)
0
(0), then by denition [] = [

], proving injectivity.
Given a smooth mapping f : N M we can dene how tangent vectors in T
p
N
are mapped to tangent vectors in T
q
M, with q = f (p). Choose charts (U, ) for
p N, and (V, ) for q M. We dene the tangent map or pushforward of f as
follows, see Figure 27. For a given tangent vector X
p
= [] T
p
N,
d f
p
= f

: T
p
N T
q
M, f

([]) = [ f ].
35
FIGURE 27. Tangent vectors in X
p
T
p
N yield tangent vectors
f

X
p
T
q
M under the pushforward of f .
The following commutative diagram shows that f

is a linear map and its denition

does not depend on the charts chosen at p N, or q M.
T
p
N
f

T
q
M

R
n
J( f
1
)|
x
R
m
Indeed, a velocity vector ()
0
(0) is mapped to ( f ())
0
(0), and
( f ())
0
(0) = ( f
1
)
0
(0) = J

f

x
()
0
(0).
Clearly, this mapping is linear and independent of the charts chosen.
If we apply the denition of pushforward to the coordinate mapping : N R
n
,
then

can be identied with

, and J( f
1
)|
x
with ( f
1
)

. Indeed,

([]) = ( )
0
(0) and

([]) = [ ], and in R
n
the equivalence class can be
labeled by ()
0
(0). The labeling map is given as follows

id
([]) = ()
0
(0),
and is an isomorphism, and satises the relations

id

=
1
id

.
36
From now one we identify T
x
R
n
with R
n
by identifying

and

. This justies
the notation

([]) = [] := ()
0
(0).
Properties of the pushforward can be summarized as follows:
Lemma 4.5.
20
Let f : N M, and g : MP be smooth mappings, and let p M,
then
(i) f

: T
p
N T
f (p)
M, and g

: T
f (p)
M T
(g f )(p)
P are linear maps (homo-
morphisms),
(ii) (g f )

= g

: T
p
N T
(g f )(p)
P,
(iii) (id)

= id : T
p
N T
p
N,
(iv) if f is a diffeomorphism, then the pushforward f

is a isomorphism from
T
p
N to T
f (p)
M.
Proof: We have that f

([]) = [ f ], and g

([ f ]) = [g f ], which denes
the mapping (g f )

([]) = [g f ]. Now
[g f ] = [g( f )] = g

([ f ]) = g

( f

([])),
which shows that (g f )

= g

.
A parametrization
1
: R
m
M coming from a chart (U, ) is a local dif-
feomorphism, and can be used to nd a canonical basis for T
p
M. Choosing local
coordinates x = (x
1
, , x
n
) = (p), and the standard basis vectors e
i
for R
m
, we
dene

x
i

p
:=
1

(e
i
).
By denition

x
i

p
T
p
M, and since the vectors e
i
form a basis for R
m
, the vectors

x
i

p
form a basis for T
p
M. An arbitrary tangent vector X
p
T
p
M can now be
written with respect to the basis {

x
i

p
}:
X
p
=
1

(X
i
e
i
) = X
i

x
i

p
.
where the notation X
i

x
i

p
=
i
X
i

x
i

p
denotes the Einstein summation convention,
and (X
i
) is a vector in R
m
!
We now dene the directional derivative of a smooth function h : M R in the
direction of X
p
T
p
M by
X
p
h := h

X
p
= [h].
20
Lee, Lemma 3.5.
37
In fact we have that X
p
h = (h)
0
(0), with X
p
= []. For the basis vectors of T
p
M
this yields the following. Let
i
(t) =
1
(x +te
i
), and X
p
=

x
i

p
, then
(2) X
p
h =

x
i

p
h =
_
(h
1
) (
i
)
_
0
(0) =

h
x
i
,
in local coordinates, which explains the notation for tangent vectors. In particular,
for general tangent vectors X
p
, X
p
h = X
i

h
x
i
.
Let go back to the curve : N = (a, b) M and express velocity vectors. Con-
sider the chart (N, id) for N, with coordinates t, then
d
dt

t=0
:= id
1

(1).
We have the following commuting diagrams:
N

M
id

R
=
R
m
T
t
N

T
p
M
id

R
m
We now dene

0
(0) =

d
dt

t=0

=
1

_
()
0
(0)
_
T
p
M,
by using the second commuting diagram.
If we take a closer look at Figure 27 we can see that using different charts at
p M, or equivalently, considering a change of coordinates leads to the following
relation. For the charts (U, ) and (U
0
,
0
) we have local coordinates x = (p)
and x
0
=
0
(p), and p U U
0
. This yields two different basis for T
p
M, namely
_

x
i

p
_
, and
_

x
0
i

p
_
. Consider the identity mapping f =id : MM, and the push-
forward yields the identity on T
p
M. If we use the different choices of coordinates
as described above we obtain

id

x
i

h = (
0

1
)

h
x
0
j
=
x
0
j
x
i

h
x
0
j
.
In terms of the different basis for T
p
M this gives

x
i

p
=
x
0
j
x
i

x
0
j

p
.
Let us prove this formula by looking a slightly more general situation. Let N, M
be smooth manifolds and f : N M a smooth mapping. Let q = f (p) and (V, )
38
is a chart for M containing q. The vectors

y
j

q
form a basis for T
q
M. If we write
X
p
= X
i

x
i

p
, then for the basis vectors

x
i

p
we have

x
i

h =

x
i

p
(h f ) =

x
i
_

h f
_
=

x
i
_
h f
1
_
=

x
i
_

h f
1
_
=

x
i
_

h

f
_
=

h
y
j

f
j
x
i
=

f
j
x
i

y
j

h
which implies that T
p
N is mapped to T
q
M under the map f

. In general a tangent
vector X
i

x
i

p
is pushed forward to a tangent vector
(3) Y
j

y
j

q
=
_

f
j
x
i
X
i
_

y
j

q
T
q
M,
expressed in local coordinates. By taking N = M and f = id we obtain the above
change of variables formula.
5. Cotangent spaces
In linear algebra it is often useful to study the space of linear functions on a
given vector space V. This space is denoted by V

and called the dual vector space

to V again a linear vector space. So the elements of V

are linear functions

: V R. As opposed to vectors v V, the elements, or vectors in V

are called
covectors.
Lemma 5.1. Let V be a n-dimensional vector space with basis {v
1
, , v
n
}, then
the there exist covectors {
1
, ,
n
} such that

i
v
j
:=
i
(v
j
) =
i
j
, Kronecker delta,
and the covectors {
1
, ,
n
} form a basis for V

.
This procedure can also be applied to the tangent spaces T
p
M described in the
previous chapter.
Denition 5.2. Let M be a smooth m-dimensional manifold, and let T
p
M be the
tangent space at some p M. The the cotangent space T

p
M is dened as the dual
vector space of T
p
M, i.e.
T

p
M := (T
p
M)

.
39
By denition the cotangent space T

p
M is also m-dimensional and it has a canon-
ical basis as described in Lemma 5.1. As before have the canonical basis vectors

x
i
|
p
for T
p
M, the associated basis vectors for T

p
M are denoted dx
i
|
p
. Let us now
describe this dual basis for T

p
M and explain the notation.
The covectors dx
i
|
p
are called differentials and we show now that these are
indeed related dh
p
. Let h : M R be a smooth function, then h

: T
p
M R
and h

p
M. Since the differentials dx
i
|
p
form a basis of T

p
M we have that
h

=
i
dx
i
|
p
, and therefore
h

x
i

p
=
j
dx
j
|
p

x
i

p
=
j

j
i
=
i
=

h
x
i
,
and thus
(4) dh
p
= h

h
x
i
dx
i

p
.
Choose h such that h

satises the identity in Lemma 5.1, i.e. let

h = x
i
( h = x
i

= h, e
i
i). These linear functions h of course span T

p
M, and
h

= (x
i
)

= d(x
i
)
p
= dx
i
|
p
.
Cotangent vectors are of the form

p
=
i
dx
i

p
.
The pairing between a tangent vector X
p
and a cotangent vector
p
is expressed
component wise as follows:

p
X
p
=
i
X
j

i
j
=
i
X
i
.
In the case of tangent spaces a mapping f : N M pushes forward to a linear
f

; T
p
N T
q
M for each p N. For cotangent spaces one expects a similar con-
struction. Let q = f (p), then for a given cotangent vector
q
T

q
M dene
( f

q
) X
p
=
q
( f

X
p
) T

p
N,
for any tangent vector X
p
T
p
N. The homomorphism f

: T

q
MT

p
N, dened by

q
7 f

q
is called the pullback of f at p. It is a straightforward consequence from
linear algebra that f

dened above is indeed the dual homomorphism of f

, also
called the adjoint, or transpose (see Lee, Ch. 6, for more details, and compare the
the denition of the transpose of a matrix). If we expand the denition of pullback
40
in local coordinates, using (3), we obtain

dy
j

X
i

x
i

p
= dy
j

q
f

X
i

x
i

= dy
j

q
_

f
j
x
i
X
i
_

y
j

q
=


f
j
x
i
X
i
Using this relation we obtain that

j
dy
j

X
i

x
i

p
=
j

f
j
x
i
X
i
=
j

f
j
x
i
dx
i

p
X
i

x
i

p
,
which produces the local formula
(5) f

j
dy
j

q
=
_

f
j
x
i
_
dx
i

p
=
_

f


f
j
x
i
_
x
dx
i

p
.
Lemma 5.3. Let f : N M, and g : M P be smooth mappings, and let p M,
then
(i) f

: T

f (p)
M T

p
N, and g

: T

(g f )(p)
P T

f (p)
M are linear maps (homo-
morphisms),
(ii) (g f )

= f

: T

(g f )(p)
P T

p
N,
(iii) (id)

= id : T

p
N T

p
N,
(iv) if f is a diffeomorphism, then the pullback f

is a isomorphismfromT

f (p)
M
to T

p
N.
Proof: By denition of the pullbacks of f and g we have
f

q
X
p
=
q
f

(X
p
), g

g(q)
Y
q
=
g(q)
g

(Y
q
).
For the composition g f it holds that (g f )

(gf )(p)
X
p
=
(gf )(p)
(g f )

(X
p
).
Using Lemma 4.5 we obtain
(g f )

(g f )(p)
X
p
=
(g f )(p)
g

_
f

(X
p
)
_
= g

(gf )(p)
f

(X
p
) = f

_
g

(g f )(p)
_
X
p
,
which proves that (g f )

= f

.
Now consider the coordinate mapping : U M R
m
, which is local diffeo-
morphism. Using Lemma 5.3 we then obtain an isomorphism

: R
m
T

p
M,
which justies the notation
dx
i
|
p
=

(e
i
).
41
6. Vector bundles
The abstract notion of vector bundle consists of topological spaces E (total
space) and M (the base) and a projection : E M (surjective). To be more
precise:
Denition 6.1. A triple (E, M, ) is called a real vector bundle of rank k over M if
(i) for each p M, the set E
p
=
1
(p) is a real k-dimensional linear vector
space, called the ber over p, such that
(ii) for every p M there exists a open neighborhood U 3 p, and a homeomor-
phism :
1
(U) U R
k
;
(a)
_

1
_
(p, ) = p, for all R
k
;
(b) 7
1
(p, ) is a vector space isomorphism between R
k
and E
p
.
The homeomorphism is called a local trivialization of the bundle.
If there is no ambiguity about the base space M we often denote a vector bundle
by E for short. Another way to denote a vector bundle that is common in the
literature is : E M. It is clear from the above denition that if M is a topological
manifold then so is E. Indeed, via the homeomorphisms it follows that E
FIGURE 28. Charts in a vector bundle E over M.
is Hausdorff and has a countable basis of open set. Dene =
_
Id
R
k
_
,
and :
1
(U) V R
k
is a homeomorphism. Figure 28 explains the choice of
bundle charts for E related to charts for M.
For two trivializations :
1
(U) UR
k
and :
1
(V) V R
k
, we have
that the transition map
1
: (U V) R
k
(U V) R
k
has the following
42
form

1
(p, ) = (p, (p))),
where : U V Gl(k, R) is continuous. It is clear from the assumptions that
FIGURE 29. Transition mappings for local trivializations.

1
(p, ) = (p, (p, )). By assumption we also have that 7
1
(p, ) =
A(p), and 7
1
(p, ) = B(p). Now,

1
(p, ) =(A(p)) = (p, (p, )),
and
1
(p, ) =A(p) =B(p). Therefore (p, ) =B
1
A =: (p). Continuity
is clear from the assumptions in Denition 6.1.
If both E and M are smooth manifolds and is a smooth projection, such that the
local trivializations can be chosen to be diffeomorphisms, then (E, M, ) is called
a smooth vector bundle. In this case the maps are smooth. The following result
allows us to construct smooth vector bundles and is important for the special vector
bundles used in this course.
Theorem 6.4.
21
Let M be a smooth manifold, and let {E
p
}
pM
be a family of k-
dimensional real vector spaces parametrized by p M. Dene E =
F
pM
E
p
, and
: E M as the mapping that maps E
p
to p M. Assume there exists
(i) an open covering {U

}
A
for M;
21
See Lee, Lemma 5.5.
43
(ii) for each A, a bijection

:
1
(U

) U

R
k
, such that 7

(p, ) is a vector space isomorphism between R

k
and E
p
;
(iii) for each , I, with U

6= , smooth mappings

: U

Gl(k, R) such that

(p, ) = (p,

(p)).
Then E has a unique differentiable (manifold) structure making (E, M, ) a smooth
vector bundle of rank k over M.
Proof: The proof of this theorem goes by verifying the hypotheses in Theorem
2.11. Let us propose charts for E. Let (V
p
,
p
), V
p
U

, be a smooth chart for

M containing p. As before dene
p
:
1
(V
p
)

V
p
R
k
. We can show, using
Theorem 2.11, that (
1
(V
p
),
p
), p M are smooth charts, which gives a unique
differentiable manifold structure. For details of the proof see Lee, Lemma 5.5.
J 6.5 Example. Consider the set E dened as
E =
_
(p
1
, p
2
,
1
,
2
) | p
1
= cos(), p
2
= sin(), cos(/2)
1
+sin(/2)
2
= 0
_
.
Clearly, E is a smooth vector bundle over S
1
of rank 1, embedded in R
4
. For
example if U = S
1
\{(1, 0)} ( (0, 2)), then (
1
(U)) = (p
1
, p
2
,
1
) is a local
trivialization. This bundle is called the M obius strip.
To show that E is a smooth vector bundle we need to verify the conditions in
Theorem 6.4. Consider a second chart U
0
= S
1
\{(1, 0)} ( (, )), and the
trivialization (
1
(U
0
)) = (p
1
, p
2
,
2
). For the we have that

1
(p
1
, p
2
,
1
) =

p
1
, p
2
,
1
,
cos(/2)
sin(/2)

,
and thus

1
(p
1
, p
2
,
1
) =

p
1
, p
2
,
cos(/2)
sin(/2)

,
which gives that (p) is represented by =
cos(/2)
sin(/2)
, for (0, ), which invert-
ible and smooth in p. I
Mappings between smooth vector bundles are called bundle maps, and are de-
ned as follows. Given vector bundles : E M and
0
: E
0
M
0
and smooth
mappings F : E E
0
and f : M M
0
, such that the following diagram commutes
E
F
E
0

_
0
M
f
M
0
44
and F|
E
p
: E
p
E
0
f (p)
is linear, then the pair (F, f ) is a called a smooth bundle
map.
A section, or cross section of a bundle E is a continuous mapping : M E,
such that = Id
M
. A section is smooth if is a smooth mapping. The space of
FIGURE 30. A cross section in a bundle E.
smooth section in E is denoted by E(M). The zero section is a mapping : ME
such that (p) = 0 E
p
for all p M.
J 6.7 Remark. By identifying M with the trivial bundle E
0
= M{0}, a section is
a bundle map from E
0
= M{0} to E. Indeed, dene
0
: E
0
E by
0
(p, 0) =
(p), and let f = Id
M
. Then
0
= = Id
M
. I
6.1. The tangent bundle and vector elds
The disjoint union of tangent spaces
TM :=
G
pM
T
p
M
is called the tangent bundle of M. We show now that TM is a fact a smooth vector
bundle over M.
Theorem 6.8. The tangent bundle TM is smooth vector bundle over M of rank m,
and as such TM is a smooth 2m-dimensional manifold.
Proof: Let (U, ) be a smooth chart for p M. Dene

X
i

x
i

=
_
p, (X
i
)
_
,
which clearly is a bijective map between
1
(U) and U R
m
. Moreover, X
p
=
(X
i
) 7
1
(p, X
p
) = X
i

x
i

p
is vector space isomorphism. Let (U

) be a cov-
ering of smooth charts of M. Let x =

(p), and x
0
=

(p). From our previous

considerations we easily see that

(p, X
p
) = (p,

(p)X
p
),
45
where

: U

Gl(m, R). It remains to show that

is smooth. We have

= (

Id)

(
1

Id). Using the change of coordinates

formula derived in Section 4 we have that

_
x, (X
i
)
_
=
_
x
0
1
(x), , x
0
m
(x),

x
0
j
x
i
X
j

_
,
which proves the smoothness of

. Theorem 6.4 can be applied now showing that

TM is asmooth vector bundle over M. From the charts (
1
(U), ) we conclude
that TM is a smooth 2m-dimensional manifold.
Denition 6.9. A smooth (tangent) vector eld is a smooth mapping
X : M TM,
with the property that X = id
M
. In other words X is a smooth (cross) section in
the vector bundle TM, see Figure 31. The space of smooth vector elds on M is
denoted by F(M).

FIGURE 31. A smooth vector eld X on a manifold M [right],

and as a curve, or section in the vector bundle TM [left].
For a chart (U, ) a vector eld X can be expressed as follows
X = X
i

x
i

p
,
where X
i
: U R. Smoothness of vector elds can be described in terms of the
component functions X
i
.
Lemma 6.11. A mapping X : M TM is a smooth vector eld at p U if and
only if the coordinate functions X
i
: U R are smooth.
Proof: In standard coordinates X is given by

X(x) = (x
1
, , x
m
,

X
1
(x), ,

X
m
(x)),
which immediately shows that smoothness of X is equivalent to the smoothness of
the coordinate functions X
i
.
46
In Section 4 we introduced the notion of push forward of a mapping f : N M.
Under submersions and immersion a vector eld X : N TN does not necessarily
push forward to a vector eld on M. If f is a diffeomorphism, then f

X =Y is a
vector eld on M. We remark that the denition of f

also allows use to restate

denitions about the rank of a map in terms of the differential, or pushforward f

at a point p N.
6.2. The cotangent bundle and differential 1-forms
The disjoint union
T

M :=
G
pM
T

p
M
is called cotangent bundle of M.
Theorem 6.12.
22
The cotangent bundle T

M is a smooth vector bundle over M of

rank m, and as such T

M is a smooth 2m-dimensional manifold.

Proof: The proof is more identical to the proof for TM, and is left to the reader
as an exercise.
The differential dh : M T

M is an example of a smooth function. The above

consideration give the coordinate wise expression for dh.
Denition 6.13. A smooth covector eld is a smooth mapping
: M T

M,
with the property that = id
M
. In order words is a smooth section in T

M.
The space of smooth covector elds on M is denoted by F

(M). Usually the space

F

(M) is denoted by
1
(M), and covector elds are referred to a (smooth) differ-
ential 1-form on M.
For a chart (U, ) a covector eld can be expressed as follows
=
i
dx
i

p
,
where
i
: U R. Smoothness of a covector elds can be described in terms of
the component functions
i
.
Lemma 6.14. A covector eld is smooth at p U if and only if the coordinate
functions
i
: U R are smooth.
Proof: See proof of Lemma 6.11.
We saw in the previous section that for arbitrary mappings f : N M, a vector
eld X F(N) does not necessarily push forward to a vector on M under f

. The
reason is that surjectivity and injectivity are both needed to guarantee this, which
22
See Lee, Proposition 6.5.
47
requires f to be a diffeomorphism. In the case of covector elds or differential 1-
forms the situation is completely opposite, because the pullback acts in the opposite
direction and is therefore onto the target space and uniquely dened. To be more
precise; given a 1-form
1
(M) we dene a 1-form f

1
(N) as follows
( f

)
p
= f

f (p)
.
Theorem 6.15.
23
The above dened pullback f

of under a smooth mapping

f : N M is a smooth covector eld, or differential 1-form on N.
Proof: Write
1
(M) is local coordinates; =
i
dy
i

q
. By Formula (5) we
then obtain
f

i
dy
i

f (p)
=
_

f
i
x
j
_
dx
j

p
=
_

f


f
i
x
j
_
x
dx
j

p
,
which proves that f

1
(N).
Given a mapping g : M R, the differential, or push forward g

= dg denes
an element in T

p
M. In local coordinates we have dg

p
=
g
x
i
dx
i

p
, and thus denes
a smooth covector eld on M; dg
1
(M). Given a smooth mapping f : N M,
it follows from (5) that
f

dg = f

g
x
i
dx
i

f (p)
=
g
x
i

f
i
x
j
dy
j

p
.
By applying the same to the 1-form d(g f ) we obtain the following identities
(6) f

dg = d(g f ), f

(g) = (g f ) f

.
Using the formulas in (6) we can also obtain (5) in a rather straightforward way.
Let g =
j
=h, e
j
i = y
j
, and = dg = dy
j
|
q
in local coordinates, then
f

_
(
j
)
_
=
_

j
f
_
f

dg =
_

j
f
_
d(g f ) =
_

f


f
j
x
i
_
x
dx
i

p
,
where the last step follows from (4).
Denition 6.16. A differential 1-form
1
(N) is called an exact 1-form if there
exists a smooth function g : N R, such that = dg.
23
Lee, Proposition 6.13.
48
The notation for tangent vectors was motivated by the fact that functions on a
manifold can be differentiated in tangent directions. The notation for the cotangent
vectors was partly motivated as the reciprocal of the partial derivative. The in-
troduction of line integral will give an even better motivation for the notation for
cotangent vectors. Let N = R, and a 1-form on N given in local coordinates by

t
= h(t)dt, which can be identied with a function h. The notation makes sense
because can be integrated over any interval [a, b] R:
Z
[a,b]
:=
Z
b
a
h(t)dt.
Let M =R, and consider a mapping f : M =RN =R, which satises f
0
(t) >0.
Then t = f (s) is an appropriate change of variables. Let [c, d] = f ([a, b]), then
Z
[c,d]
f

=
Z
d
c
h( f (s)) f
0
(s)ds =
Z
b
a
h(t)dt =
Z
[a,b]
,
which is the change of variables formula for integrals. We can use this now to
dene the line integral over a curve on a manifold N.
Denition 6.17. Let : [a, b] R N, and let be a 1-form on N and

the
pullback of , which is a 1-form on R. Denote the image of in N also by , then
Z

:=
Z
[a,b]

=
Z
[a,b]

i
((t))
0
i
(t)dt,
24
the expression in local coordinates.
The latter can be seen by combining some of the notion introduced above:

d
dt
= (

)
t
=
(t)

d
dt
=

(t)
0
(t).
Therefore,

= (

)
t
dt =
(t)

0
(t)dt =
i
((t))
0
i
(t)dt, and
Z

=
Z
[a,b]

=
Z
[a,b]

(t)

0
(t)dt =
Z
[a,b]

i
((t))
0
i
(t)dt.
If
0
is nowhere zero then the map : [a, b] N is either an immersion or em-
bedding. For example in the embedded case this gives an embedded submanifold
N with boundary = {(a), (b)}. Let = dg be an exact 1-form, then
Z

dg = g

= g((b)) g((a)).
Indeed,
Z

dg =
Z
[a,b]

dg =
Z
[a,b]
d(g) =
Z
b
a
(g)
0
(t)dt = g((b)) g((a)).
24
The expressions
i
(t) R are the components of
0
(t) T
(t)
M.
49
This identity is called the Fundamental Theorem for Line Integrals and is a spe-
cial case of the Stokes Theorem (see Section 16).
50
III. Tensors and differential
forms
7. Tensors and tensor products
In the previous chapter we encountered linear functions on vector spaces, linear
functions on tangent spaces to be precise. In this chapter we extend to notion of
linear functions on vector spaces to multilinear functions.
Denition 7.1. Let V
1
, ,V
r
, and W be real vector spaces. A mapping T : V
1

V
r
W is called a multilinear mapping if
T(v
1
, , v
i
+v
0
i
, , v
r
) =T(v
1
, , v
i
, v
r
) +T(v
1
, , v
0
i
, v
r
), i,
and for all , R i.e. f is linear in each variable v
i
separately.
Now consider the special case that W =R, then T becomes a multilinear func-
tion, or form, and a generalization of linear functions. If in addition V
1
= =
V
r
=V, then
T : V V R,
is a multilinear function on V, and is called a covariant r-tensor on V. The number
of copies r is called the rank of T. The space of covariant r-tensors on V is denoted
by T
r
(V), which clearly is a real vector space using the multilinearity property in
Denition 7.1. In particular we have that T
0
(V)

= R, T
1
(V) =V

, and T
2
(V) is
the space of bilinear forms on V. If we consider the case V
1
= =V
r
=V

, then
T : V

R,
is a multilinear function on V

, and is called a contravariant r-tensor on V. The

space of contravariant r-tensors on V is denoted by T
r
(V). Here we have that
T
0
(V)

=R, and T
1
(V) = (V

)

=V.
J 7.2 Example. The cross product on R
3
is an example of a multilinear (bilinear)
function mapping not to R to R
3
. Let x, y R
3
, then
T(x, y) = x y R
3
,
which clearly is a bilinear function on R
3
. I
51
Since multilinear functions on V can be multiplied, i.e. given vector spaces V,W
and tensors T T
r
(V), and S T
s
(W), the multilinear function
R(v
1
, v
r
, w
1
, , w
s
) = T(v
1
, v
r
)S(w
1
, , w
s
)
is well dened and is a multilnear function on V
r
W
s
. This brings us to the
following denition. Let T T
r
(V), and S T
s
(W), then
T S : V
r
W
s
R,
is given by
T S(v
1
, , v
r
, w
1
, w
s
) = T(v
1
, , v
r
)S(w
1
, w
s
).
This product is called the tensor product. By taking V =W, T S is a covariant
(r +s)-tensor on V, which is a element of the space T
r+s
(V) and : T
r
(V)
T
s
(V) T
r+s
(V).
Lemma 7.3. Let T T
r
(V), S, S
0
T
s
(V), and R T
t
(V), then
(i) (T S) R = T (SR) (associative),
(ii) T (S+S
0
) = T S+T S
0
(distributive),
(iii) T S 6= ST (non-commutative).
The tensor product is also dened for contravariant tensors and mixed tensors.
As a special case of the latter we also have the product between covariant and
contravariant tensors.
J7.4 Example. The last property can easily be seen by the following example. Let
V =R
2
, and T, S T
1
(R
2
), given by T(v) = v
1
+v
2
, and S(w) = w
1
w
2
, then
T S(1, 1, 1, 0) = 2 6= 0 = ST(1, 1, 1, 0),
which shows that is not commutative in general. I
The following theorem shows that the tensor product can be used to build the
tensor space T
r
(V) from elementary building blocks.
Theorem 7.5.
25
Let {v
1
, , v
n
} be a basis for V, and let {
1
, ,
n
} be the dual
basis for V

. Then the set

B=
_

i
1

i
r
: 1 i
1
, , i
r
n
_
,
is a basis for the n
r
-dimensional vector space T
r
(V).
25
See Lee, Prop. 11.2.
52
Proof: Compute
T
i
1
i
r

i
1

i
r
(v
j
1
, , v
j
r
) = T
i
1
i
r

i
1
(v
j
1
)
i
r
(v
j
r
)
= T
i
1
i
r

j
1
j
1

j
r
j
r
= T
j
1
j
r
= T(v
j
1
, , v
j
r
),
which shows by using the multilinearity of tensors that T can be expanded in the
basis B as follows;
T = T
i
1
i
r

i
1

i
r
,
where T
i
1
i
r
=T(v
i
1
, , v
i
r
), the components of the tensor T. Linear independence
follows form the same calculation.
J 7.6 Example. Consider the the 2-tensors T(x, y) = x
1
y
1
+x
2
y
2
, T
0
(x, y) = x
1
y
2
+
x
2
y
2
and T
00
=x
1
y
1
+x
2
y
2
+x
1
y
2
on R
2
. With respect to the standard bases
1
(x) =
x
1
,
2
(x) = x
1
, and

1
(x, y) = x
1
y
1
,
1

2
(x, y) = x
1
y
2
,

2
(x, y) = x
2
y
1
, and
2

2
(x, y) = x
2
y
2
.
Using this the components of T are given by T
11
=1, T
12
=0, T
21
=0, and T
22
=1.
Also notice that T
0
= S S
0
, where S(x) = x
1
+x
2
, and S
0
(y) = y
2
. Observe that
not every tensor T T
2
(R
2
) is of the form T = S S
0
. For example T
00
6= S S
0
,
for any S, S
0
T
1
(R
2
). I
In Lee, Ch. 11, the notion of tensor product between arbitrary vector spaces is
explained. Here we will discuss a simplied version of the abstract theory. Let
V and W be two (nite dimensional) real vector spaces, with bases {v
1
, v
n
}
and {w
1
, w
m
} respectively, and for their dual spaces V

and W

we have the
dual bases {
1
, ,
n
} and {
1
, ,
m
} respectively. If we use the identication
{V

}

=V, and {W

}

=W we can dene V W as follows:
Denition 7.7. The tensor product of V and W is the real vector space of (nite)
linear combinations
V W :=
_

i j
v
i
w
j
:
i j
R
_
=
_
_
v
i
w
j
_
i, j
_
,
where v
i
w
j
(v

, w

) :=v

(v
i
)w

(w
j
), using the identication v
i
(v

) :=v

(v
i
), and
w
j
(w

) := w

(w
j
), with (v

, w

) V

.
53
To get a feeling of what the tensor product of two vector spaces represents con-
sider the tensor product of the dual spaces V

and W

. We obtain the real vector

space of (nite) linear combinations
V

:=
_

i j

j
:
i j
R
_
=
_
_

j
_
i, j
_
,
where
i

j
(v, w) =
i
(v)
j
(w) for any (v, w) V W. One can show that V

is isomorphic to space of bilinear maps fromV W to R. In particular elements

v

all lie in V

, but not all elements in V

are of this form. The

isomorphism is easily seen as follows. Let v =
i
v
i
, and w =
j
w
j
, then for a given
bilinear form b it holds that b(v, w) =
i

j
b(v
i
, w
j
). By denition of dual basis
we have that
i

j
=
i
(v)
j
(w) =
i

j
(v, w), which shows the isomorphism by
setting
i j
= b(v
i
, w
j
).
In the case V

W the tensors represent linear maps from V to W. Indeed,

from the previous we know that elements in V

W represent bilinear maps from

V W

to R. For an element b V

W this means that b(v, ) : W

R, and thus
b(v, ) (W

)

=W.
J 7.8 Example. Consider vectors a V and b

W, then a

(b

can be iden-
tied with a matrix, i.e a

(b

(v, ) = a

(v)(b

()

= a

(v)b. For example let

a

(v) = a
1
v
1
+a
2
v
2
+a
3
v
3
, and
Av =a

(v)b =
_
a
1
b
1
v
1
+a
2
b
1
v
2
+a
3
b
1
v
3
a
1
b
2
v
1
+a
2
b
2
v
2
+a
3
b
2
v
3
_
=
_
a
1
b
1
a
2
b
1
a
3
b
1
a
1
b
2
a
2
b
2
a
3
b
2
_
_
_
_
v
1
v
2
v
3
_
_
_.
Symbolically we can write
A = a

b =
_
a
1
b
1
a
2
b
1
a
3
b
1
a
1
b
2
a
2
b
2
a
3
b
2
_
=

a
1
a
2
a
3

_
b
1
b
2
_
,
which shows how a vector and covector can be tensored to become a matrix. Note
that it also holds that A = (a b

= b a

. I
Lemma 7.9. We have that
(i) V W and W V are isomorphic;
(ii) (U V) W and U (V W) are isomorphic.
With the notion of tensor product of vector spaces at hand we now conclude that
the above describe tensor spaces T
r
(V) and V
r
(V) are given as follows;
T
r
(V) =V

. .
r times
, T
r
(V) =V V
. .
r times
.
54
By considering tensor products of Vs and V

s we obtain the tensor space of

mixed tensors;
T
r
s
(V) :=V

. .
r times
V V
. .
s times
.
Elements in this space are called (r, s)-mixed tensors on V r copies of V

, and s
copies of V. Of course the tensor product described above is dened in general for
tensors T T
r
s
(V), and S T
r
0
s
0
(V):
: T
r
s
(V) T
r
0
s
0 (V) T
r+r
0
s+s
0
(V).
The analogue of Theorem 7.5 can also be established for mixed tensors. In the next
sections we will see various special classes of covariant, contravariant and mixed
tensors.
J7.10 Example. The inner product on a vector space V is an example of a covariant
2-tensor. This is also an example of a symmetric tensor. I
J 7.11 Example. The determinant of n vectors in R
n
is an example of covariant n-
tensor on R
n
. The determinant is skew-symmetric, and an example of an alternating
tensor. I
If f : V W is a linear mapping between vector spaces and T is an covariant
tensor on W we can dene concept of pullback of T. Let T T
r
(W), then f

T
T
r
(V) is dened as follows:
f

T(v
1
, v
r
) = T( f (v
1
), , f (v
r
)),
and f

: T
r
(W) T
r
(V) is a linear mapping. Indeed, f

(T +S) = T f +S f =
f

T + f

S, and f

T = T f = f

T. If we represent f by a matrix A with

respect to bases {v
i
} and {w
j
} for V and W respectively, then the matrix for the
linear f

is given by
A

. .
r times
,
with respect to the bases {
i
1

i
r
} and {
j
1

j
r
} for T
r
(W) and T
r
(V)
respectively.
J 7.12 Remark. The direct sum
T

(V) =

M
r=0
T
r
(V),
consisting of nite sums of covariant tensors is called the covariant tensor algebra
of V with multiplication given by the tensor product . Similarly, one denes the
55
contravariant tensor algebra
T

(V) =

M
r=0
T
r
(V).
For mixed tensors we have
T(V) =

M
r,s=0
T
r
s
(V),
which is called the tensor algebra of mixed tensor of V. Clearly, T

(V) and T

(V)
subalgebras of T(V). I
8. Symmetric and alternating tensors
There are two special classes of tensors which play an important role in the
analysis of differentiable manifolds. The rst class we describe are symmetric
tensors. We restrict here to covariant tensors.
Denition 8.1. A covariant r-tensor T on a vector space V is called symmetric if
T(v
1
, , v
i
, , v
j
, , v
r
) = T(v
1
, , v
j
, , v
i
, , v
r
),
for any pair of indices i j. The set of symmetric covariant r-tensors on V is
denoted by
r
(V) T
r
(V), which is a (vector) subspace of T
r
(V).
If a S
r
is a permutation, then dene
a
T(v
1
, , v
r
) = T(v
a(1)
, , v
a(r)
),
where a({1, , r}) = {a(1), , a(r)}. From this notation we have that for two
permutations a, b S
r
,
b
(
a
T) =
ba
T. Dene
Sym T =
1
r!

aS
r
a
T.
It is straightforward to see that for any tensor T T
r
(V), Sym T is a symmetric.
Moreover, a tensor T is symmetric if and only if Sym T = T. For that reason
Sym T is called the (tensor) symmetrization.
J 8.2 Example. Let T, T
0
T
2
(R
2
) be dened as follows: T(x, y) = x
1
y
2
, and
T
0
(x, y) = x
1
y
1
. Clearly, T is not symmetric and T
0
is. We have that
Sym T(x, y) =
1
2
T(x, y) +
1
2
T(y, x)
=
1
2
x
1
y
2
+
1
2
y
1
x
2
,
56
which clearly is symmetric. If we do the same thing for T
0
we obtain:
Sym T
0
(x, y) =
1
2
T
0
(x, y) +
1
2
T
0
(y, x)
=
1
2
x
1
y
1
+
1
2
y
1
x
1
= T
0
(x, y),
showing that operation Sym applied to symmetric tensors produces the same ten-
sor again. I
Using symmetrization we can dene the symmetric product. Let S
r
(V) and
T
s
(V) be symmetric tensors, then
S T = Sym (ST).
The symmetric product of symmetric tensors is commutative which follows di-
rectly from the denition:
S T(v
1
, , v
r+s
) =
1
(r +s)!

aS
r+s
S(v
a(1)
, , v
a(r)
)T(v
a(r+1)
, , v
a(r+s)
).
J 8.3 Example. Consider the 2-tensors S(x) = x
1
+x
2
, and T(y) = y
2
. Now S
T(x, y) = x
1
y
2
+x
2
y
2
, and T S(x, y) = x
2
y
1
+x
2
y
2
, which clearly gives that S
T 6= T S. Now compute
Sym (ST)(x, y) =
1
2
x
1
y
2
+
1
2
x
2
y
2
+
1
2
y
1
x
2
+
1
2
x
2
y
2
=
1
2
x
1
y
2
+
1
2
x
2
y
1
+x
2
y
2
= S T(x, y).
Similarly,
Sym (T S)(x, y) =
1
2
x
2
y
1
+
1
2
x
2
y
2
+
1
2
y
2
x
1
+
1
2
x
2
y
2
=
1
2
x
1
y
2
+
1
2
x
2
y
1
+x
2
y
2
= T S(x, y),
which gives that S T = T S. I
Lemma 8.4. Let {v
1
, , v
n
} be a basis for V, and let {
1
, ,
n
} be the dual
basis for V

. Then the set

B

=
_

i
1

i
r
: 1 i
1
i
r
n
_
,
is a basis for the (sub)space
r
(V) of symmetric r-tensors. Moreover, dim
r
(V) =
_
n+r 1
r
_
=
(n+r1)!
r!(n1)!
.
57
Proof: Proving that B

is a basis follows from Theorem 7.5, see also Lemma

8.10. It remains to establish the dimension of
r
(V). Note that the elements in the
basis are given by multi-indices (i
1
, , i
r
), satisfying the property that 1 i
1

i
r
n. This means choosing r integers satisfying this restriction. To do this
redene j
k
:=i
k
+k1. The integers j range from 1 through n+r 1. Now choose
r integers out of n+r 1, i.e.
_
n+r 1
r
_
combinations ( j
1
, , j
r
), which are
in one-to-one correspondence with (i
1
, , i
r
).
Another important class of tensors are alternating tensors and are dened as
follows.
Denition 8.5. A covariant r-tensor T on a vector space V is called alternating if
T(v
1
, , v
i
, , v
j
, , v
r
) =T(v
1
, , v
j
, , v
i
, , v
r
),
for any pair of indices i j. The set of alternating covariant r-tensors on V is
denoted by
r
(V) T
r
(V), which is a (vector) subspace of T
r
(V).
As before we dene
Alt T =
1
r!

aS
r
(1)
a a
T,
where (1)
a
is +1 for even permutations, and 1 for odd permutations. We say
that Alt T is the alternating projection of a tensor T, and Alt T is of course a
alternating tensor.
J 8.6 Example. Let T, T
0
T
2
(R
2
) be dened as follows: T(x, y) = x
1
y
2
, and
T
0
(x, y) = x
1
y
2
x
2
y
1
. Clearly, T is not alternating and T
0
(x, y) = T
0
(y, x) is
alternating. We have that
Alt T(x, y) =
1
2
T(x, y)
1
2
T(y, x)
=
1
2
x
1
y
2

1
2
y
1
x
2
=
1
2
T
0
(x, y),
which clearly is alternating. If we do the same thing for T
0
we obtain:
Alt T
0
(x, y) =
1
2
T
0
(x, y)
1
2
T
0
(y, x)
=
1
2
x
1
y
2

1
2
x
2
y
1

1
2
y
1
x
2
+
1
2
y
2
x
1
= T
0
(x, y),
showing that operation Alt applied to alternating tensors produces the same tensor
again. Notice that T
0
(x, y) = det(x, y). I
58
This brings us to the fundamental product of alternating tensors called the wedge
product. Let S
r
(V) and T
s
(V) be symmetric tensors, then
ST =
(r +s)!
r!s!
Alt (ST).
The wedge product of alternating tensors is anti-commutative which follows di-
rectly from the denition:
ST(v
1
, , v
r+s
) =
1
r!s!

aS
r+s
(1)
a
S(v
a(1)
, , v
a(r)
)T(v
a(r+1)
, , v
a(r+s)
).
In the special case of the wedge of two covectors , V

gives
=.
In particular we have that
(i) (T S) R = T (SR);
(ii) (T +T
0
) S = T S+T
0
S;
(iii) T S = (1)
rs
ST, for T
r
(V) and S
s
(V);
(iv) T T = 0.
The latter is a direct consequence of the denition of Alt . In order to prove these
properties we have the following lemma.
Lemma 8.7. Let T T
r
(V) and S T
s
(V), then
Alt (T S) = Alt ((Alt T) S) = Alt (T Alt S).
Proof: Let G

= S
r
be the subgroup of S
r+s
consisting of permutations that only
permute the element {1, , r}. For a G, we have a
0
S
r
. Now
a
(T S) =
a0
T S, and thus
1
r!

aG
(1)
a a
(T S) = (Alt T) S.
For the right cosets {ba : a G} we have

aG
(1)
ba ba
(T S) = (1)
b b

aG
(1)
a a
(T S)

= r!(1)
b b

(Alt T) S

.
Taking the sum over all right cosets with the factor
1
(r+s)!
gives
Alt (T S) =
1
(r +s)!

aG
(1)
ba ba
(T S)
=
r!
(r +s)!

b
(1)
b b

(Alt T) S

= Alt ((Alt T) S),

59
where the latter equality is due to the fact that r! terms are identical under the
denition of Alt ((Alt T) S).
Property (i) can nowbe proved as follows. Clearly Alt (Alt (T S)T S) =0,
and thus from Lemma 8.7 we have that
0 = Alt ((Alt (T S) T S) R) = Alt (Alt (T S) R) Alt (T SR).
By denition
(T S) R =
(r +s +t)!
(r +s)!t!
Alt ((T S) R)
=
(r +s +t)!
(r +s)!t!
Alt

(r +s)!
r!s!
Alt (T S)

=
(r +s +t)!
r!s!t!
Alt (T SR).
The same formula holds for T (SR), which prove associativity. More generally
it holds that for T
i

r
i
(V)
T
1
T
k
=
(r
1
+ +r
k
)!
r
1
! r
k
!
Alt (T
1
T
k
).
Property (iii) can be seen as follows. Each term in T S can be found in S T.
This can be done by linking the permutations a and a
0
. To be more precise, how
many permutation of two elements are needed to change
a (i
1
, , i
r
, j
r+1
, , j
r+s
) into a
0
( j
r+1
, , j
r+s
, i
1
, , i
r
).
This clearly requires rs permutations of two elements, which shows Property (iii).
J 8.8 Example. Consider the 2-tensors S(x) = x
1
+x
2
, and T(y) = y
2
. As before
ST(x, y) = x
1
y
2
+x
2
y
2
, and T S(x, y) = x
2
y
1
+x
2
y
2
. Now compute
Alt (ST)(x, y) =
1
2
x
1
y
2
+
1
2
x
2
y
2

1
2
y
1
x
2

1
2
x
2
y
2
=
1
2
x
1
y
2

1
2
x
2
y
1
= 2

ST(x, y)

.
Similarly,
Alt (T S)(x, y) =
1
2
x
2
y
1
+
1
2
x
2
y
2

1
2
y
2
x
1

1
2
x
2
y
2
=
1
2
x
1
y
2
+
1
2
x
2
y
1
=2

T S(x, y)

,
which gives that S T = T S. Note that if T = e

1
, i.e. T(x) = x
1
, and S = e

2
,
i.e. S(x) = x
2
, then
T S(x, y) = x
1
y
2
x
2
y
1
= det(x, y).
I
60
J 8.9 Remark. Some authors use the more logical denition
S

T = Alt (ST),
which is in accordance with the denition of the symmetric product. This denition
is usually called the alt convention for the wedge product, and our denition is
usually referred to as the determinant convention. For computational purposes the
determinant convention is more appropriate. I
If {e

1
, , e

n
} is the standard dual basis for (R
n
)

, then for vectors a

1
, , a
n

R
n
,
det(a
1
, , a
n
) = e

1
e

n
(a
1
, , a
n
).
Using the multilinearity the more general statement reads
(7)
1

n
(a
1
, , a
n
) = det
_

i
(a
j
)
_
,
where
i
are co-vectors.
The alternating tensor det =e

1
e

n
is called the determinant function on R
n
.
If f : V W is a linear map between vector spaces then the pullback f

T
r
(V)
of any alternating tensor T
r
(W) is given via the relation:
f

T(v
1
, , v
r
) = T
_
f (v
1
), , f (v
r
)
_
, f

:
r
(W)
r
(V).
In particular, f

(T S) = ( f

T) f

(S). As a special case we have that if f : V

V, linear, and dimV = n, then
(8) f

T = det( f )T,
for any alternating tensor T
n
(V). This can be seen as follows. By multilinearity
we verify the above relation for the vectors {e
i
}. We have that
f

T(e
1
, , e
n
) = T( f (e
1
), , f (e
n
))
= T( f
1
, , f
n
) = cdet( f
1
, , f
n
) = cdet( f ),
where we use the fact that
n
(V)

=R (see below). On the other hand

det( f )T(e
1
, e
n
) = det( f )c det(e
1
, , e
n
)
= cdet( f ),
which proves (8).
Lemma 8.10. Let {
1
, ,
n
} be a basis for V

, then the set

B

=
_

i
1

i
r
: 1 i
1
< < i
r
n
_
,
is a basis for
r
(V), and dim
r
(V) =
n!
(nr)!r!
. In particular, dim
r
(V) = 0 for
r > n.
61
Proof: From Theorem 7.5 we know that any alternating tensor T
r
(V) can
be written as
T = T
j
1
j
r

j
1

j
r
.
We have that Alt T = T, and so
T = T
j
1
j
r
Alt (
j
1

j
r
) =
1
r!
T
j
1
j
r

j
1

j
r
In the expansion the terms with j
k
= j
`
are zero since
j
k

j
`
= 0. If we order the
indices in increasing order we obtain
T =
1
r!
T
i
1
i
r

i
1

i
r
,
which show that B

spans
r
(V).
Linear independence can be proved as follows. Let 0 =
i
1
i
r

i
1

i
r
, and
thus
i
1
i
r
=
i
1

i
r
(v
i
1
, , v
i
r
) = 0, which proves linear independence.
It is immediately clear that B

consists of
_
n
r
_
elements.
As we mentioned before the operation Alt is called the alternating projection.
As a matter of fact Sym is also a projection.
Lemma 8.11. Some of the basic properties can be listed as follows;
(i) Sym and Alt are projections on T
r
(V), i.e. Sym
2
= Sym, and Alt
2
= Alt;
(ii) T is symmetric if and only if Sym T = T, and T is alternating if and only
if Alt T = T;
(iii) Sym(T
r
(V)) =
r
(V), and Alt(T
r
(V)) =
r
(V);
(iv) Sym Alt = Alt Sym = 0, i.e. if T
r
(V), then Sym T = 0, and if
T
r
(V), then Alt T = 0;
(v) let f : V W, then Sym and Alt commute with f

: T
r
(W) T
r
(V), i.e.
Sym f

= f

Sym, and Alt f

= f

Alt.
9. Tensor bundles and tensor elds
Generalizations of tangent spaces and cotangent spaces are given by the tensor
spaces
T
r
(T
p
M), T
s
(T
p
M), and T
r
s
(T
p
M),
62
where T
r
(T
p
M) = T
r
(T

p
M). As before we can introduce the tensor bundles:
T
r
M =
G
pM
T
r
(T
p
M),
T
s
M =
G
pM
T
s
(T
p
M),
T
r
s
M =
G
pM
T
r
s
(T
p
M),
which are called the covariant r-tensor bundle, contravariant s-tensor bundle,
and the mixed (r, s)-tensor bundle on M. As for the tangent and cotangent bundle
the tensor bundles are also smooth manifolds. In particular, T
1
M = T

M, and
T
1
M = TM. Recalling the symmetric and alternating tensors as introduced in the
previous section we also dene the tensor bundles
r
M and
r
M.
Theorem 9.1. The tensor bundles T
r
M, T
r
M and T
r
s
M are smooth vector bundles.
Proof: Using Section 6 the theorem is proved by choosing appropriate local
trivializations . For coordinates x =

(p) and x
0
=

(p) we recall that

x
i

p
=
x
0
j
x
i

x
0
j

p
, dx
0 j
|
p
=
x
0
j
x
i
dx
i
|
p
.
For a covariant tensor T T
r
M this implies the following. The components are
dened in local coordinates by
T = T
i
1
i
r
dx
i
1
dx
i
r
, T
i
1
i
r
= T

x
i
1
, ,

x
i
r

.
The change of coordinates x x
0
then gives
T
i
1
i
r
= T

x
0
j
1
x
i
1

x
0
j
1
, ,
x
0
j
r
x
i
r

x
0
j
r

= T
0
j
1
j
r
x
0
j
1
x
i
1

x
0
j
r
x
i
r
.
Dene
(T
i
1
i
r
dx
i
1
dx
i
r
) =
_
p, (T
i
1
i
r
)
_
,
then

01
_
p, (T
0
j
1
j
r
)
_
=
_
p,

T
0
j
1
j
r
x
0
j
1
x
i
1

x
0
j
r
x
i
r

_
.
The products of the partial derivatives are smooth functions. We can now apply
Theorem 6.4 as we did in Theorems 6.8 and 6.12.
On tensor bundles we also have the natural projection
: T
r
s
M M,
dened by (p, T) = p. A smooth section in T
r
s
M is a smooth mapping
: M T
r
s
M,
63
such that = id
M
. The space of smooth sections in T
r
s
M is denoted by F
r
s
(M).
For the co- and contravariant tensors these spaces are denoted by F
r
(M) and F
s
(M)
respectively. Smooth sections in these tensor bundles are also called smooth tensor
elds. Clearly, vector elds and 1-forms are examples of tensor elds. Sections in
tensor bundles can be expressed in coordinates as follows:
=
_

i
1
i
r
dx
i
1
dx
i
r
, F
r
(M),

j
1
j
s

x
j
1


x
js
, F
s
(M),

j
1
j
s
i
1
i
r
dx
i
1
dx
i
r

x
j
1


x
js
, F
r
s
(M).
Tensor elds are often denoted by the component functions. The tensor and tensor
elds in this course are, except for vector elds, all covariant tensors and covariant
tensor elds. Smoothness of covariant tensor elds can be described in terms of
the component functions
i
1
i
r
.
Lemma 9.2.
26
A covariant tensor eld is smooth at p U if and only if
(i) the coordinate functions
i
1
i
r
: U R are smooth, or equivalently if and
only if
(ii) for smooth vector elds X
1
, , X
r
dened on any open set U M, then the
function (X
1
, , X
r
) : U R, given by
(X
1
, , X
r
)(p) =
p
(X
1
(p), , X
r
(p)),
is smooth.
The same equivalences hold for contravariant and mixed tensor elds.
Proof: Use the identities in the proof of Theorem 9.1 and then the proof goes as
Lemma 6.11.
J9.3 Example. Let M =R
2
and let =dx
1
dx
1
+x
2
1
dx
2
dx
2
. If X =
1
(x)

x
1
+

2
(x)

x
2
and Y =
1
(x)

x
1
+
2
(x)

x
2
are arbitrary smooth vector elds on T
x
R
2
=
R
2
, then
(X,Y) =
1
(x)
1
(x) +x
2
1

2
(x)
2
(x),
which clearly is a smooth function in x R
2
. I
For covariant tensors we can also dene the notion of pullback of a mapping
f between smooth manifolds. Let f : N M be a smooth mappings, then the
pullback f

: T
r
(T
f (p)
M) T
r
(T
p
N) is dened as
( f

T)(X
1
, X
r
) := T( f

X
1
, , f

X
r
),
where T T
r
(T
f (p)
M), and X
1
, , X
r
T
p
N. We have the following properties.
26
See Lee, Lemma 11.6.
64
Lemma 9.4.
27
Let f : N M, g : M P be smooth mappings, and let p N,
S T
r
(T
f (p)
M), and T T
r
0
(T
f (p)
M), then:
(i) f

: T
r
(T
f (p)
M) T
r
(T
p
N) is linear;
(ii) f

(ST) = f

S f

T;
(iii) (g f )

= f

: T
r
(T
(g f )(p)
P) T
r
(T
p
M);
(iv) id

M
S = S;
(v) f

: T
r
M T
r
N is a smooth bundle map.
Proof: Combine the proof of Lemma 5.3 with the identities in the proof of
Theorem 9.1.
J 9.5 Example. Let us continue with the previous example and let N = M = R
2
.
Consider the mapping f : N M dened by
f (x) = (2x
1
, x
3
2
x
1
).
For given tangent vectors X,Y T
x
N, given by X =
1

x
1
+
2

x
2
and Y =
1

x
1
+

2

x
2
, we can compute the pushforward
f

=
_
2 0
1 3x
2
2
_
, and
f

X = 2
1

y
1
+(
1
+3x
2
2

2
)

y
2
,
f

Y = 2
1

y
1
+(
1
+3x
2
2

2
)

y
2
.
Let be given by = dy
1
dy
1
+y
2
1
dy
2
dy
2
. This then yields
( f

X, f

Y) = 4
1

1
+4x
2
1
(
1
3x
2
1

2
)(
1
3x
2
1

2
)
= 4(1+x
2
1
)
1

1
12x
2
1
x
2
2

2
12x
2
1
x
2
2

1
+36x
2
1
x
4
2
x
2
1

2
.
We have to point out here that f

X and f

Y are tangent vectors in T

x
N and not
necessarily vector elds, although we can use this calculation to compute f

,
which clearly is a smooth 2-tensor eld on T
2
N. I
27
See Lee, Proposition 11.8.
65
J 9.6 Example. A different way of computing f

is via a local representation of

directly. We have = dy
1
dy
1
+y
2
1
dy
2
dy
2
, and
f

= d(2x
1
) d(2x
1
) +4x
2
1
d(x
3
2
x
1
) d(x
3
2
x
1
)
= 4dx
1
dx
1
+4x
2
1
_
3x
2
2
dx
2
dx
1
_

_
3x
2
2
dx
2
dx
1
_
= 4(1+x
2
1
)dx
1
dx
1
12x
2
1
x
2
2
dx
1
dx
2
12x
2
1
x
2
2
dx
2
dx
1
+36x
2
1
x
4
2
x
2
1
dx
2
dx
2
.
Here we used the fact that computing the differential of a mapping to R produces
the pushforward to a 1-form on N. I
J 9.7 Example. If we perform the previous calculation for an arbitrary 2-tensor
= a
11
dy
1
dy
1
+a
12
dy
1
dy
2
+a
21
dy
2
dy
1
+a
22
dy
2
dy
2
.
Then,
f

= (4a
11
2a
12
2a
21
+a
22
)dx
1
dx
1
+(6a
12
3a
22
)x
2
2
dx
1
dx
2
+(6a
21
3a
22
)x
2
2
dx
2
dx
1
+9x
4
2
a
22
dx
2
dx
2
,
which produces the following matrix if we identify T
2
(T
x
N) and T
2
(T
y
M) with R
4
:
f

=
_
_
_
_
_
4 2 2 1
0 6x
2
2
0 3x
2
2
0 0 6x
2
2
3x
2
2
0 0 0 9x
4
2
_
_
_
_
_
,
which clearly equal to the tensor product of the matrices (J f )

, i.e.
f

= (J f )

(J f )

=
_
2 1
0 3x
2
2
_

_
2 1
0 3x
2
2
_
.
This example show how to interpret f

: T
2
(T
y
M) T
2
(T
x
N) as a linear mapping.
I
As for smooth 1-forms this operation extends to smooth covariant tensor elds:
( f

)
p
= f

(
f (p)
), F
r
(N), which in coordinates reads
( f

)
p
(X
1
, X
r
) :=
f (p)
( f

X
1
, , f

X
r
),
for tangent vectors X
1
, , X
r
T
p
N.
Lemma 9.8.
28
Let f : N M, g : MP be smooth mappings, and let h C

(M),
F
r
(M), and F
r
(N), then:
(i) f

: F
r
(M) F
r
(N) is linear;
28
See Lee, Proposition 11.9.
66
(ii) f

(h) = (h f ) f

;
(iii) f

() = f

;
(iv) f

is a smooth covariant tensor eld;

(v) (g f )

= f

;
(vi) id

M
=;
Proof: Combine the Lemmas 9.4 and 9.2.
10. Differential forms
A special class of covariant tensor bundles and associated bundle sections are
the so-called alternating tensor bundles. Let
r
(T
p
M) T
r
(T
p
M) be the space
alternating tensors on T
p
M. We know from Section 8 that a basis for
r
(T
p
M) is
given by
_
dx
i
1
dx
i
r
: 1 i
1
, , i
r
m
_
,
and dim
r
(T
p
M) =
m!
r!(mr)!
. The associated tensor bundles of atlernating covariant
tensors is denoted by
r
M. Smooth sections in
r
M are called differential r-forms,
and the space of smooth sections is denoted by
r
(M) F
r
(M). In particular

0
(M) = C

(M), and
1
(M) = F

(M). In terms of components a differential r-

form, or r-form for short, is given by

i
1
i
r
dx
i
1
dx
i
r
,
and the components
i
1
i
r
are smooth functions. An r-form acts on vector elds
X
1
, , X
r
as follows:
(X
1
, , X
r
) =

aS
r
(1)
a

i
1
i
r
dx
i
1
(X
a(1)
) dx
i
r
(X
a(r)
)
=

aS
r
(1)
a

i
1
i
r
X
i
1
a(1)
X
i
r
a(r)
.
J 10.1 Example. Let M =R
3
, and = dx dz. Then for vector elds
X
1
= X
1
1

x
+X
2
1

y
+X
3
1

z
,
and
X
2
= X
1
2

x
+X
2
2

y
+X
3
2

z
,
we have that
(X
1
, X
2
) = X
1
1
X
3
2
X
3
1
X
1
2
.
I
67
An important notion that comes up in studying differential forms is the notion
of contracting an r-form. Given an r-form
r
(M) and a vector eld X F(M),
then
i
X
:=(X, , , ),
is called the contraction with X, and is a differential (r 1)-form on M. Another
notation for this is i
X
= Xy. Contraction is a linear mapping
i
X
:
r
(M)
r1
(M).
Contraction is also linear in X, i.e. for vector elds X,Y it holds that
i
X+Y
= i
X
+i
Y
, i
X
= i
X
.
Lemma 10.2.
29
Let
r
(M) and X F(M) a smooth vector eld, then
(i) i
X

r1
(M) (smooth (r 1)-form);
(ii) i
X
i
X
= 0;
(iii) i
X
is an anti-derivation, i.e. for
r
(M) and
s
(M),
i
X
() = (i
X
) +(1)
r
(i
X
).
A direct consequence of (iii) is that if =
1

r
, where
i

1
(M) =
F

(M), then
(10) i
X
= (1)
i1

i
(X)
1

i

r
,
where the hat indicates that
i
is to be omitted, and we use the summation conven-
tion.
J 10.3 Example. Let = x
2
dx
1
dx
3
be a 2-form on R
3
, and X = x
2
1

x
1
+x
3

x
2
+
(x
1
+x
2
)

x
3
a given vector eld on R
3
. If Y =Y
1
y
1
+Y
2
y
2
+Y
3
y
3
is an arbitrary
vector elds then
(i
X
)(Y) =(X,Y) = dx
1
(X)dx
3
(Y) dx
1
(Y)dx
3
(X)
= x
2
1
Y
3
(x
1
+x
2
)Y
1
,
which gives that
i
X
= x
2
1
dx
3
(x
1
+x
2
)dx
1
.
I
29
See Lee, Lemma 13.11.
68
Since is a 2-form the calculation using X,Y is still doable. For higher forms
this becomes to involved. If we use the multilinearity of forms we can give a simple
procedure for computing i
X
using Formula (10).
J 10.4 Example. Let = dx
1
dx
2
dx
3
be a 3-form on R
3
, and X the vector eld
as given in the previous example. By linearity
i
X
= i
X
1
+i
X
2
+i
X
3
,
where X
1
= x
2
1

x
1
, X
2
= x
3

x
2
, and X
3
= (x
1
+x
2
)

x
3
. This composition is chosen
so that X is decomposed in vector elds in the basis directions. Now
i
X
1
= dx
1
(X
1
)dx
2
dx
3
= x
2
1
dx
2
dx
3
,
i
X
2
= dx
2
(X
2
)dx
1
dx
3
=x
3
dx
1
dx
3
,
i
X
3
= dx
2
(X
3
)dx
1
dx
2
= (x
1
+x
2
)dx
1
dx
2
,
which gives
i
X
= x
2
1
dx
2
dx
3
x
3
dx
1
dx
3
+(x
1
+x
2
)dx
1
dx
2
,
a 2-form on R
3
. One should now verify that the same answer is obtained by com-
puting (X,Y, Z). I
For completeness we recall that for a smooth mapping f : N M, the pullback
of a r-form is given by
( f

)
p
(X
1
, , X
r
) = f

f (p)
(X
1
, , X
r
) =
f (p)
( f

X
1
, , f

X
r
).
We recall that for a mapping h : M R, then pushforward, or differential of h
dh
p
= h

p
M. In coordinates dh
p
=

h
x
i
dx
i
|
p
, and thus the mapping p 7dh
p
is a smooth section in
1
(M), and therefore a differential 1-form, with component

i
=

h
x
i
(in local coordinates).
If f : N M is a mapping between m-dimensional manifolds with charts (U, ),
and (V, ) respectively, and f (U) V. Set x =(p), and y =(q), then
(11) f

_
dy
1
dy
m
_
= ( f )det
_
J

f |
x
_
dx
1
dx
m
.
This can be proved as follows. From the denition of the wedge product and
Lemma 9.8 it follows that f

(dy
1
dy
m
) = f

dy
1
f

dy
m
, and f

dy
j
=

f
j
x
i
dx
i
= dF
j
, where F = f and

f = f
1
. Now
f

(dy
1
dy
m
) = f

dy
1
f

dy
m
= dF
1
dF
m
,
69
and furthermore, using (7),
dF
1
dF
m

x
1
, ,

x
m

= det
_
dF
i

x
j

_
= det

f
j
x
j

,
which proves the above claim.
As a consequence of this a change of coordinates

f =
1
yields
(12) f

(dy
1
dy
m
) = det
_
J

f |
x
_
dx
1
dx
m
.
J10.5 Example. Consider =dxdy on R
2
, and mapping f given by x =r cos()
and y = r sin(). The map f the identity mapping that maps R
2
in Cartesian coor-
dinates to R
2
in polar coordinates (consider the chart U = {(r, ) : r > 0, 0 < <
2}). As before we can compute the pullback of to R
2
with polar coordinates:
= dx dy = d(r cos()) d(r sin())
= (cos()dr r sin()d) (sin()dr +r cos()d)
= r cos
2
()dr dr sin
2
()ddr
= rdr d.
Of course the same can be obtained using (12). I
J 10.6 Remark. If we dene
(M) =

M
r=0

r
(M),
which is an associative, anti-commutative graded algebra, then f

: (N) (M)
is a algebra homomorphism. I
11. Orientations
In order to explain orientations on manifolds we rst start with orientations of
nite-dimensional vector spaces. Let V be a real m-dimensional vector space. Two
ordered basis {v
1
, , v
m
} and {v
0
1
, , v
0
m
} are said to be consistently oriented if
the transition matrix A = (a
i j
), dened by the relation
v
i
= a
i j
v
0
j
,
has positive determinant. This notion denes an equivalence relation on ordered
bases, and there are exactly two equivalence classes. An orientation for V is a
choice of an equivalence class of order bases. Given an ordered basis {v
1
, , v
m
}
the orientation is determined by the class O= [v
1
, , v
m
]. The pair (V, O) is called
70
an oriented vector space. For a given orientation any ordered basis that has the
same orientation is called positively oriented, and otherwise negatively oriented.
Lemma 11.1. Let 0 6=
m
(V), then the set of all ordered bases {v
1
, , v
m
}
for which (v
1
, , v
m
) > 0 is an orientation for V.
Proof: Let {v
0
i
} be any ordered basis and v
i
= a
i j
v
0
j
. Then (v
1
, , v
m
) =
det(A)(v
0
1
, , v
0
m
) > 0, which proves that the set of bases {v
i
} for which it holds
that (v
1
, , v
m
) > 0, characterizes an orientation for V.
Let us now describe orientations for smooth manifolds M. We will assume that
dimM = m 1 here. For each point p M we can choose an orientation O
p
for
the tangent space T
p
M, making (T
p
M, O
p
) oriented vector spaces. The collection
{O
p
}
pM}
is called a pointwise orientation. Without any relation between these
choices this concept is not very useful. The following denition relates the choices
of orientations of T
p
M, which leads to the concept of orientation of a smooth man-
ifold.
Denition 11.2. Asmooth m-dimensional manifold M with a pointwise orientation
{O
p
}
pM
, O
p
=

X
1
, , X
m

, is (positively) oriented if for each point in M there

exists a neighborhood U and a diffeomorphism , mapping from U to an open
subset of R
m
, such that
(13)

(X
1
), ,

(X
m
)

p
= [e
1
, , e
m
], p U,
where [e
1
, , e
m
] is the standard orientation of R
m
. In this case the pointwise
orientation {O
p
}
pM
is said to be consistently oriented. The choice of {O
p
}
pM
is
called an orientation on M, denoted by O = {O
p
}
pM
.
If a consistent choice of orientations O
p
does not exist we say that a manifold is
non-orientable.
J 11.3 Remark. If we look at a single chart (U, ) we can choose orientations O
p
that are consistently oriented for p O
p
. The choices X
i
=
1

(e
i
) =

x
i
|
p
are
ordered basis for T
p
M. By denition

(X
i
) =e
i
, and thus by this choice we obtain
a consistent orientation for all p U. This procedure can be repeated for each chart
in an atlas for M. In order to get a globally consistent ordering we need to worry
about the overlaps between charts. I
J11.4 Example. Let M =S
1
be the circle in R
2
, i.e. M ={p = (p
1
, p
2
) : p
2
1
+p
2
2
=
1}. The circle is an orientable manifold and we can nd a orientation as follows.
Consider the stereographic charts U
1
= S
1
\Np and U
2
= S
1
\Sp, and the associated
71
mappings

1
(p) =
2p
1
1p
2
,
2
(p) =
2p
1
1+ p
2

1
1
(x) =

4x
x
2
+4
,
x
2
4
x
2
+4

,
1
2
(x) =

4x
x
2
+4
,
4x
2
x
2
+4

.
Let us start with a choice of orientation and verify its validity. Choose X(p) =
(p
2
, p
1
), then
(
1
)

(X) = J
1
|
p
(X) =

2
1p
2
2p
1
(1p
2
)
2

_
p
2
p
1
_
=
2
1p
2
> 0,
on U
1
(standard orientation). If we carry out the same calculation for U
2
we obtain
(
2
)

(X) = J
2
|
p
(X) =
2
1+p
2
< 0 on U
2
, with corresponds to the opposite orienta-
tion. Instead by choosing
2
(p) =
2
(p
1
, p
2
) we obtain
(
2
)

(X) = J
2
|
p
(X) =
2
1+ p
2
> 0,
which chows that X(p) is denes an orientation O on S
1
. The choice of vectors
X(p) does not have to depend continuously on p in order to satisfy the denition
of orientation, as we will explain now.
For p U
1
choose the canonical vectors X(p) =

x
|
p
, i.e.

p
= J
1
1
(1) =
_
164x
2
(x
2
+4)
2
16x
(x
2
+4)
2
_
.
In terms of p this gives X(p) =
1
2
(1p
2
)
_
p
2
p
1
_
. By denition (
1
)

(X) = 1
for p U
1
. For p =Np we choose X(p) = (1 0)
t
, so X(p) is dened for p S
1
and is not a continuous function of p! It remains to verify that (
2
)

(X) > 0 for

some neighborhood of Np. First, at p = Np,
J
2
|
p
_
1
0
_
=
2
1+ p
2
= 1 > 0.
The choice of X(p) at p = Np comes from the canonical choice

x

p
with respect
to
2
. Secondly,
J
2
|
p
(X(p)) =
1
2
(1p
2
)

2
1+ p
2
2p
1
(1+ p
2
)
2

_
p
2
p
1
_
=
1p
2
1+ p
2
,
for all p U
2
\Np. Note that
2

1
1
(x) =
4
x
, and
J
2
|
p
(X(p)) = J
2
|
p
(J
1
1
|
p
(1)) = J(
2

1
1
)|
p
(1) =
4
x
2
,
72
which shows that consistency at overlapping charts can be achieved if the transition
mappings have positive determinant. Basically, the above formula describes the
change of basis as was carried out in the case of vector spaces. I
What the above example shows us is that if we choose X(p) =

x
i
|
p
for all
charts, it remains to be veried if we have the proper collection of charts, i.e. does

{J

J
1

(e
i
)}

= [{e
i
}] hold? This is equivalent to having det(J

J
1

) >0.
These observations lead to the following theorem.
Theorem 11.5. A smooth manifold M is orientable if and only if there exists an
atlas A= {(U

} such that for any p M

det(J

|
x
) > 0,

,
for any pair , for which x =

(p), and p U

.
Proof: Obvious from the previous example and Lemma 11.1.
As we have seen for vector spaces m-forms can be used to dene an orientation
on a vector space. This concept can also be used for orientations on manifolds.
Theorem 11.6.
30
Let M be a smooth m-dimensional manifold. A nowhere van-
ishing differential m-form
m
(M) determines a unique orientation O on M for
which
p
is positively orientated at each T
p
M. Conversely, given an orientation O
on M, then there exists a nowhere vanishing m-form
m
(M) that is positively
oriented at each T
p
M.
Proof: Let
m
(M) be a nowhere vanishing m-form, then in local coordinates
(U, ), = f dx
1
dx
m
. For the canonical bases for T
p
M we have

x
1
, ,

x
m

= f 6= 0.
Without loss of generality can be assumed to be positive, otherwise change by
means of x
1
x
1
. Now let {

y
i
} be canonical bases for T
p
M with respect to a
chart (V, ). As before = gdy
1
dy
m
, g > 0 on V. Assume that U V 6=,
then at the overlap we have
0 < g =

y
1
, ,

y
m

= det

J(
1
)

x
1
, ,

x
m

= f det

J(
1
)

,
which shows that det

J(
1
)

> 0, and thus M is orientable.

For the converse we construct non-vanishing m-forms on each chart (U, ). Via
a partition of unity we can construct a smooth m-form on M, see Lee.
30
See Lee, Prop. 13.4.
73
This theorem implies in particular that non-orientable m-dimensional manifolds
do not admit a nowhere vanishing m-form, or volume form.
J 11.7 Example. Consider the M obius strip M. The M obius strip is an example
of a non-orientable manifold. Let us parametrize the M obius strip as an embedded
manifold in R
3
:
g : R(1, 1) R
3
, g(, r) =
_
_
_
r sin(/2)cos() +cos()
r sin(/2)sin() +sin()
r cos(/2)
_
_
_,
where g is a smooth embedding when regarded as a mapping from R/2Z
(1, 1) R
3
. Let us assume that the M obius strip M is oriented, then by the
above theorem there exists a nonwhere vanishing 2-form which can be given as
follows
= a(x, y, z)dy dz +b(x, y, z)dz dx +c(x, y, z)dx dy,
where (x, y, z) = g(, r). Since g is a smooth embedding (parametrization) the
pullback form g

= (, r)d dr is a nowhere vanishing 2-form on R/2Z

(1, 1). In particular, this means that is 2-periodic in . Notice that =
i
X
dx dy dz, where
X = a(x, y, z)

x
+b(x, y, z)

y
+c(x, y, z)

z
.
For vector elds =
1

x
+
2

y
+
3

z
and =
1

x
+
2

y
+
3

z
we then have
that
(, ) = i
X
dx dy dz(, ) = X (),
where is the cross product of and . The condition that is nowhere
vanishing can be translated to R
2
as follows:
g

(e
1
, e
2
) =(g

(e
1
), g

(e
2
)).
Since g

(e
1
) g

(e
2
)|
=0
=g

(e
1
) g

(e
2
)|
=2
, the pullback g

form cannot
be nowhere vanishing on R/2ZR, which is a contradiction. This proves that
the M obius band is a non-orientable manifold. I
Let N, M be oriented manifolds of the same dimension, and let f : N M be
a smooth mapping. A mapping is called orientation preserving at a point p N
if f

maps positively oriented bases of T

p
N to positively oriented bases in T
f (p)
M.
A mapping is called orientation reversing at p N if f

maps positively oriented

bases of T
p
N to negatively oriented bases in T
f (p)
M.
Let us now look at manifolds with boundary; (M, M). For a point p M we
distinguish three types of tangent vectors:
74
FIGURE 32. Out(in)ward vectors and the induced orientation to
M.
(i) tangent boundary vectors X T
p
(M) T
p
M, which form an (m1)-
dimensional subspace of T
p
M;
(ii) outward vectors; let
1
: W H
m
M, then X T
p
M is called an out-
ward vector if
1

(Y) = X, for some Y = (y

1
, , y
m
) with y
1
< 0;
(iii) inward vectors; let
1
: W H
m
M, then X T
p
M is called an inward
vector if
1

(Y) = X, for some Y = (y

1
, , y
m
) with y
1
> 0.
Using this concept we can now introduce the notion of induced orientation on M.
Let p M and choose a basis {X
1
, , X
m
} for T
p
M such that [X
1
, , X
m
] =O
p
,
{X
2
, , X
m
} are tangent boundary vectors, and X
1
is an outward vector. In this case
[X
2
, , X
m
] = (O)
p
determines an orientation for T
p
(M), which is consistent,
and therefore O = {(O)
p
}
pM
is an orientation on M induced by O. Thus for
an oriented manifold M, with orientation O, M has an orientation O, called the
induced orientation on M.
J 11.8 Example. Any open set M R
m
(or H
m
) is an orientable manifold. I
J 11.9 Example. Consider a smooth embedded co-dimension 1 manifold
M = {p R
m+1
: f (p) = 0}, f : R
m+1
R, rk( f )|
p
, p M.
Then M is an orientable manifold. Indeed, M =N, where N ={p R
m+1
: f (p) >
0}, which is an open set in R
m+1
and thus an oriented manifold. Since M =N the
manifold M inherits an orientation from M and hence it is orientable. I
75
IV. Integration on manifolds
12. Integrating m-forms on R
m
We start off integration of m-forms by considering m-forms on R
m
.
Denition 12.1. A subset D R
m
is called a domain of integration if
(i) D is bounded, and
(ii) D has m-dimensional Lebesgue measure d = dx
1
dx
m
equal to zero.
In particular any nite union or intersection of open or closed rectangles is a
domain of integration. Any bounded
31
continuous function f on D is integrable,
i.e.
<
Z
D
f dx
1
dx
m
<.
Since
m
(R
m
)

=R, a smooth m-form on R

m
is given by
= f (x
1
, , x
m
)dx
1
dx
m
,
where f : R
m
R is a smooth function. For a given (bounded) domain of integra-
tion D we dene
Z
D
:=
Z
D
f (x
1
, , x
m
)dx
1
dx
m
=
Z
D
f d
=
Z
D

x
(e
1
, , e
m
)d.
An m-form is compactly supported if supp() = cl{x R
m
: (x) 6= 0}
is a compact set. The set of compactly supported m-forms on R
m
is denoted by

m
c
(R
m
), and is a linear subspace of
m
(R
m
). Similarly, for any open set U R
m
we can dene
m
c
(U). Clearly,
m
c
(U)
m
c
(R
m
), and can be viewed as a
linear subspace via zero extension to R
m
. For any open set U R
m
there exists a
domain of integration D such that U D supp() (see Exercises).
31
Boundedness is needed because the rectangles are allowed to be open.
76
Denition 12.2. Let U R
m
be open and
m
c
(U), and let D be a domain of
integration D such that U D supp(). We dene the integral
Z
U
:=
Z
D
.
If U H
m
open, then
Z
U
:=
Z
DH
m
.
The next theorem is the rst step towards dening integrals on m-dimensional
manifolds M.
Theorem 12.3. Let U,V R
m
be open sets, f : U V an orientation preserving
diffeomorphism, and let
m
c
(V). Then,
Z
V
=
Z
U
f

.
If f is orientation reversing, then
R
U
=
R
V
f

.
Proof: Assume that f is an orientation preserving diffeomorphism from U to
V. Let E be a domain a domain of integration for , then D = f
1
(E) is a do-
main of integration for f

. We now prove the theorem for the domains D and

E. We use coordinates {x
i
} and {y
i
} on D and E respectively. We start with
= g(y
1
, , y
m
)dy
1
dy
m
. Using the change of variables formula for inte-
grals and the pullback formula in (11) we obtain
Z
E
=
Z
E
g(y)dy
1
dy
m
(Denition)
=
Z
D
(g f )(x)det(J

f |
x
)dx
1
dx
m
=
Z
D
(g f )(x)det(J

f |
x
)dx
1
dx
m
=
Z
D
f

(Denition).
One has to introduce a sign in the orientation reversing case.
13. Partitions of unity
We start with introduce the notion partition of unity for smooth manifolds. We
should point out that this denition can be used for arbitrary topological spaces.
77
Denition 13.1. Let M be smooth m-manifold with atlas A = {(
i
,U
i
)}
iI
. A
partition of unity subordinate to A is a collection for smooth functions {
i
: M
R}
iI
satisfying the following properties:
(i) 0
i
(p) 1 for all p M and for all i I;
(ii) supp(
i
) U
i
;
(iii) the set of supports {supp(
i
)}
iI
is locally nite;
(iv)
iI

i
(p) = 1 for all p M.
Condition (iii) says that every p M has a neighborhood U 3 p such that only
nitely many
i
s are nonzero in U. As a consequence of this the sum in Condition
(iv) is always nite.
Theorem 13.2. For any smooth m-dimensional manifold M with atlas A there
exists a partition of unity {
i
} subordinate to A.
In order to prove this theorem we start with a series of auxiliary results and
notions.
Lemma 13.3. There exists a smooth function h : R
m
R such that 0 h 1 on
R
m
, and h|
B
1
(0)
1 and supp(h) B
2
(0).
Proof: Dene the function f
1
: R R as follows
f
1
(t) =
_
_
_
e
1/t
, t > 0,
0, t 0.
One can easily prove that f
1
is a C

-function on R. If we set
f
2
(t) =
f
1
(2t)
f
1
(2t) + f
1
(t 1)
.
This function has the property that f
2
(t) 1 for t 1, 0 < f
2
(t) < 1 for 1 < t <
2, and f
2
(t) 0 for t 2. Moreover, f
2
is smooth. To construct f we simply
write f (x) = f
2
(|x|) for x R
m
\{0}. Clearly, f is smooth on R
m
\{0}, and since
f |
B
1
(0)
1 it follows that f is smooth on R
m
.
An atlas A gives an open covering for M. The set U
i
in the atlas need not be
compact, nor locally nite. We say that a covering U = {U
i
} of M is locally nite
if every point p M has a neighborhood that intersects only nitely U
i
U. If
there exists another covering V = {V
j
} such that every V
j
V is contained in some
V
j
U
i
U, then V is called a renement of U. A topological space X for which
each open covering admits a locally nite renement is called paracompact.
Lemma 13.6. Any topological manifold M allows a countable, locally nite cov-
ering by precompact open sets.
78
FIGURE 33. The functions f
1
and f
2
.
FIGURE 34. A locally nite covering [left], and a renement
[right].
Proof: We start with a countable covering of open balls B = {B
i
}. We now
construct a covering U that satises
(i) all sets U
i
U are precompact open sets in M;
(ii) U
i1
U
i
, i > 1;
(iii) B
i
U
i
.
We build the covering U from B using an inductive process. Let U
1
= B
1
, and
assume U
1
, ,U
k
have been constructed satisfying (i)-(iii). Then
U
k
B
i
1
B
i
k
,
where B
i
1
= B
1
. Now set
U
k+1
= B
i
1
B
i
k
.
Choose i
k
large enough so that i
k
k +1 and thus B
k+1
U
k+1
.
From the covering U we can now construct a locally nite covering V by setting
V
i
=U
i
\U
i2
, i > 1.
79
FIGURE 35. Constructing a nested covering U from a covering
with balls B [left], and a locally nite covering obtained from the
previous covering [right].
Next we seek a special locally nite renement {W
j
} =Wof V which has spe-
cial properties with respect to coordinate charts of M;
(i) Wis a countable and locally nite;
(ii) each W
j
Wis in the domain of some smooth coordinate map
j
:
j
R
m
for M;
(iii) the collection Z = {Z
j
}, Z
j
=
1
j
(B
1
(0)) covers M.
The covering Wis called regular.
Lemma 13.8. For any open covering U for a smooth manifold M there exists a
regular renement. In particular M is paracompact.
Proof: Let V be a countable, locally nite covering as described in the previous
lemma. Since V is locally nite we can nd a neighborhood W
p
for each p M
that intersects only nitely many set V
j
V. We want to choose W
p
in a smart
way. First we replace W
p
by W
p
{V
j
: p V
j
}. Since p U
i
for some i we then
replace W
p
by W
p
V
i
. Finally, we replace W
p
by a small coordinate ball B
r
(p)
so that W
p
is in the domain of some coordinate map
p
. This provides coordinate
charts (W
p
,
p
). Now dene Z
p
=
1
p
(B
1
(0)).
For every k, {Z
p
: p V
k
} is an open covering of V
k
, which has a nite subcov-
ering, say Z
1
k
, , Z
m
k
k
. The sets Z
i
k
are sets of the form Z
p
i
for some p
i
V
k
. The
associated coordinate charts are (W
1
k
,
1
k
), , (W
m
k
k
,
m
k
k
). Each W
i
k
is obtained via
the construction above for p
i
V
k
. Clearly, {W
i
k
} is a countable open covering of
M that renes U and which, by construction, satises (ii) and (iii) in the denition
of regular renement. It is clear from compactness that {W
i
k
} is locally nite. Let
80
FIGURE 36. The different stages of constructing the sets W
p
col-
ored with different shades of grey.
us relabel the sets {W
i
k
} and denote this covering by W= {W
i
}. As a consequence
M is paracompact.
Proof of Theorem 13.2: From Lemma 13.8 there exists a regular renement W
for A. By construction we have Z
i
=
1
i
(B
1
(0)), and we dene

Z
i
:=
1
i
(B
2
(0)).
Using Lemma 13.3 we now dene the functions
:
M R;

i
=
_
_
_
f
2

i
on W
i
0 on M\V
i
.
These functions are smooth and supp(
i
) W
i
. The quotients

=

j
(p)

i
(p)
,
are, due to the local niteness of W
i
and the fact that the denominator is positive
for each p M, smooth functions on M. Moreover, 0

j
1, and
j

j
1.
Since W is a renement of A we can choose an index k
j
such that W
j
U
k
j
.
Therefore we can group together some of the function

j
:

i
=

j : k
j
=i

j
,
which give us the desired partition functions with 0
i
1,
i

i
1, and
supp(
i
) U
i
.
Some interesting byproducts of the above theorem using partitions of unity are.
In all these case M is assumed to be an smooth m-dimensional manifold.
81
Theorem 13.10.
32
For any close subset A M and any open set U A, there
exists a smooth function f : M R such that
(i) 0 f 1 on M;
(ii) f
1
(1) = A;
(iii) supp( f ) U.
Such a function f is called a bump function for A supported in U.
By considering functions g =c(1 f ) 0 we obtain smooth functions for which
g
1
(0) can be an arbitrary closed subset of M.
Theorem 13.11.
33
Let A M be a closed subset, and f : A R
k
be a smooth
mapping.
34
Then for any open subset U M containing A there exists a smooth
mapping f

: M R
k
such that f

|
A
= f , and supp( f

) U.
14. Integration on of m-forms on m-dimensional manifolds.
In order to introduce the integral of an m-form on M we start with the case of
forms supported in a single chart. In what follows we assume that M is an oriented
manifold with an oriented atlas A, i.e. a consistent choice of smooth charts such
that the transitions mappings are orientation preserving.
FIGURE 37. An m-form supported in a single chart.
32
See Lee, Prop. 2.26.
33
See Lee, Prop. 2.26.
34
A mapping from a closed subset A M is said to be smooth if for every point p A there exists
an open neighborhood W M of p, and a smooth mapping f

: W R
k
such that f

|
WA
= f .
82
Let
m
c
(M), with supp() U for some chart U in A. Now dene the
integral of over M as follows:
Z
M
:=
Z
(U)
(
1
)

,
where the pullback form (
1
)

is compactly supported in V = (U) ( is a

diffeomorphism). The integral over V of the pullback form (
1
)

is dened in
Denition 12.2. It remains to show that the integral
R
M
does not depend on the
particular chart. Consider a different chart U
0
, possibly in a different oriented atlas
FIGURE 38. Different charts U and U
0
containing supp(), i.e.
supp() U U
0
.
A
0
for M (same orientation), then
Z
V
0
(
01
)

=
Z

0
(UU
0
)
(
01
)

=
Z
(UU
0
)
(
0

1
)

(
01
)

=
Z
(UU
0
)
(
1
)

(
0
)

(
01
)

=
Z
(UU
0
)
(
1
)

=
Z
V
(
1
)

,
which show that the denition is independent of the chosen chart. We crucially do
use the fact that AA
0
is an oriented atlas for M. To be more precise the mappings

1
are orientation preserving.
By choosing a partition of unity as described in the previous section we can now
dene the integral over M of arbitrary m-forms
m
c
(M).
83
Denition 14.3. Let
m
c
(M) and let A
I
= {(U
i
,
i
}
iI
A be a nite subcov-
ering of supp() coming from an oriented atlas A for M. Let {
i
} be a partition
of unity subordinate to A
I
. Then the integral of over M is dened as
Z
M
:=

i
Z
M

i
,
where the integrals
R
M

i
are integrals of form that have support in single charts
as dened above.
We need to show now that the integral is indeed well-dened, i.e. the sum is
nite and independent of the atlas and partition of unity. Since supp() is compact
A
I
exists by default, and thus the sum is nite.
Lemma 14.4. The above denition is independent of the chosen partition of unity
and covering A
I
.
Proof: Let A
0
J
A
0
be another nite covering of supp(), where A
0
is a com-
FIGURE 39. Using a partition of unity we can construct m-forms
which are all supported in one, but possibly different charts U
i
.
patible oriented atlas for M, i.e. AA
0
is a oriented atlas. Let {
0
j
} be a partition
of unity subordinate to A
0
J
. We have
Z
M

i
=
Z
M

0
j

i
=

j
Z
M

0
j

i
.
By summing over i we obtain
i
R
M

i
=
i, j
R
M

0
j

i
. Each term
0
j

i
is sup-
ported in some U
i
and by previous independent of the coordinate mappings. Sim-
ilarly, if we interchange the is and js, we obtain that
j
R
M

0
j
=
i, j
R
M

0
j

i
,
which proves the lemma.
84
J 14.6 Remark. If M is a compact manifold that for any
m
(M) it holds that
supp() M is a compact set, and therefore
m
c
(M) =
m
(M) in the case of com-
pact manifolds. I
So far we considered integral of m-forms over M. One can of course integrate
n-forms on M over n-dimensional immersed or embedded submanifolds N M.
Given
n
(M) for which the restriction to N, |
N
has compact support in N,
then
Z
N
:=
Z
N
|
N
.
As a matter of fact we have i : N M, and |
N
= i

, so that
R
N
:=
R
N
|
N
=
R
N
i

. If M is compact with boundary M, then

R
M
is well-dened for any
(m1)-form
m1
(M).
Theorem 14.7. Let
m
c
(M), and f : N M is a diffeomorphism. Then
Z
M
=
Z
N
f

.
Proof: By the denition of integral above we need to prove the above statement
for the terms
Z
M

i
=
Z
N
f

i
.
Therefore it sufces to prove the theorem for forms whose support is in a single
chart U.
FIGURE 40. The pullback of an m-form.
Z
U
f

=
Z
(U)
(
1
)

=
Z
(U)
(
1
)

(
1
)

=
Z
(U
0
)
(
1
)

=
Z
U
0
.
85
By applying this to
i
and summing we obtain the desired result.
For sake of completeness we summarize the most important properties of the
integral.
Theorem 14.9.
35
Let N, M be oriented manifolds (with or without boundary) of
dimension m , and ,
m
c
(M), are m-forms. Then
(i)
R
M
a+b = a
R
M
+b
R
M
;
(ii) if O is the opposite orientation to O, then
Z
M,O
=
Z
M,O
;
(iii) if is an orientation form, then
R
M
> 0;
(iv) if f : N M is a diffeomorphism, then
Z
M
=
Z
N
f

.
For practical situations the denition of the integral above is not convenient since
constructing partitions of unity is hard in practice. If the support of a differential
form can be parametrized be appropriate parametrizations, then the integral can
be easily computed from this. Let D
i
R
m
nite set of indices i be compact
domains of integration, and g
i
: D
i
M are smooth mappings satisfying
(i) E
i
= g
i
(D
i
), and g
i
: int (D
i
) int (E
i
) are orientation preserving diffeo-
morphisms;
(ii) E
i
E
j
intersect only along their boundaries, for all i 6= j.
Theorem 14.10.
36
Let {(g
i
, D
i
)} be a nite set of parametrizations as dened
above. Then for any
m
c
(M) such that supp()
i
E
i
it holds that
Z
M
=

i
Z
D
i
g

i
.
Proof: As before it sufces the prove the above theorem for a single chart U,
i.e. supp() U. One can choose U to have a boundary U so that (U) has
measure zero, and maps cl(U) to a compact domain of integration K H
m
. Now
set
A
i
= cl(U) E
i
, B
i
= g
1
i
(A
i
), C
i
=
i
(A
i
).
We have
Z
C
i
(
1
)

=
Z
B
i
(g
i
)

(
1
)

=
Z
B
i
g

i
.
35
See Lee Prop. 14.6.
36
See Lee Prop. 14.7.
86
FIGURE 41. Carving up supp() via domains of integration for
parametrizations g
i
for M.
Since the interiors of the sets C
i
(and thus A
i
) are disjoint it holds that
Z
M
=
Z
K
(
1
)

i
Z
C
i
(
1
)

=

i
Z
B
i
g

i
=

i
Z
D
i
g

i
,
which proves the theorem.
J 14.12 Remark. From the previous considerations computing
R
M
boils down
to computing (
1
)

, or g

for appropriate parametrizations, and summing the

various contrubutions. Recall from Section 12 that in order to integrate one needs
to evaluate (
1
)

x
(e
1
, , e
m
), which is given by the formula
_
(
1
)

_
x
(e
1
, , e
m
) =

1
(x)
(
1

(e
1
), ,
1

(e
m
)).
For a single chart contribution this then yields the formula
Z
U
=
Z
(U)

1
(x)
(
1

(e
1
), ,
1

(e
m
))d
=
Z
(U)

1
(x)

x
1
, ,

x
1

d
or in the case of a parametrization g : D M:
Z
g(D)
=
Z
D

g(x)
(g

(e
1
), , g

(e
m
))d.
These expression are useful for computing integrals. I
87
J 14.13 Example. Consider the 2-sphere parametrized by the mapping g : R
2

S
2
R
3
given as
g(, ) =
_
_
_
x
y
z
_
_
_=
_
_
_
sin()cos()
sin()sin()
cos()
_
_
_.
This mapping can be viewed as a covering map. From this expression we derive
various charts for S
2
. Given the 2-form = zdx dz let us compute the pullback
form g

on R
2
. We have that
g

(e
1
) = cos()cos()

x
+cos()sin()

y
sin()

z
,
g

(e
2
) = sin()sin()

x
+sin()cos()

y
,
and therefore
g

(e
1
, e
2
) =
g(x)
(g

(e
1
), g

(e
2
)) =cos()sin
2
()sin(),
and thus
g

=cos()sin
2
()sin()dd.
The latter gives that
R
S
2 =
R
2
0
R

0
cos()sin
2
()sin()dd = 0, which shows
that is not a volume form on S
2
.
If we perform the same calculations for =xdydz+ydzdx+zdxdy, then
Z
S
2
=
Z
2
0
Z

0
sin()dd = 4.
The pullback form g

= sin()dd, which shows that is a volume form on

S
2
. I
15. The exterior derivative
The exterior derivative d of a differential form is an operation that maps an
k-form to a (k +1)-form. Write a k-form on R
m
in the following notation
=
i
1
i
k
dx
i
1
dx
i
k
=
I
dx
I
, I = (i
1
i
k
),
then we dene
(14) d = d
I
dx
I
.
Written out in all its differentials this reads
d = d

i
1
i
k
dx
i
1
dx
i
k

=

I
x
i
dx
i
dx
i
1
dx
i
k
.
88
Of course the Einstein summation convention is again applied here. This is the
denition that holds for all practical purposes, but it is a local denition. We need
to prove that d extends to differential forms on manifolds.
J 15.1 Example. Consider
0
= f (x, y), a 0-form on R
2
, then
d
0
=
f
x
dx +
f
y
dy.
For a 1-form
1
= f
1
(x, y)dx + f
2
(x, y)dy we obtain
d
1
=
f
1
x
dx dx +
f
1
y
dy dx +
f
2
x
dx dy +
f
2
y
dy dy
=
f
1
y
dy dx +
f
2
x
dx dy =

f
2
x

f
1
y

dx dy.
Finally, for a 2-form
2
= g(x, y)dx dy the d-operation gives
d
2
=
g
x
dx dx dy +
g
y
dy dx dy = 0.
The latter shows that d applied to a top-form always gives 0. I
J 15.2 Example. In the previous example d
0
is a 1-form, and d
1
is a 2-form.
Let us now apply the d operation to these forms:
d(d
0
) =

2
f
xy

2
f
yx

dx dy = 0,
and
d(d
1
) = 0,
since d acting on a 2-form always gives 0. These calculations seem to suggest that
in general d d = 0. I
As our examples indicate d
2
=0. One can also have forms for which d=0,
but 6= d. We say that is a closed form, and when = d, then is called
an exact form. Clearly, closed forms form a possibly larger class than exact forms.
In the next chapter on De Rham cohomology we are going to come back to this
phenomenon in detail. On R
m
one will not be able to nd closed forms that are
not exact, i.e. on R
m
all closed forms are exact. However, if we consider different
manifolds we can have examples where this is not the case.
J 15.3 Example. Let M =R
2
\{(0, 0)}, and consider the 1-form
=
y
x
2
+y
2
dx +
x
x
2
+y
2
dy,
89
which clearly is a smooth 1-form on M. Notice that does not extend to a 1-form
on R
2
! It holds that d= 0, and thus is a closed form on M. Let : [0, 2) M,
t 7(cos(t), sin(t)) be an embedding of S
1
into M, then
Z

=
Z
2
0

=
Z
2
0
dt = 2.
Assume that is an exact 1-form on M, then = d f , for some smooth function
f : M R. In Section 5 we have showed that
Z

=
Z

d f =
Z
2
0

d f = ( f )
0
(t)dt = f (1, 0) f (1, 0) = 0,
which contradicts the fact that
R

=
R

d f = 2. This proves that is not exact.

I
For the exterior derivative in general we have the following theorem.
Theorem 15.4.
37
Let M be a smooth m-dimensional manifold. Then for all k 0
there exist unique linear operations d :
k
(M)
k+1
(M) such that;
(i) for any 0-form = f , f C

(M) it holds that

d(X) = d f (X) = X f , X F(M);
(ii) if
k
(M) and
`
(M), then
d() = d+(1)
k
d;
(iii) d d = 0;
(iv) if
m
(M), then d = 0.
This operation d is called the exterior derivative on differential forms, and is a
unique anti-derivation (of degree 1) on (M) with d
2
= 0.
Proof: Let us start by showing the existence of d on a chart U M. We have
local coordinates x =(p), p U, and we dene
d
U
:
k
(U)
k+1
(U),
via (14). Let us write d instead of d
U
for notational convenience. We have that
d( f dx
I
) = d f dx
I
. Due to the linearity of d it holds that
d( f dx
I
gdx
J
) = d( f gdx
I
dx
J
) = d( f g)dx
K
= gd f dx
K
+ f dgdx
K
= (d f dx
I
) (gdx
J
) + f dgdx
I
dx
J
= (d f dx
I
) (gdx
J
) +(1)
k
f dx
I
dgdx
J
= d( f dx
I
) (gdx
J
) +(1)
k
( f dx
I
) d(gdx
J
),
37
See Lee Thm. 12.14
90
which proves (ii). As for (iii) we argue as follows. In the case of a 0-form we have
that
d(d f ) = d

f
x
i
dx
i
=

2
f
x
j
x
i

dx
j
dx
i
=

i<j

2
f
x
i
x
j

2
f
x
j
x
i

dx
i
dx
j
= 0.
Given a k-form =
I
dx
I
, then, using (ii), we have
d(d) = d

d
I
dx
I

= d(d
I
) dx
I
+(1)
k+1
d
I
d(dx
I
) = 0,
since
I
is a 0-form, and d(dx
I
) = (1)
j
dx
i
1
d(dx
i
j
) dx
i
k
= 0. The
latter follows from d(dx
i
j
) = 0. The latter also implies (iv), nishing the existence
proof in the one chart case. The operation d = d
U
is well-dened, satisfying (i)-
(iv), for any chart (U, ).
The operation d
U
is unique, for if there exists yet another exterior derivative

d
U
,
which satises (i)-(iv), then for =
I
dx
I
,

d =

d
I
dx
I
+
I

d(dx
I
),
where we used (ii). From(ii) it also follows that

d(dx
I
) = (1)
j
dx
i
1

d(dx
i
j
)
dx
i
k
= 0. By (i) d((p)
i
j
) = dx
i
j
, and thus by (iv)

d(dx
i
j
) =

d((p)
i
j
) = 0,
which proves the latter. From (i) it also follows that

d
I
= d
I
, and therefore

d =

d
I
dx
I
+
I

d(dx
I
) = d
I
dx
I
= d,
which proves the uniqueness of d
U
.
Before giving the dening d for
k
(M) we should point out that d
U
trivially
satises (i)-(iii) of Theorem 15.5 (use (14). Since we have a unique operation d
U
for every chart U, we dene for p U
(d)
p
= (d
U
|
U
)
p
,
as the exterior derivative for forms on M. If there is a second chart U
0
3 p, then by
uniqueness it holds that on
d
U
(|
UU
0 ) = d
UU
0 (|
UU
0 ) = d
U
0 (|
UU
0 ),
which shows that the above denition is independent of the chosen chart.
The last step is to show that d is also uniquely dened on
k
(M). Let p U, a
coordinate chart, and consider the restriction
|
U
=
I
dx
I
,
91
where
I
C

(U). Let W U be an open set containing p, with the additional

property that cl(W) U is compact. By Theorem 13.10 we can nd a bump-
function g C

(M) such that g|

W
= 1, and supp(g) U. Now dene
= g
I
d(gx
i
1
) d(gx
i
k
).
Using (i) we have that d(gx
i
)|
W
= dx
i
, and therefore |
W
= |
W
. Set = ,
then |
W
= 0. Let p W and h C

(M) satisfying h(p) = 1, and supp(h) W.

Thus, h 0 on M, and
0 = d(h) = dh+hd.
This implies that (d)
p
=(d f )
p
= 0, which proves that (d )|
W
= (d)|
W
.
If we use (ii)-(iii) then
d = d
_
g
I
d(gx
i
1
) d(gx
i
k
)
_
= d(g
I
) d(gx
i
1
) d(gx
i
k
) +g
I
d
_
d(gx
i
1
) d(gx
i
k
)
_
= d(g
I
) d(gx
i
1
) d(gx
i
k
).
It now follows that since g|
W
= 1, and since (d )|
W
= (d)|
W
, that
(d)|
W
=

I
x
i
dx
i
dx
I
,
which is exactly (14). We have only used properties (i)-(iii) the derive this ex-
pression, and since p is arbitrary it follows that d :
k
(M)
k+1
(M) is uniquely
dened.
The exterior derivative has other important properties with respect to restrictions
and pullbacks that we now list here.
Theorem 15.5. Let M be a smooth m-dimensional manifold, and let
k
(M),
k 0. Then
(i) in each chart (U, ) for M, d in local coordinates is given by (14);
(ii) if =
0
on some open set U M, then also d = d
0
on U;
(iii) if U M is open, then d(|
U
) = (d)|
U
;
(iv) if f : N M is a smooth mapping, then
f

(d) = d( f

),
i.e. f

:
k
(M)
k
(N), and d commute as operations.
Proof: Let us restrict ourselves here to proof of (iv). It sufces to prove (iv) in a
chart U, and =
I
dx
I
. Let us rst compute f

(d):
f

(d) = f

_
d
I
dx
i
1
dx
i
k
_
= d(
I
f ) d( f )
i
1
d( f )
i
k
.
92
Similarly,
d( f

) = d
_
(
I
f ) d( f )
i
1
d( f )
i
k
_
= d(
I
f ) d( f )
i
1
d( f )
i
k
,
which proves the theorem.
16. Stokes Theorem
The last section of this chapter deals with a generalization of the Fundamental
Theorem of Integration: Stokes Theorem. The Stokes Theorem allows us to
compute
R
M
d in terms of a boundary integral for . In a way the (m1)-form
act as the primitive, anti-derivative of the d. Therefore if we are interested in
R
M
using Stokes Theorem we need to rst integrate , i.e. write = d. This
is not always possible as we saw in the previous section.
Stokes Theorem can be now be phrased as follows.
Theorem 16.1. Let M be a smooth m-dimensional manifold with or without bound-
ary M, and
m1
c
(M). Then,
(15)
Z
M
d =
Z
M
j

,
where j : M ,M is the natural embedding of the boundary M into M.
In order to prove this theorem we start with the following lemma.
Lemma 16.2. Let
m1
c
(H
m
) be given by
=
i
dx
1

dx
i
dx
m
.
Then,
(i)
R
H
M d = 0, if supp() int (H
m
);
(ii)
R
H
m d =
R
H
m j

, if supp() H
m
6=,
where j : H
m
,H
m
is the canonical inclusion.
Proof: We may assume without loss of generality that supp() [0, R]
[0, R] = I
m
R
. Now,
Z
H
m
d = (1)
i+1
Z
I
m
R

i
x
i
dx
1
dx
m
= (1)
i+1
Z
I
m1
R

i
|
x
i
=R

i
|
x
i
=0

dx
1

dx
i
dx
m
.
If supp() int I
m
R
, then
i
|
x
i
=R

i
|
x
i
=0
= 0 for all i, and
Z
H
m
d = 0.
93
If supp()H
m
6=, then
i
|
x
i
=R

i
|
x
i
=0
=0 for all i m1, and
m
|
x
m
=R
=0.
Therefore,
Z
H
m
d = (1)
i+1
Z
I
m
R

i
x
i
dx
1
dx
m
= (1)
i+1
Z
I
m1
R

i
|
x
i
=R

i
|
x
i
=0

dx
1

dx
i
dx
m
= (1)
m+1
Z
I
m1
R

i
|
x
m
=0

dx
1
dx
m1
= (1)
m
Z
I
m1
R

i
|
x
m
=0

dx
1
dx
m1
.
The mapping j : H
m
H
m
, which, in coordinates, is given by

j = j : (x
1
, , x
m1
) 7
(x
1
, , x
m1
, 0), where (x
1
, , x
m1
, 0) = (x
1
, , x
m1
). The latter mapping is
orientation preserving if m is even and orientation reversing if m is odd. Under the
mapping j we have that
j

(e
i
) = e
i
, i = 1, , m1,
where the e
i
s on the left hand side are the unit vectors in R
m1
, and the bold face
e
k
are the unit vectors in R
m
. We have that the induced orientation for H
m
is
obtained by the rotation e
1
e
m
, e
m
e
1
, and therefore
O = [e
2
, , e
m1
, e
1
].
Under this corresponds with the orientation [e
2
, , e
m1
, e
1
] of R
m1
, which is
indeed the standard orientation for m even, and the opposite orientation for m odd.
The pullback form on H
m
, using the induced orientation on H
m
, is given by
( j

)
(x
1
, ,x
m1
)
(e
2
, , e
m1
, e
1
)
=
(x
1
, ,x
m1
,0)
_
j

(e
2
), , j

(e
m2
), j

(e
1
)
_
= (1)
m

m
(x
1
, , x
m1
, 0),
where we used the fact that
dx
1
dx
m1
_
j

(e
2
), , j

(e
m1
), j

(e
1
))
_
= dx
1
dx
m1
(e
2
, , e
m1
, e
1
)
= (1)
m
dx
2
dx
m1
dx
1
(e
2
, , e
m1
, e
1
) = (1)
m
.
Combining this with the integral over H
m
we nally obtain
Z
H
m
d =
Z
H
m
j

,
which completes the proof.
94
Proof of Theorem 16.1: Let us start with the case that supp() U, where
(U, ) is an oriented chart for M. Then by the denition of the integral we have
that
Z
M
d =
Z
(U)
(
1
)

(d) =
Z
H
m
(
1
)

(d) =
Z
H
m
d

(
1
)

,
where the latter equality follows from Theorem 15.5. It follows from Lemma 16.2
that if supp() int (H
m
), then the latter integral is zero and thus
R
M

.
= 0. Also,
using Lemma 16.2, it follows that if supp() H
m
6=, then
Z
M
d =
Z
H
m
d

(
1
)

=
Z
H
m
j
0
(
1
)

=
Z
H
m
(
1
j
0
)

=
Z
M
j

,
where j
0
: H
m
H
m
is the canonical inclusion as used Lemma 16.2, and j =

1
j
0
: M M is the inclusion of the boundary of M into M.
Now consider the general case. As before we choose a nite covering A
I
A
of supp() and an associated partition of unity {
i
} subordinate to A
I
. Consider
the forms
i
, and using the rst part we obtain
Z
M
j

=

i
Z
M
j

i
=

i
Z
M
d(
i
)
=

i
Z
M

d
i
+
i
d

=
Z
M
d

+
Z
M

d
=
Z
M
d,
which proves the theorem.
J 16.3 Remark. If M is a closed (compact, no boundary), oriented manifold mani-
fold, then by Stokes Theorem for any
m1
(M),
Z
M
d = 0,
since M =. As a consequence volume form cannot be exact. Indeed,
R
M
>0,
and thus if were exact, then = d, which implies that
R
M
=
R
M
d = 0, a
contradiction. I
95
Let us start with the obvious examples of Stokes Theorem in R, R
2
, and R
3
.
J 16.4 Example. Let M = [a, b] R, and consider the 0-form = f , then, since
M is compact, is compactly supported 0-form. We have that d = f
0
(x)dx, and
Z
b
a
f
0
(x)dx =
Z
{a,b}
f = f (b) f (a),
which is the fundamental theorem of integration. Since on M any 1-form is exact
the theorem holds for 1-forms. I
J 16.5 Example. Let M = R
2
be a curve parametrized given by : [a, b] R
2
.
Consider a 0-form = f , then
Z

f
x
(x, y)dx + f
y
(x, y)dy =
Z
{(a),(b)}
f = f ((b)) f ((a)).
Now let M = R
2
, a closed subset of R
2
with a smooth boundary , and
consider a 1-form = f (x, y)dx +g(x, y)dy. Then, d =
_
g
x
f
y
)dx dy, and
Stokes Theorem yields
Z

g
x

f
y

dxdy =
Z

f dx +gdy,
also known as Greens Theorem in R
2
. I
J 16.6 Example. In the case of a curve M = R
3
we obtain the line-integral for
0-forms as before:
Z

f
x
(x, y, z)dx + f
y
(x, y, z)dy + f
z
(x, y, z) = f ((b)) f ((a)).
If we write the vector eld
F = f

x
+g

y
+h

z
,
then
grad f =f =
f
x

x
+
f
y

y
+
f
z

z
,
and the above expression can be rewritten as
Z

f ds = f |

.
Next, let M = S R
3
be an embedded or immersed hypersurface, and let =
f dx +gdy +hdz be a 1-form. Then,
d =

h
y

g
z

dy dz +

f
z

h
x

dz dx +

g
x

f
y

dx dy.
Write the vector elds
96
curl F =F =

h
y

g
z

x
+

f
z

h
x

y
+

g
x

f
y

z
.
Furthermore set
ds =
_
_
_
dx
dy
dz
_
_
_, dS =
_
_
_
dydz
dzdx
dxdy
_
_
_,
then from Stokes Theorem we can write the following surface and line integrals:
Z
S
F dS =
Z
S
F ds,
which is usually referred to as the classical Stokes Theorem in R
3
. The version in
Theorem 16.1 is the general Stokes Theorem. Finally let M = a closed subset
of R
3
with a smooth boundary , and consider a a 2-form = f dy dz +gdz
dx +hdx dy. Then
d =

f
x
+
g
y
+
h
z

dx dy dz.
Write
div F = F =
f
x
+
g
y
+
h
z
, dV = dxdydz,
then from Stokes Theorem we obtain
Z

FdV =
Z

F dS,
which is referred to as the Gauss Divergence Theorem. I
97
V. De Rham cohomology
17. Denition of De Rham cohomology
In the previous chapters we introduced and integrated m-forms over manifolds
M. We recall that k-form
k
(M) is closed if d = 0, and a k-form
k
(M)
is exact if there exists a (k 1)-form
k1
(M) such that = d. Since d
2
= 0,
exact forms are closed. We dene
Z
k
(M) =
_

k
(M) : d = 0
_
= Ker(d),
B
k
(M) =
_

k
(M) :
k1
(M) 3 = d
_
= Im(d),
and in particular
B
k
(M) Z
k
(M).
The sets Z
k
and B
k
are real vector spaces, with B
k
a vector subspace of Z
k
. This
leads to the following denition.
Denition 17.1. Let M be a smooth m-dimensional manifold then the de Rham
cohomology groups are dened as
(16) H
k
dR
(M) := Z
k
(M)/B
k
(M), k = 0, m,
where B
0
(M) := 0.
It is immediate from this denition that Z
0
(M) are smooth functions on M that
are constant on each connected component of M. Therefore, when M is con-
nected, then H
0
dR
(M)

= R. Since
k
(M) = {0}, for k > m = dim M, we have
that H
k
dR
(M) = 0 for all k > m. For k < 0, we set H
k
dR
(M) = 0.
J 17.2 Remark. The de Rham groups dened above are in fact real vector spaces,
and thus groups under addition in particular. The reason we refer to de Rham
cohomology groups instead of de Rham vector spaces is because (co)homology
theories produce abelian groups. I
An equivalence class [] H
k
dR
(M) is called a cohomology class, and two form
,
0
Z
k
(M) are cohomologous if [] = [
0
]. This means in particular that and

0
differ by an exact form, i.e.

0
=+d.
98
Now let us consider a smooth mapping f : N M, then we have that the pull-
back f

acts as follows: f

:
k
(M)
k
(N). From Theorem 15.5 it follows that
d f

= f

d and therefore f

descends to homomorphism in cohomology. This

can be seen as follows:
d f

= f

d = 0, and f

d = d( f

),
and therefore the closed forms Z
k
(M) get mapped to Z
k
(N), and the exact form
B
k
(M) get mapped to B
k
(N). Now dene
f

[] = [ f

],
which is well-dened by
f

0
= f

+ f

d = f

+d( f

)
which proves that [ f

0
] = [ f

], whenever [
0
] = []. Summarizing, f

maps
cohomology classes in H
k
dR
(M) to classes in H
k
dR
(N):
f

: H
k
dR
(M) H
k
dR
(N),
Theorem 17.3. Let f : N M, and g : M K, then
g

= ( f g)

: H
k
dR
(K) H
k
dR
(N),
Moreover, id

is the identity map on cohomology.

Proof: Since g

= ( f g)

the proof follows immediately.

As a direct consequence of this theorem we obtain the invariance of de Rham
cohomology under diffeomorphisms.
Theorem 17.4. If f : N M is a diffeomorphism, then H
k
dR
(M)

= H
k
dR
(N).
Proof: We have that id = f f
1
= f
1
f , and by the previous theorem
id

= f

( f
1
)

= ( f
1
)

,
and thus f

is an isomorphism.
18. Homotopy invariance of cohomology
We will prove now that the de Rham cohomology of a smooth manifold M is
even invariant under homeomorphisms. As a matter of fact we prove that the de
Rham cohomology is invariant under homotopies of manifolds.
99
Denition 18.1. Two smooth mappings f , g : N M are said to be homotopic if
there exists a continuous map H : N[0, 1] M such that
H(p, 0) = f (p)
H(p, 1) = g(p),
for all p N. Such a mapping is called a homotopy from/between f to/and g. If
in addition H is smooth then f and g are said to be smoothly homotopic, and H is
called a smooth homotopy.
Using the notion of smooth homotopies we will prove the following crucial
property of cohomology:
Theorem 18.2. Let f , g : N M be two smoothly homotopic maps. Then for k 0
it holds for f

, g

: H
k
dR
(M) H
k
dR
(N), that
f

= g

.
J 18.3 Remark. It can be proved in fact that the above results holds for two homo-
topic (smooth) maps f and g. This is achieved by constructing a smooth homotopy
from a homotopy between maps. I
Proof of Theorem 18.2: A map h :
k
(M)
k1
(N) is called a homotopy map
between f

and g

if
(17) dh() +h(d) = g

,
k
(M).
Now consider the embedding i
t
: N N I, and the trivial homotopy between
i
0
and i
1
(just the identity map). Let
k
(NI), and dene the mapping
h() =
Z
1
0
i
t
dt,
which is a map from
k
(NI)
k1
(N). Choose coordinates so that either
=
I
(x, t)dx
I
, or =
I
0 (x, t)dt d
x
I
0
.
In the rst case we have that i
t
= 0 and therefore dh() = 0. On the other hand
h(d) = h

I
t
dt dx
I
+

I
x
i
dx
i
dx
I

Z
1
0

I
t
dt

dx
I
=
_

I
(x, 1)
I
(x, 0)
_
dx
I
= i

1
i

0
,
which prove (17) for i

0
and i

1
, i.e.
dh() +h(d) = i

1
i

0
.
100
In the second case we have
h(d) = h

I
x
i
dx
i
dt dx
I
0

=
Z
1
0

I
dx
i
i
x
i
_
dx
i
dt dx
I
0
_
dt
=

Z
1
0

I
x
i
dt

dx
i
dx
I
0
.
On the other hand
dh() = d

Z
1
0

I
0 (x, t)dt

dx
I
0

=

x
i

Z
1
0

I
0 (x, t)dt

dx
i
dx
I
0
=

Z
1
0

I
x
i
dt

dx
i
dx
I
0
= h(d).
This gives the relation that
dh() +h(d) = 0,
and since i

1
= i

0
= 0 in this case, this then completes the argument in both
cases, and h as dened above is a homotopy map between i

0
and i

1
.
By assumption we have a smooth homotopy H : N [0, 1] M bewteen f and
g, with f = Hi
0
, and g = Hi
1
. Consider the composition

h = hH

. Using the
relations above we obtain

h(d) +d

h() = h(H

d) +dh(H

)
= h(d(H

)) +dh(H

)
= i

1
H

0
H

= (Hi
1
)

(Hi
0
)

= g

.
If we assume that is closed then
g

= dh(H

),
and thus
0 = [dh(H

)] = [g

] = g

[] f

[],
which proves the theorem.
101
J 18.4 Remark. Using the same ideas as for the Whitney embedding theorem one
can prove, using approximation by smooth maps, that Theorem 18.2 holds true for
continuous homotopies between smooth maps. I
Denition 18.5. Two manifolds N and M are said to be homotopy equivalent, if
there exist smooth maps f : N M, and g : M N such that
g f

= id
N
, f g

= id
M
(homotopic maps).
We write N M., The maps f and g are homotopy equivalences are each other
homotopy inverses . If the homotopies involved are smooth we say that N and M
smoothly homotopy equivalent.
J 18.6 Example. Let N = S
1
, the standard circle in R
2
, and M =R
2
\{(0, 0)}. We
have that N M by considering the maps
f = i : S
1
,R
2
\{(0, 0)}, g = id/| |.
Clearly, (g f )(p) = p, and ( f g)(p) = p/|p|, and the latter is homotopic to the
identity via H(p, t) =t p+(1t)p/|p|. I
Theorem 18.7. Let N and M be smoothly homotopically equivalent manifold, N
M, then
H
k
dR
(N)

= H
k
dR
(M),
and the homotopy equivalences f g between N and M, and M and N respectively
are isomorphisms.
As before this theorem remains valid for continuous homotopy equivalences of
manifolds.
102
VI. Exercises
Anumber of the exercises given here are taken fromthe Lecture notes by J. Bochnak.
Manifolds
Topological Manifolds
+ 1
Given the function g : R R
2
, g(t) = (cos(t), sin(t)). Show that f (R) is a mani-
fold.
+ 2
Given the set T
2
= {p = (p
1
, p
2
, p
3
) | 16(p
2
1
+ p
2
2
) = (p
2
1
+ p
2
2
+ p
2
3
+3)
2
} R
3
,
called the 2-torus.
(i ) Consider the product manifold S
1
S
1
={q = (q
1
, q
2
, q
3
, q
4
) | q
2
1
+q
2
2
=1, q
2
3
+
q
2
4
= 1}, and the mapping f : S
1
S
1
T
2
, given by
f (q) =
_
q
1
(2+q
3
), q
2
(2+q
3
), q
4
_
.
Show that f is onto and f
1
(p) =
_
p
1
r
,
p
2
r
, r 2, p
3
_
, where r =
|p|
2
+3
4
.
(ii ) Show that f is a homeomorphism between S
1
S
1
and T
2
.
+ 3
(i ) Find an atlas for T
2
using the mapping f in 2.
(ii ) Give a parametrization for T
2
.
4
Show that
A
4,4
= {(p
1
, p
2
) R
2
| p
4
1
+ p
4
2
= 1},
is a manifold and A
4,4

= S
1
(homeomorphic).
+ 5 (i ) Show that an open subset U M of a manifold M is again a manifold.
(ii ) Let N and M be manifolds. Show that their cartesian production NM is also
a manifold.
103
+ 6
Show that
(i ) {A M
2,2
(R) | det(A) = 1} is a manifold.
(ii ) Gl(n, R) = {A M
n,n
(R) | det(A) 6= 0} is a manifold.
(iii ) Determine the dimensions of the manifolds in (a) and (b).
7
Construct a simple counterexample of a set that is not a manifold.
8
Show that in Denition 1.1 an open set U
0
R
n
can be replaced by an open disc
D
n
R
n
.
9
Show that PR
n
is a Hausdorff space and is compact.
10
Dene the Grassmann manifold G
k
R
n
as the set of all k-dimensional linear sub-
spaces in R
n
. Show that G
k
R
n
is a manifold.
11
Consider X to be the parallel lines
_
R{0}
_

_
R{1}
_
. Dene the equivalence
relation (x, 0) (x, 1) for all x 6= 0. Show that M = X/ is a topological space
that satises (ii) and (iii) of Denition 1.1.
+ 12
Let M be an uncountable union of copies of R. Show that M is a topological space
that satises (i) and (ii) of Denition 1.1.
13
Let M = {(x, y) R
2
: xy = 0}. Show that M is a topological space that satises
(i) and (iii) of Denition 1.1.
Differentiable manifolds
+ 14
Show that cartesian products of differentiable manifolds are again differentaible
manifolds.
+ 15
Show that PR
n
is a smooth manifold.
16
Prove that PR is diffeomorphic to the standard unit circle in R
2
by constructing a
diffeomorphism.
+ 17
Which of the atlases for S
1
given in Example 2, Sect. 1, are compatible. If not
compatible, are they diffeomorphic?
+ 18
Show that the standard torus T
2
R
3
is diffeomorphic to S
1
S
1
, where S
1
R
2
is the standard circle in R
2
.
19
Prove that the taking the union of C

-atlases denes an equivalence relation (Hint:

use Figure 7 in Section 2).
104
+ 20
Show that diffeomorphic denes an equivalence relation.
+ 21
Prove Theorem 2.15.
Immersion, submersion and embeddings
+ 22
Show that the torus the map f dened in Exer. 2 of 1.1 yields a smooth embedding
of the torus T
2
in R
3
.
23
Let k, m, n be positive integers. Show that the set
A
k,m,n
:= {(x, y, z) R
3
: x
k
+y
m
+z
n
= 1},
is a smooth embedded submanifold in R
3
, and is diffeomorphic to S
2
when these
numbers are even.
24
Prove Lemma 3.18.
+ 25
Let f : R
n+1
\{0} R
k+1
\{0} be a smooth mapping such that f (x) =
d
f (x),
d N, for all R\{0}. This is called a homogeneous mapping of degree d.
Show that

f : PR
n
PR
k
, dened by

f ([x]) = [ f (x)], is a smooth mapping.

+ 26
Show that the open ball B
n
R
n
is diffeomorphic to R
n
.
+ 27
(Lee) Consider the mapping f (x, y, s) = (x
2
+y, x
2
+y
2
+s
2
+y) from R
3
to R
2
.
Show that q = (0, 1) is a regular value, and f
1
(q) is diffeomorphic to S
1
(stan-
dard).
28
Prove Theorem 3.24 (Hint: prove the steps indicated above Theorem 3.24).
29
Prove Theorem 3.26.
30
Let N, M be two smooth manifolds of the same dimension, and f : N M is a
smooth mapping. Show (using the Inverse Function Theorem) that if f is a bijec-
tion then it is a diffeomorphism.
31
Prove Theorem 3.28.
+ 32
Show that the map f : S
n
PR
n
dened by f (x
1
, , x
n+1
) = [(x
1
, , x
n+1
)] is
smooth and everywhere of rank n (see Boothby).
105
+ 33
(Lee) Let a R, and M
a
= {(x, y) R
2
: y
2
= x(x 1)(x a)}.
(i ) For which values of a is M
a
an embedded submanifold of R
2
?
(ii ) For which values of a can M
a
be given a topology and smooth structure so that
M
a
is an immersed submanifold?
+ 34
Consider the map f : R
2
R given by f (x
1
, x
2
) = x
3
1
+x
1
x
2
+x
3
2
. Which level sets
of f are smooth embedded submanifolds of R
2
.
35
Let M be a smooth m-dimensional manifold with boundary M (see Lee, pg. 25).
Show that M is an (m1)-dimensional manifold (no boundary), and that there
exists a unique differentiable structure on M such that i : M ,M is a (smooth)
embedding.
+ 36
Let f : N M be a smooth mapping. Let S = f
1
(q) N is a smooth embedded
submanifold of codimension m. Is q necessarily a regular value of f ?
+ 37
Use Theorem 3.27 to show that the 2-torus T
2
is a smooth embedded manifold in
R
4
.
38
Prove Theorem 3.37.
39
Show that an m-dimensional linear subspace in R
`
is an m-dimensional manifold.
40
Prove that the annulus A ={(p
1
, p
2
) R
2
: 1 p
2
1
+p
2
2
4} is a smooth manifold
two dimensional manifold in R
2
with boundary.
106
Tangent and cotangent spaces
Tangent spaces
+ 41
Let U R
k
, y R
m
, and f : U R
m
a smooth function. By Theorem 3.28 M =
f
1
(y) ={p R
k
: f (p) =y} is a smooth embedded manifold in R
k
if rk(J f )|
p
=
m for all p M. Show that
T
p
M

= {X
p
R
k
: X
p

p
i
f = 0, i = 1, , m} = ker J f |
p
,
i.e. the tangent space is the kernel of the Jacobian. Using this identication from
now on, prove that
M
p
= p+T
p
M R
k
is tangent to M at p.
+ 42
Given the set M = {(p
1
, p
2
, p
3
) R
3
: p
3
1
+ p
3
2
+ p
3
3
3p
1
p
2
p
3
= 1}.
(i ) Show that M is smooth embedded manifold in R
3
of dimension 2.
(ii ) Compute T
p
M, at p = (0, 0, 1).
+ 43
Given the set M = {(p
1
, p
2
, p
3
) R
3
: p
2
1
p
2
2
+ p
1
p
3
2p
2
p
3
= 0, 2p
1
p
2
+
p
3
= 3}.
(i ) Show that M is smooth embedded manifold in R
3
of dimension 1.
(ii ) Compute T
p
M, at p = (1, 1, 0).
44
Prove Lemma 4.5.
+ 45
Express the following planar vector elds in polar coordinates:
(i ) X = x

x
+y

y
;
(ii ) X =y

x
+x

y
;
(iii ) X = (x
2
+y
2
)

x
.
46
Find a vector eld on S
2
that vanishes at exactly one point.
47
Let S
2
R
2
be the standard unit sphere and f : S
2
S
2
a smooth map dened as
a -degree rotation around the polar axis. Compute f

: T
p
S
2
T
f (p)
S
2
and give a
matrix representation of f

in terms of a canonical basis.

107
Cotangent spaces
+ 48
(Lee) Let f : R
3
R be given by f (p
1
, p
2
, p
3
) = |p|
2
, and g : R
2
R
3
by
g(x
1
, x
2
) =

2x
1
|x|
2
+1
,
2x
2
|x|
2
+1
,
|x|
2
1
|x|
2
+1

.
Compute both g

d f , and d( f g) and compare the two answers.

+ 49
(Lee) Compute d f in coordinates, and determine the set points where d f vanishes.
(i ) M = {(p
1
, p
2
) R
2
: p
2
> 0}, and f (p) =
p
1
|p|
2
standard coordinates in R
2
.
(ii ) As the previous, but in polar coordinates.
(iii ) M = S
2
= {p R
3
: |p| = 1}, and f (p) = p
3
stereographic coordinates.
(iv) M =R
n
, and f (p) = |p|
2
standard coordinates in R
n
.
+ 50
Express in polar coordinates
(i ) = x
2
dx +(x +y)dy;
(ii ) = xdx +ydy +zdz.
51
Prove Lemma 5.3.
+ 52
(Lee) Let f : N M (smooth),
1
(N), and : [a, b] N is a smooth curve.
Show that
Z

=
Z
f
.
+ 53
Given the following 1-forms on R
3
:
=
4z
(x
2
+1)
2
dx +
2y
y
2
+1
dy +
2x
x
2
+1
dz,
=
4xz
(x
2
+1)
2
dx +
2y
y
2
+1
dy +
2
x
2
+1
dz.
(i ) Let (t) = (t, t, t), t [0, 1], and compute
R

and
R

.
(ii ) Let be a piecewise smooth curve going from (0, 0, 0) to (1, 0, 0) to (1, 1, 0) to
(1, 1, 1), and compute the above integrals.
(iii ) Which of the 1-forms and is exact.
(iv) Compare the answers.
108
Vector bundles
+ 54
Let MR
`
be a manifold in R
`
. Show that TM as dened in Section 6 is a smooth
embedded submanifold of R
2`
.
55 Let S
1
R
2
be the standard unit circle. Show that TS
1
is diffeomorphic to S
1
R.
56
Let MR
`
be a manifold in R
`
. Show that T

M as dened in Section 6 is a smooth

embedded submanifold of R
2`
.
57 Show that the open M obius band is a smooth rank-1 vector bundle over S
1
.
109
Tensors
Tensors and tensor products
+ 58
Describe the standard inner product on R
n
as a covariant 2-tensor.
+ 59
Construct the determinant on R
2
and R
3
as covariant tensors.
60
Given the vectors a =
_
_
_
1
3
0
_
_
_, and b =
_
2
4
_
. Compute a

b and b

a.
+ 61
Given nite dimensional vector spaces V and W, prove that V W and W V are
isomorphic.
62
Given nite dimensional vector spaces U, V and W, prove that (U V) W and
U (V W) are isomorphic.
+ 63
Show that V R 'V 'RV.
Symmetric and alternating tensors
+ 64
Show that for T T
s
(V) the tensor Sym T is symmetric.
65
Prove that a tensor T T
s
(V) is symmetric if and only if T = Sym T.
66
Show that the algebra

(V) =

M
k=0

k
(V),
is a commutative algebra.
+ 67
Prove that for any T T
s
(V) the tensor Alt T is alternating.
68
Show that a tensor T T
s
(V) is alternating if and only if T = Alt T.
69
Given the vectors a =
_
_
_
1
2
4
_
_
_, b =
_
_
_
0
1
3
_
_
_, and c =
_
_
_
1
1
2
_
_
_. Compute abc,
and compare this with det(a, b, c).
110
70
Given vectors a
1
, , a
n
show that a
1
a
n
= det(a
1
, , a
n
).
71
Prove Lemma 8.11.
Tensor bundles and tensor elds
72
Let MR
`
be an embedded m-dimensional manifold. Show that T
r
M is a smooth
manifold in R
`+`
r
.
73
Similarly, show that T
s
M is a smooth manifold in R
`+`
s
, and T
r
s
M is a smooth
manifold in R
`+`
r
+`
s
.
74
Prove that the tensor bundles introduced in Section 7 are smooth manifolds.
+ 75 One can consider symmetric tensors in T
p
M as dened in Section 8. Dene and
describe
r
(T
p
M) T
r
(T
p
M) and
r
M T
r
M.
+ 76
Describe a smooth covariant 2-tensor eld in
2
M T
2
M. How does this relate to
an (indenite) inner product on T
p
M?
77
Prove Lemma 9.2.
+ 78
Given the manifolds N = M =R
2
, and the smooth mapping
q = f (p) = (p
2
1
p
2
, p
1
+2p
3
2
),
acting from N to M. Consider the tensor spaces T
2
(T
p
N)

= T
2
(T
q
M)

= T
2
(R
2
)
and compute the matrix for the pullback f

.
+ 79
In the above problem consider the 2-tensor eld on M, given by
= dy
1
dy
2
+q
1
dy
2
dy
2
.
(i ) Show that F
2
(M).
(ii ) Compute the pulback f

and show that f

F
2
(N).
(iii ) Compute f

(X,Y), where X,Y F(N).

111
Differential forms
+ 80
Given the differential form = dx
1
d
x2
dx
2
dx
3
on R
3
. For each of the fol-
lowing vector elds X compute i
X
:
(i ) X =

x
1
+

x
2


x
3
;
(ii ) X = x
1

x
1
x
2
2

x
3
;
(iii ) X = x
1
x
2

x
1
sin(x
3
)

x
2
;
+ 81
Given the mapping
f (p) = (sin(p
1
+ p
2
), p
2
1
p
2
),
acting from R
2
to R
2
and the 2-form = p
2
1
dx
1
dx
2
. Compute the pullback form
f

.
+ 82
Prove Lemma 10.2.
+ 83
Derive Formula (10).
Orientations
+ 84
Show that S
n
is orientable and give an orientation.
85
Show that the standard n-torus T
n
is orientable and nd an orientation.
+ 86
Prove that the Klein bottle and the projective space PR
2
are non-orientable.
+ 87
Give an orientation for the projective space PR
3
.
88
Prove that the projective spaces PR
n
are orientable for n = 2k +1, k 0.
89
Show that the projective spaces PR
n
are non-orientable for n = 2k, k 1.
112
Integration on manifolds
Integrating m-forms on R
m
+ 90
Let U R
m
be open and let K U be compact. Show that there exists a domain
of integration D such that K D R
m
.
91
Show that Denition 12.2 is well-posed.
Partitions of unity
92
Show that the function f
1
dened in Lemma 13.3 is smooth.
+ 93
If U is open covering of M for which each set U
i
intersects only nitely many other
sets in U, prove that U is locally nite.
94
Give an example of uncountable open covering U of the interval [0, 1] R, and a
countable renement V of U.
Integration of m-forms on m-dimensional manifolds
+ 95
Let S
2
= B
3
R
3
oriented via the standard orientation of R
3
, and consider the
2-form
= xdy dz +ydz dx +zdx dy.
Given the parametrization
F(, ) =
_
_
_
x
y
z
_
_
_=
_
_
_
sin()cos()
sin()sin()
cos()
_
_
_,
for S
2
, compute
R
S
2 .
+ 96
Given the 2-form = xdy dz +zdy dx, show that
Z
S
2
= 0.
113
+ 97
Consider the circle S
1
R
2
parametrized by
F() =
_
x
y
_
=
_
r cos()
r sin()
_
.
Compute the integral over S
1
of the 1-form = xdy ydx.
98
If in the previous problem we consider the 1-form
=
x
_
x
2
+y
2
dy
y
_
x
2
+y
2
dx.
Show that
R
S
1 represents the induced Euclidian length of S
1
.
+ 99
Consider the embedded torus
T
2
= {(x
1
, x
2
, x
3
, x
4
) R
4
: x
2
1
+x
2
2
= 1, x
2
3
+x
2
4
= 1}.
Compute the integral over T
2
of the 2-form
= x
2
1
dx
1
dx
4
+x
2
dx
3
dx
1
.
+ 100
Consider the following 3-manifold M parametrized by g : [0, 1]
3
R
4
,
_
_
_
r
s
t
_
_
_7
_
_
_
_
_
r
s
t
(2r t)
2
_
_
_
_
_
.
Compute
Z
M
x
2
dx
2
dx
3
dx
4
+2x
1
x
3
dx
1
dx
2
dx
3
.
The exterior derivative
+ 101
Let (x, y, z) R
3
. Compute the exterior derivative d, where is given as:
(i ) = e
xyz
;
(ii ) = x
2
+zsin(y);
(iii ) = xdx +ydy;
(iv) = dx +xdy +(z
2
x)dz;
(v) = xydx dz +zdx dy;
(vi ) = dx dz.
114
+ 102
Which of the following forms on R
3
are closed?
(i ) = xdx dy dz;
(ii ) = zdy dx +xdy dz;
(iii ) = xdx +ydy;
(iv) = zdx dz.
+ 103
Verify which of the following forms on R
2
are exact, and if so write = d:
(i ) = xdy ydx;
(ii ) = xdy +ydx;
(iii ) = dx dy;
(iv) = (x
2
+y
3
)dx dy;
+ 104
Verify which of the following forms on R
3
are exact, and if so write = d:
(i ) = xdx +ydy +zdz;
(ii ) = x
2
dx dy +z
3
dx dz;
(iii ) = x
2
ydx dy dz.
105
Verify that on R
2
and R
3
all closed k-forms, k 1, are exact.
106
Find a 2-form on R
3
\{(0, 0, 0)} which is closed but not exact.
Stokes Theorem
107
Let R
3
be a parametrized 3-manifold,i.e. a solid, or 3-dimensional domain.
Show that the standard volume of is given by
Vol(M) =
1
3
Z

xdy dz ydx dz +zdx dy.

108
Let R
n
be an n-dimensional domain. Prove the analogue of the previous prob-
lem.
+ 109
Prove the integration by parts formula
Z
M
f d =
Z
M
f
Z
M
d f ,
where f is a smooth function and a k-form.
115
110
Compute the integral
Z
S
2
x
2
ydx dz +x
3
dy dz +(z 2x
2
)dx dy,
where S
2
R
3
is the standard 2-sphere.
111
Use the standard polar coordinates for the S
2
R
3
with radius r to compute
Z
S
2
xdy dz ydx dz +zdx dy
(x
2
+y
2
+z
2
)
3/2
,
and use the answer to compute the volume of the r-ball B
r
.
Extras
112
Use the examples in Section 16 to show that
curl grad f = 0, and div curl F = 0,
where f : R
3
R, and F : R
3
R
3
.
+ 113
Given the mapping f : R
2
R
3
, = dx dz, and
f (s, t) =
_
_
_
cos(s)sin(t)

s
2
+t
2
st
_
_
_.
Compute f

.
+ 114
Given the mapping f : R
3
R
3
, = xydx dy dz, and
f (s, t, u) =
_
_
_
scos(t)
ssin(t)
u
_
_
_.
Compute f

.
+ 115
Let M =R
3
\{(0, 0, 0)}, and dene the following 2-form on M:
=
1
(x
2
+y
2
+z
2
)
3/2
_
xdy dz +ydz dx +zdx dy
_
.
(i ) Show that is closed (d = 0) (Hint: compute
R
2
S
).
(ii ) Prove that is not exact on M!
On N =R
3
\{x = y = 0} M consider the 1-form:
=
z
(x
2
+y
2
+z
2
)
1/2
xdy ydx
x
2
+y
2
.
116
(iii ) Show that is exact as 2-form on N, and verify that d =.
+ 116
Let M R
4
be given by the parametrization
(s, t, u) 7
_
_
_
_
_
s
t +u
t
s u
_
_
_
_
_
, (s, t, u) [0, 1]
3
.
(i ) Compute
R
M
dx
1
dx
2
dx
4
.
(ii ) Compute
R
M
x
1
x
3
dx
1
dx
2
dx
3
+x
2
3
dx
2
dx
3
dx
4
.
+ 117
Let C = be the boundary of the triangle OAB in R
2
, where O = (0, 0), A =
(/2, 0), and B = (/2, 1). Orient the traingle by traversing the boundary counter
clock wise. Given the integral
Z
C
(y sin(x))dx +cos(y)dy.
(i ) Compute the integral directly.
(ii ) Compute the integral via Greens Theorem.
118
Compute the integral
R

F dS, where =[0, 1]

3
, and
F(x, y, z) =
_
_
_
4xz
y
2
yz
_
_
_
(Hint: use the Gauss Divergence Theorem).
117
De Rham cohomology
Denition of De Rham cohomology
119
Prove Theorem 17.3.
120
Let M =
S
j
M
j
be a (countable) disjoint union of smooth manifolds. Prove the
isomorphism
H
k
dR
(M)

j
H
k
dR
(M
j
).
121
Show that the mapping : H
k
dR
(M)H
`
dR
(M) H
k+`
dR
(M), called the cup-product,
and dened by
[] [] := [],
is well-dened.
122
Let M = S
1
, show that
H
0
dR
(S
1
)

=R, H
1
dR
(S
1
)

=R, H
k
dR
(S
1
) = 0,
for k 2.
123
Show that
H
0
dR
(R
2
\{(0, 0)})

=R, H
1
dR
(R
2
\{(0, 0)})

=R, H
k
dR
(R
2
\{(0, 0)}) = 0,
for k 2.
124
Find a generator for H
1
dR
(R
2
\{(0, 0)}).
125
Compute the de Rham cohomology of the n-torus M =T
n
.
126
Compute the de Rham cohomology of M = S
2
.
118
VII. Solutions
1
Note that (x, y) g(R) if and only if x
2
+y
2
= 1. So g(R) is just the circle S
1
in
the plane. The fact that this is a manifold is in the lecture notes.
2
We will show that f has an inverse by showing that the funcrion suggested in the
exercise is indeed the inverse of f . It follows that f is a bijection and therefore it
is onto. By continuity of f and its inverse, it follows that it is a homeomorphism.
We rst show that f maps S
1
S
1
into T
2
: We use the parametrisation of S
1
S
1
given by
(s, t) 7(coss, sins, cost, sint),
where s, t [0, 2). Suppose that q = (coss, sins, cost, sint) S
1
S
1
. Then
f (q) = (coss(2+cost), sins(2+cost), sint),
so f (q) T
2
if and only if LHS = RHS where
(LHS) = 16(cos
2
s(2+cost)
2
+sin
2
s(2+cost)
2
),
(RHS) = (cos
2
s(2+cost)
2
+sin
2
s(2+cost)
2
+sin
2
t +3)
2
.
We have
LHS = 16(2+cost)
2
RHS = ((2+cost)
2
+sin
2
t +3)
2
= (4+4cost +cos
2
t +sin
2
t +3)
2
= (4(2+cost)
2
= 16(2+cost).
This shows that f maps S
1
S
1
into T
2
. Now we let g : T
2
S
1
S
1
be the
suggested inverse; i.e.
g(p) =

p
1
r
,
p
2
r
, r 2, p
3

,
where r =
|p|
2
+3
4
. We will rst show that g maps T
2
into S
1
S
1
. So we let p T
2
and we will show that g(p) S
1
S
1
. But this is the case if and only if p
2
1
+p
2
2
=r
2
and (r 2)
2
+ p
2
3
= 1. The rst equality follows from the denition of r and the
119
fact that p T
2
. For the second equality we note that
(r 2)
2
+ p
2
3
= r
2
4r +4+ p
2
3
= p
2
1
+ p
2
2
4r +4+ p
2
3
= (p
2
1
+ p
2
2
+ p
2
3
+3) (|p|
2
+3) +1
= 1.
This shows that g maps T
2
into S
1
S
1
. We will now show that g is the inverse of
f ; let p T
2
and q S
1
S
1
. Then
f g(p) = f

p
1
r
,
p
2
r
, r 2, p
3

p
1
r
r,
p
2
r
r, p
3

= p
g f (q) = g(q
1
(2+q
3
), q
2
(2+q
3
), q
4
)
=

q
1
(2+q
3
)
r
0
,
q
2
(2+q
3
)
r
0
, r
0
2, q
4

where r
0
=
q
2
1
(2+q
3
)
2
+q
2
2
(2+q
3
)
2
+q
2
4
+3
4
=
1
/4
_
(2+q
3
)
2
+q
2
4
+3
_
=
1
/4
_
4+4q
3
+q
2
3
+q
2
4
_
=
1
/4
_
4(2+q
3
)
_
= 2+q
3
and therefore g f (q) = (q
1
, q
2
, q
3
, q
4
) = q.
This show that f is a bijection, so in particular it is onto. Since f and g are both
continuos, it follows that f is a homeomorphism.
3
(i ) Given co-ordinate charts (U, ) and (V, ) for S
1
, we dene a new co-ordinate
chart for T
2
as follows; W = f (U) f (V), : W (U) (V) is dened by
(p) = ((x), (y)) where (x, y) = f
1
(p) and p W. This shows how to make
an atlas for T
2
using an atlas for S
1
.
(ii ) A parametrisation for S
1
S
1
was already given in the solution of 2. Using the
map f , we obtain the following parametrisation for T
2
;
(s, t) 7
_
coss (2+cost), sins (2+cost), sint
_
,
where s, t [0, 2).
5 (i ) If (V, ) is a co-ordinate chart for M, then let W =U V. Let : W (W)
be dened by (x) = (x), i.e. is just the restriction of to W. Then clearly
is a homeomorphism from W to an open subset of (V). In particular, if (V) is
120
m-dimensional, then so is (W). This shows how to make an atlas for U, given an
atlas for M.
(ii ) Let (U, ) and (V, ) be charts for N and M respectively. The map : UV
(U) (V) is dened by (x, y) = ((x), (y)). This is clearly a homeomor-
phism. This shows how to make an atlas for N M, given atlases for N and M.
Note that if N is n-dimensional and M is m-dimensional, then N M is n +m-
dimensional.
6
(i ) We always let A be a matrix of the form
_
a b
c d
_
We let M = {A M
2,2
(R)| det A = 1} and U = {A M : a 6= 0}. Note that U is an
open subset of M. We now dene : U (R\{0}) RR by
(A) = (a, b, c).
The inverse of is given by

1
(x, y, z) =
_
x y
z
1+yz
x
_
.
Cleary, and its inverse are continuous, and therefore is a homeomorphism. So
(,U) is a co-ordinate chart for M. In a similar way we can dene charts on sets
in M where b 6= 0, c 6= 0 and d 6= 0. These four sets cover all of M and so we
obtain an atlas for M which shows that M is a manifold. Since the range of is
3-dimensional, it follows that M has dimension 3.
(ii ) Note that det : M
n,n
(R) R is a continuous function and that Gl(n, R) =
det
1
(R\ {0}). It follows that Gl(n, R) is an open subset of the space M
n,n
(R).
But M
n,n
(R) is homeomorphic to R
n
2
, which is a manifold, and therefore Gl(n, R)
is an open subset of a manifold. It follows from 5 that this space is a manifold. The
dimension of Gl(n, R) is n
2
.
12
The space M is not second countable; observe that every second countable space
is separable. To see that M is not separable, note that it contains an uncountable
collection of non-empty pairwise disjoint open subsets.
14
Let M and N be differentiable manifolds with maximal atlases A = {(U
i
,
i
)}
iI
and B= {(V
k
,
k
)}
kK
respectively. A typical co-ordinate chart for MN is given
by
i

k
: U
i
V
k
R
m+n
.
121
Now, for i, j I and k, l K, consider the transition map (
i

k
) (
j

l
)
1
.
This os just the product of the maps
i

1
j
and
j

1
k
. Since these maps are
diffeomorphisms, their product is also a diffeomorphisms.
This shows that the atlas for MN given by {(U
i
V
k
,
i

k
)
iI,kK
is a C

-
atlas. Taking a maximal extension of this atlas endows MN with a differentiable
structure.
15
As usual, : R
n+1
\{0} PR
n
is the quotient mapping, given by (x) = [x]. For
i = 1, . . . , n+1, we have V
i
= {x R
n+1
\{0} : x
i
6= 0} and U
i
=(V
i
). For [x] U
i
we dene

i
([x]) =

x
1
x
i
, . . . ,
x
i1
x
i
,
x
i+1
x
i
, . . . ,
x
n+1
x
i

,
and its inverse is given by

1
i
(z
1
, . . . , z
n
) = [(z
1
, . . . , z
i1
, 1, z
i
, . . . , z
n
)].
We will show that for i < j, the transition map
i

1
j
:
j
(U
i
U
j
)
i
(U
i
U
i
)
is a diffeomorphism. We compute
i

1
j
;

1
j
(z
1
, . . . , z
n
) =
i
([z
1
, . . . , z
j1
, 1, z
j
, . . . , z
n
])
=

z
1
z
i
, . . . ,
z
i1
z
i
,
z
i+1
z
i
, . . . ,
z
j1
z
i
,
1
z
i
,
z
j
z
i
, . . .
z
n
z
i

.
This function is clearly a diffeomorphisms. Note that

i
(U
i
U
j
) = {z R
n
: z
j1
6= 0},

j
(U
i
U
j
) = {z R
n
: z
i
6= 0}.
17
All the atlases are compatible. We will compute some of the transition maps. We
list three co-ordinate charts, one from each atlas;
(1) : U U
0
, where U = {p S
1
: p
2
> 0} and U
0
= (1, 1).
(p) = p
1
and
1
(x) = (x,
_
1x
2
).
(2) : V V
0
, where V
0
= {p S
1
: p 6= (0, 1)} and V
0
=R.
(p) =
2p
1
1p
2
and
1
(x) =

4x
x
2
+4
,
x
2
4
x
2
+4

.
(3) : W W
0
, where W = {p S
1
: p 6= (1, 0)} and W
0
= (0, 2).
(p) = arccos p
1
and
1
() = (cos, sin).
Now we compute some of the transition maps;
122
(1) The map
1
: (1, 1) \{0} (, 2) (2, ) is given by
x 7
2x
1

1x
2
.
(2) The map
1
: (, 2) (2, ) (1, 1) \{0} is given by
x 7
4x
x
2
+4
.
(3) The map
1
: (0, ) (1, 1) is given by
7cos.
All the functions mentioned above are C

-functions. If we check this for all tran-

sition maps, then we have shown that the atlases are compatible.
18
We have proved in 15 that the product of differentiable manifolds is again a dif-
ferentiable manifolds. The C

-atlas constructed there gives a C

-atlas for S
1
S
1
.
We have seen in 2 that S
1
S
1
is homeomorphic to the torus T
2
and the map
f : S
1
S
1
T
2
is a homeomorphism. In this exercise we are asked to show that f
is even a diffeomorphism. The differentiable structure of T
2
is the structure which
it inherits from R
3
.
Note that using the map f we can endowT
2
with a differentiable structure inherited
fromS
1
S
1
. So we have two differentiable structures on T
2
; the structure inherited
from R
3
and the structure inherited from S
1
S
1
. This exercise amounts to saying
that T
2
endowed with the structure inherited from S
1
S
1
is a smooth embedded
submanifold of R
3
.
Fixing appropriate charts for S
1
S
1
and T
2
one can show that f is a diffeomor-
phism. For example, consider U S
1
S
1
given by
{q S
1
S
1
| q
1
> 0 & q
3
> 0 }
with : U (1, 1) (1, 1) given by (q) = (q
2
, q
4
). For T
2
we x a co-
ordinate chart (V, ) such that V f (U) and is a projection onto the 2
nd
and 3
rd
co-ordinate. Computing f
1
gives;
f
1
(x, y) = (x(2+
_
1y
2
), y).
This map is a diffeomorphism. For other co-ordinate charts this is similar.
20
Let M, N and O be differentiable manifolds. To show that diffeomorphic is an
equivalence relation, we prove reexivity, symmetry and transitivity.
For reexivity: M is diffeomorphic to M since the transition maps are diffeomor-
phism. So the identity is a diffeomorphism from M onto M.
123
For symmetry: The inverse of a diffeomorphism is a diffeomorphism.
For transitivity: Suppose f : M N and g : N O are diffeomorphisms and let
p M. Fix charts (U, ), (V, ), (V
0
,
0
) and (W, ) such that p U, f (p)
V V
0
and g f (p) W. Since f and g are diffeomorphisms, we may asssume
that f
1
and g(
0
)
1
are diffeomorphims. Also note that since N is a
differentiable manifold, the map
0

1
is a diffeomorphism. Now note that
g f
1
= (g(
0
)
1
) (
0

1
) ( f
1
).
The map on the right hand side is a composition of diffeomorphisms, so it is again a
diffeomorphism. It follows that the map on the left hand side is a diffeomorphism.
Since p M was arbitrary, this show that the composition of diffeomorphisms
between manifolds is again a diffeomorphism; i.e. M and O are diffeomorphic. In
this proof we have implicitly used part (iii) of Theorem 2.14.
21
See Lee, pg 33&34.
22
Recall that f : S
1
S
1
R
3
. We x a co-ordinate chart (U, ) on S
1
S
1
where
U = {q = (q
1
, q
2
, q
3
, q
4
)|q
2
> 0 & q
4
> 0},
and : U V is dened by (q
1
, q
2
, q
3
, q
4
) = (q
1
, q
3
) where V = (1, 1)(1, 1).
The chart (, R
3
) is just the identity. We express f in local co-ordinates, i.e.

f = f
1
where

f (x, y) =
_
x(1+y),
_
1x
2
(2+y),
_
1y
2
_
.
We have that
J

f
x=(p)
=
_
_
_
1+y x
x(2+y)(1x
2
)
1/2
(1x
2
)
1/2
0 y(1y
2
)
1/2
_
_
_
It is not hard to verify that the matrix J

f
x=(p)
has rank 2 for all p U. For example,
if p =
1
(0, 0), then
J

f
(0,0)
=
_
_
_
1 0
0 1
0 0
_
_
_
This shows that f is of rank 2 at all p U. Of course, we can also prove this for
other similar charts. So it follows that rk( f ) = 2, so it is an immersion. Since f is
a homeomorphism onto its image, it follows that f is a smooth embedding.
124
25
We compute g in local co-ordinates. So x the usual charts (U
i
, ) and (U
j
,
j
)
for PR
n
and PR
k
respectively. For z R
n
, we let z
i
be the point in R
n+1
given by
(z
1
, . . . , z
i1
, 1, z
i
, . . . z
n
). If f
j
(z
i
) 6= 0, then
g(z) =
j
g
1
i
(z) = ([ f (z
i
)])
=

f
1
(z
i
)
f
j
(z
i
)
, . . . ,
f
j1
(z
i
)
f
j
(z
i
)
,
f
j+1
(z
i
)
f
j
(z
i
)
, . . . ,
f
n
(z
i
)
f
j
(z
i
)

.
By assumption, f (z
i
) 6= 0, so there is some j such that f
j
(z
i
) 6= 0. The fact that g is
smooth follows from the smoothness of all co-ordinate functions f
k
of f .
26
The mapping f : B
n
S
n
dened by
f (x) =
1
1|x|
x,
is a diffeomorphism. Note that one can chose charts for B
n
and R
n
such that the co-
ordinate maps are just the identity. So the expression

f of f in local co-ordinates
is equal to f .
27
We rst note that
J f
(x,y,s)
=
_
2x 1 0
2x 2y +1 2s
_
Now consider a point p = (x, y, s) f
1
(q). Then x
2
+y = 0 and y
2
+s
2
= 1. If
s 6= 0, then the last two columns of J f
p
are independent. If s = 0, then y {1, 1}
and since x
2
+y = 0, we have in fact that y = 1 and x
2
= 1. In this case the rst
two colums on J f
p
are independent. Sow e have shown that if p f
1
(q), then
rk(J f
p
) = 2 and it follows that q is a regular value.
To show that f
1
(q) is diffeomorphic to S
1
, consider the set
A = {(x, y, s) R
3
| x
4
+s
2
= 1 & y = 0}.
Then A is diffeomorphic to S
1
, see 23. Now consider the map g : A R
3
dened
by
f (x, y, s) = (x, x
2
, s).
Then g is a constant rank map with rk(g) = 1 and g is injective. So g is a smooth
embedding of A (and thus of S
1
) into R
3
. It follows that g[A] = f
1
(q) is diffeo-
morphic to S
1
.
32
Let U
i
= {x S
n
: x
i
> 0} and be the projection on all co-ordinates but the i
th
.
We have that

1
(z) = (z
1
, . . . , z
i1
,
_
1|z|
2
, z
i
, . . . , z
n
).
125
where |z| < 1. Next we let V
i
=[U
i
] and

i
([x]) =

x
1
x
i
, . . . ,
x
i1
x
i
,
x
i+1
x
i
, . . . ,
x
n+1
x
i

.
We calculate f in local co-ordinates, so

f = f
1
,

f (z) =
1
_
1|z|
2
(z
1
, . . . , z
i1
, z
i+1
, . . . , z
n
).
We note that

z
i
z
j
_
1|z|
2
=
_

_
1+2z
i
(1|z|
2
)
3/2
i = j
2z
i
(1|z|
2
)
3/2
i 6= j
so
J

f
z
=
_
1|z|
2
_
3/2
_
_
_
_
_
_
1+2z
1
2z
1
2z
1
2z
2
1+2z
2
2z
2
.
.
.
.
.
.
.
.
.
.
.
.
2z
n
2z
n
1+2z
n
_
_
_
_
_
_
For all z with |z| < 1, this matrix has rank n.
33
We let f
a
: R
2
R be given by f
a
(x, y) = y
2
x(x 1)(x a) and note that:
f
a
(x, y) = y
2
x
3
+(a+1)x
2
ax.
Of course, M
a
= f
1
(0). We also have that:
J

f
a
|
(x,y)
= (3x
2
+2(a+1)x a, 2y).
We compute the rank of this matrix for various values of a and (x, y) M
a
;
If x 6 {0, 1, a} : then y 6= 0 the rank is 1
x = 0 : 3x
2
+2(a+1)x a =a the rank is 1 iff a 6= 0
x = 1 : 3x
2
+2(a+1)x a = a1 the rank is 1 iff a 6= 1
x = a : 3x
2
+2(a+1)x a = a(a1) the rank is 1 iff a(a1) = 0
So we conclude that if a 6 {0, 1}, then M
a
is an embedded submanifold of R
2
.
For (ii ) : this is possible for all a R.
34
So the question really is; which values q R are regular values of f ? For this, we
compute the Jacobian;
J

f |
(x,y)
= (3x
2
+y, 3y
2
+x).
Let q R and (x, y) f
1
(q). The rank of J f |
(x,y)
is 0 if and only if:
3x
2
+y = 3y
2
+x = 0.
126
In this case we have
y = 3x
2
y
2
= x/3
It follows from these equations that y
3
= x
3
and therefore x = y. Furthermore:
0 = 3x
2
+y = 3(3y
2
)
2
+y =
= 27y
4
+y = y(27y
3
+1).
We conclude that (x, y) is either (0, 0) or (1/3, 1/3). So q {0, 1/27}.
We conclude that if q 6 {0, 1/27}, then q is a regular value of f and therefore
f
1
(q) is a smooth embedded submanifold of R
2
. The level sets corresponding to
q = 0 and q = 1/27 are not embedded submanifolds.
36
NO; consider the mapping f : R
3
R
2
given by f (x, y, z) = (x
2
+y
2
, z
2
). We let
q = (1, 0) and S = f
1
(q). Note that
S = {(x, y, z) R
3
|x
2
+y
2
= 1 & z = 0}.
So S is just a circle embedded in R
3
. The co-dimension of S is 2. The Jacobian of
f in p = (x, y, z) is given by
J f |
p
=
_
2x 2y 0
0 0 2z
_
So for example, at the point p = (1, 0, 0) S, this matrix is of rank 1 and not of
rank 2. So q is not a regular value of f .
37
Let N = R
4
\ {p|p
2
1
+ p
2
2
= p
2
3
+ p
2
4
= 0} and consider the mapping f : N R
2
given by f (p) = (p
2
1
+ p
2
2
, p
2
3
+ p
2
4
). The Jacobian of f at the point p is given by
J f |
p
=
_
2p
1
2p
2
0 0
0 0 2p
3
2p
4
_
On N, this matrix is always of rank 2, so f is a constant rank mapping with rk( f ) =
2. The 2-torus in R
4
is the level set f
1
(1, 1), so it follows from Theorem 3.27 that
this space is a smooth embedded submanifold of N. Since N is an open subset of
R
4
, it follows that N, and hence the 2-torus, is a smooth embedded submanifold of
R
4
.
41
Fix p f
1
(y) and as in the proof of Theorem 3.28, construct a map g : R
n

R
nm
R
m
where
g() = (L, f () y),
127
where L : R
n
R
nm
is invertible on ker J f |
p
R
n
. We have that
Jg|
p
= LJ f |
p
,
and this map is onto. As in the proof of Theorem 3.28, g is invertible on a neigh-
bourhood V of (L(p), 0) such that
g
1
:
_
R
nm
{0}
_
V f
1
(y).
This gives a local co-ordinate map with

1
: R
nm
V|
nm
f
1
(y) U,
where () = L for all f
1
(y) U. It follows that
1
= L
1
and
J
1
|
p
= JL
1
|
p
= L
1
.
So we have that JL
1
|
p
(R
nm
) = ker J f |
p
by construction, so it follows that
T
p
M = ker J f |
p
,
and this is what we wanted to show.
42
Let f : R
3
R be dened by f (p
1
, p
2
, p
3
) = p
3
1
+ p
3
2
+ p
3
3
3p
1
p
2
p
3
, so that
M = f
1
(1). The Jacobian of f at the point p is given by
J f |
p
=

3p
2
1
3p
2
p
3
, 3p
2
2
3p
1
p
3
, 3p
2
3
3p
1
p
2

So we have that rk(J f )|

p
= 0 if and only if:
p
2
1
= p
2
p
3
& p
2
2
= p
1
p
3
& p
2
3
= p
1
p
2
.
So it follows that if rk(J f )|
p
=0, then f (p) =0. We conclude that q =1 is a regular
value of f and therefore M is 2-dimensional smooth embedded submanifold of R
3
.
We now compute T
p
M at the point p = (0, 0, 1). We use 41, so T
p
M =ker J f |
p
. We
have:
J f |
p
=

0 0 3

,
so it follows that
T
p
M = {(x, y, z) R
3
: z = 0}.
So this is just the (x, y)-plane.
43
Let f : R
3
R
2
be dened by f (p
1
, p
2
, p
3
) = (p
2
1
p
2
2
+p
1
p
3
2p
2
p
3
, 2p
1
p
2
+
p
3
) so that M = f
1
(0, 3). The Jacobian of f in p is given by:
J f |
p
=
_
2p
1
+ p
3
2p
2
2p
3
p
1
2p
2
2 1 1
_
128
Note that if p f
1
(0, 3), then p
2
1
+ p
1
p
3
= p
2
2
+2p
2
p
3
and therefore
p
1
(p
1
+ p
3
) = p
2
(p
2
+2p
3
)
2p
1
+ p
3
= p
2
+3.
It is left to the reader to verify that in this case, the rank of the Jacobian of f at p is
2.
We now compute T
p
M at the point p = (1, 1, 0) M. In this case we have
J f |
p
=
_
2 2 3
2 1 1
_
So, using the fact that T
p
M = ker J f |
p
, we have
T
p
M = {(x, y, z) R
3
: 2x +2y +3z = 0 & 2x y +z = 0}
This is the line spanned by the vector (5, 4, 6).
45
We recall that

x
= (1, 0)

r
= (cos, sin)

y
= (0, 1)

= (r sin, r cos).
So if r 6= 0, then

x
= cos

r

1
r
sin

y
= sin

r
+
1
r
cos

.
Also recall that x = r cos and y = r sin. we obtain the following: (i )
x

x
+y

y
= r cos(cos

r

1
r
sin

)
+r sin(sin

+
1
r
cos

)
= r cos
2

r
+r sin
2

r
= r

r
129
(ii )
y

x
+x

y
= r sin(cos

r

1
r
sin

)
+r cos(sin

r
+
1
r
cos

)
= sin
2

+cos
2

(iii )
(x
2
+y
2
)

x
= (r
2
cos
2
+r
2
sin
2
)(cos

r

1
r
sin

)
= r
2
cos

r
r sin

48
We let = u
2
+v
2
and = (+1)
2
. We rst compute d( f g). Note that f g :
R
2
R is given by
f g(u, v) =
4u
2
+4v
2
+(u
2
+v
2
1)
2

=
4+
2
2+1

=

2
+2+1

=
(+1)
2

= 1.
So we have that d( f g) = 0. Next we compute g

d f . Note that d f = 2xdx +

2ydy +2zdz. We have the following;
g

d f = 2
2u
+1
d

2u
+1

+2
2v
+1
d

2v
+1

+2
1
+1
d

1
+1

130
We rst make the following computations:
d

2u
+1

=
2(+1) 4u
2

du+
4uv

dv
=
2(v
2
u
2
+1)

du+
4uv

dv
d

2v
+1

=
4uv

du+
2(
2
+1) 4v

dv
=
4uv

du+
2(u
2
v
2
+1)

dv
d

2
1
+1

=
2u(
2
+1) 2u(
2
1)

du+
2v(
2
+1) 2v(
2
1)

dv
=
4u

du+
4v

dv
Combining these results with the above expression for g

d f , one can verify that

g

d f = 0; i.e. we have that d( f g) = g

d f .
49
(i )
f (p) =
p
1
p
2
1
+ p
2
2

f (x, y) =
x
x
2
+y
2
.
d f =
f
x
dx +
f
y
dy
=
y
2
x
2
(x
2
+y
2
)
2
dx +
2xy
(x
2
+y
2
)
2
dy
(ii )
f (p) =
p
1
p
2
1
+ p
2
2

f (r, ) =
r cos
r
2
=
cos
r
d f =


f
r
dr +


f

d
=
cos
r
2
dr +
sin
r
d.
(iv)
f (p) = |p|
2

f (x
1
, . . . , x
n
) = x
2
1
+. . . +x
2
n
d f = 2x
i
dx
i
131
50
Recall that x = r cos and y = r sin and note that
dx = cosdr r sind
dy = sindr +r cosd
(i ) We obtain:
= r
2
cos
2
(cosdr r sind) +(r cos+r sin)(sindr +r cosd)
= (r
2
cos
3
+r cossin+r sin
2
)dr +(r
2
cos
2
+r
2
sincosr
3
cos
2
sin)d
(ii ) We obtain:
= r cos(cosdr r sind) +r sin(sindr +r cosd)
= rdr.
52
We use the fact that ( f )

, see Lemma 6.3. We have the following se-

quence of equalities:
Z

=
Z
[a,b]

=
Z
[a,b]
( f )

=
Z
f
.
53
(i )
Z

=
Z
[0,1]

=
Z
[0,1]
4t
(t
2
+1)
2
dt +
2t
t
2
+1
dt +
2t
t
2
+1
dt
=
Z
[0,1]
4t
(t
2
+1)
2
dt +
Z
[0,1]
4t
t
2
+1
dt
=
2
t
2
+1

1
0
+2ln(t
2
+1)

1
0
= 2ln21.
and:
Z

=
Z
[0,1]

=
Z
[0,1]
4t
2
(t
2
+1)
2
dt +
2t
t
2
+1
dt +
2
t
2
+1
dt
=
Z
[0,1]
2(t
2
+1) 4t
2
(t
2
+1)
2
dt +
Z
[0,1]
2t
t
2
+1
dt
=
2t
t
2
+1

1
0
+ln(t
2
+1)

1
0
= 1+ln2.
(ii ) We can split into three pieces;
1
,
2
and
3
all ranging from [0, 1] to R
3
where

1
(t) = (t, 0, 0),
2
(t) = (1, t, 0) and
3
(t) = (1, 1, t). We now compute the integrals
132
as follows:
Z

=
Z

1
+
Z

2
+
Z

=
Z
[0,1]
4 0
(t
2
+1)
2
dt +
Z
[0,1]
2t
t
2
+1
dt +
Z
[0,1]
2
1+1
dt
=
Z
[0,1]
2t
t
2
+1
+
Z
[0,1]
1dt = ln(t
2
+1)

1
0
+t

1
0
= ln2+1.
and:
Z

=
Z

1
+
Z

2
+
Z

3
=
Z
[0,1]
2t
t
2
+1
dt +
Z
[0,1]
1dt = ln2+1.
(iii ) and (iv) : Recall that a 1-form is exact if it is of the form dg for some smooth
function g. The Fundamental Theorem for Line Integrals states that the integral of
an exact 1-form over a path only depends on the end-points of . So if either
or is exact, then the answers in (i) and (ii) do not differ. So we can conclude
immediately that is not exact. So what about ? Well, we guess that it is exact,
but we have to nd a function g : R
3
R such that = dg. This function g is
given by
g(x, y, z) =
2z
x
2
+1
+ln(y
2
+1).
54
Fix p M and a homeomorphism : U V where U,V R
l
and
(U M) =V (R
m
{0}).
Suppose (x
1
, . . . , x
l
) is a co-ordinate representation for . Note that J
1
|
p
(R
l
) is
just R
l
. Now dene a map : U R
l
V R
l
as follows;

q, v
i

x
i

=
_
x
1
(q), . . . , x
l
(q), v
1
, . . . , v
l
_
,
where

x
i
|
q
= J
1
|
p
(e
i
). Now note that

1
(x, v
1
, . . . , v
l
) =

1
(x), v
i

x
i

1
(x)

.
So is a homeomorphism fromUR
l
onto V R
l
. Since is a slice chart for M,
we also have that
(
1
[U]) =
_
V

R
m
{0}
_

_
R
m
{0}
_
R
l
R
l
where : TM M is the natural projection map. So maps a neighbourhood of
(p, v) TM onto a 2m-slice in R
2l
. This shows how to make co-ordinate charts
for TM. It remains to verify that if is another such chart, then the transition map

1
is smooth. For this, see Lee pg81.