Math 52H: Multilinear algebra, dierential

forms and Stokes theorem

Yakov Eliashberg
January 2014
2
Contents
I Multilinear Algebra 7
1 Linear and multilinear functions 9
1.1 Dual space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Canonical isomorphism between (V

and V . . . . . . . . . . . . . . . . . . . . . . 11
1.3 The map /

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Multilinear functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Symmetric bilinear functions and quadratic forms . . . . . . . . . . . . . . . . . . . . 15
2 Tensor and exterior products 17
2.1 Tensor product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Spaces of multilinear functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Symmetric and skew-symmetric tensors . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Symmetrization and anti-symmetrization . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Exterior product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.6 Spaces of symmetric and skew-symmetric tensors . . . . . . . . . . . . . . . . . . . . 24
2.7 Operator /

on spaces of tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Orientation and Volume 29
3.1 Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Orthogonal transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Determinant and Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Volume and Gram matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3
4 Dualities 35
4.1 Duality between k-forms and (n k)-forms on a n-dimensional Euclidean space V . 35
4.2 Euclidean structure on the space of exterior forms . . . . . . . . . . . . . . . . . . . 41
4.3 Contraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5 Complex vector spaces 49
5.1 Complex numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2 Complex vector space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.3 Complex linear maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
II Calculus of dierential forms 57
6 Topological preliminaries 59
6.1 Elements of topology in a vector space . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.2 Everywhere and nowhere dense sets . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.3 Compactness and connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.4 Connected and path-connected components . . . . . . . . . . . . . . . . . . . . . . . 64
6.5 Continuous maps and functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7 Vector elds and dierential forms 69
7.1 Dierential and gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.2 Smooth functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.3 Gradient vector eld . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.4 Vector elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
7.5 Dierential forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.6 Coordinate description of dierential forms . . . . . . . . . . . . . . . . . . . . . . . 75
7.7 Smooth maps and their dierentials . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
7.8 Operator f

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
7.9 Coordinate description of the operator f

. . . . . . . . . . . . . . . . . . . . . . . . 79
7.10 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.11 Pfaan equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4
8 Exterior dierential 83
8.1 Coordinate denition of the exterior dierential . . . . . . . . . . . . . . . . . . . . . 83
8.2 Properties of the operator d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.3 Curvilinear coordinate systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.4 Geometric denition of the exterior dierential . . . . . . . . . . . . . . . . . . . . . 89
8.5 More about vector elds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
8.6 Case n = 3. Summary of isomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.7 Gradient, curl and divergence of a vector eld . . . . . . . . . . . . . . . . . . . . . . 94
8.8 Example: expressing vector analysis operations in spherical coordinates . . . . . . . 95
8.9 Complex-valued dierential k-forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
9 Integration of dierential forms and functions 101
9.1 Useful technical tools: partition of unity and cut-o functions . . . . . . . . . . . . . 101
9.2 One-dimensional Riemann integral for functions and dierential 1-forms . . . . . . . 104
9.3 Integration of dierential 1-forms along curves . . . . . . . . . . . . . . . . . . . . . . 107
9.4 Integrals of closed and exact dierential 1-forms . . . . . . . . . . . . . . . . . . . . . 112
9.5 Integration of functions over domains in high-dimensional spaces . . . . . . . . . . . 113
9.6 Fubinis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.7 Integration of n-forms over domains in n-dimensional space . . . . . . . . . . . . . . 128
9.8 Manifolds and submanifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
9.8.1 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
9.8.2 Gluing construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
9.8.3 Examples of manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
9.8.4 Submanifolds of an n-dimensional vector space . . . . . . . . . . . . . . . . . 146
9.8.5 Submanifolds with boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
9.9 Tangent spaces and dierential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
9.10 Vector bundles and their homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . 153
9.11 Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
9.12 Integration of dierential k-forms over k-dimensional submanifolds . . . . . . . . . . 154
5
III Stokes theorem and its applications 161
10 Stokes theorem 163
10.1 Statement of Stokes theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
10.2 Proof of Stokes theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
10.3 Integration of functions over submanifolds . . . . . . . . . . . . . . . . . . . . . . . . 169
10.4 Work and Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
10.5 Integral formulas of vector analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
10.6 Expressing div and curl in curvilinear coordinates . . . . . . . . . . . . . . . . . . . . 179
11 Applications of Stokes formula 183
11.1 Integration of closed and exact forms . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
11.2 Approximation of continuous functions by smooth ones . . . . . . . . . . . . . . . . . 184
11.3 Homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
11.4 Winding and linking numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
11.5 Properties of k-forms on k-dimensional manifolds . . . . . . . . . . . . . . . . . . . . 195
6
Part I
Multilinear Algebra
7
Chapter 1
Linear and multilinear functions
1.1 Dual space
Let V be a nite-dimensional real vector space. The set of all linear functions on V will be denoted
by V

.
Proposition 1.1. V

is a vector space of the same dimension as V .

Proof. One can add linear functions and multiply them by real numbers:
(l
1
+l
2
)(x) = l
1
(x) +l
2
(x)
(l)(x) = l(x) for l, l
1
, l
2
V

, x V, R
It is straightforward to check that all axioms of a vector space are satised for V

. Let us now
check that dimV = dimV

.
Choose a basis v
1
. . . v
n
of V . For any x V let
_
_
_
_
_
x
1
.
.
.
x
n
_
_
_
_
_
be its coordinates in the basis v
1
. . . v
n
.
Notice that each coordinate x
1
, . . . , x
n
can be viewed as a linear function on V . Indeed,
1) the coordinates of the sum of two vectors are equal to the sum of the corresponding coordinates;
9
2) when a vector is multiplied by a number, its coordinates are being multiplied by the same
number.
Thus x
1
, . . . , x
n
are vectors from the space V

. Let us show now that they form a basis of V

.
Indeed, any linear function l V

can be written in the form l(x) = a

1
x
1
+. . . +a
n
x
n
which means
that l is a linear combination of x
1
. . . x
n
with coecients a
1
, . . . , a
n
. Thus x
1
, . . . , x
n
generate V

.
On the other hand, if a
1
x+. . . +a
n
x
n
is the 0-function, then all the coecients must be equal to 0;
i.e. functions x
1
, . . . , x
n
are linearly independent. Hence x
1
, . . . , x
n
form a basis of V and therefore
dimV

= n = dimV.
The space V

is called dual to V and the basis x

1
, . . . , x
n
dual to v
1
. . . v
n
.
1
Exercise 1.2. Prove the converse: given any basis l
1
, . . . , l
n
of V

we can construct a dual basis

w
1
, . . . , w
n
of V so that the functions l
1
, . . . , l
n
serve as coordinate functions for this basis.
Recall that vector spaces of the same dimension are isomorphic. For instance, if we x bases in
both spaces, we can map vectors of the rst basis into the corresponding vectors of the second basis,
and extend this map by linearity to an isomorphism between the spaces. In particular, sending a
basis S = v
1
, . . . , v
n
of a space V into the dual basis x
1
, . . . , x
n
of the dual space V

we can
establish an isomorphism i
S
: V V

. However, this isomorphism is not canonical, i.e. it depends

on the choice of the basis v
1
, . . . , v
n
.
If V is a Euclidean space, i.e. a space with a scalar product x, y), then this allows us to dene
another isomorphism V V

, dierent from the one described above. This isomorphism associates

with a vector v V a linear function l
v
(x) = v, x). We will denote the corresponding map V V

by T. Thus we have T(v) = l

v
for any vector v V .
Exercise 1.3. Prove that T : V V

is an isomorphism. Show that T = i

S
for any orthonormal
basis S.
The isomorphismT is independent of a choice of an orthonormal basis, but is still not completely
canonical: it depends on a choice of a scalar product.
1
It is sometimes customary to denote dual bases in V and V

by the same letters but using lower indices for V

and upper indices for V

, e.g. v1, . . . , vn and v

1
, . . . , v
n
. However, in these notes we do not follow this convention.
10
Remark 1.4. The denition of the dual space V

also works in the innite-dimensional case.

Exercise 1.5. Show that both maps i
S
and T are injective in the innite case as well. However,
neither one is surjective if V is innite-dimensional.
1.2 Canonical isomorphism between (V

)

and V
The space (V

, dual to the dual space V , is canonically isomorphic in the nite-dimensional case

to V . The word canonically means that the isomorphism is god-given, i.e. it is independent of
any additional choices.
When we write f(x) we usually mean that the function f is xed but the argument x can vary.
However, we can also take the opposite point of view, that x is xed but f can vary. Hence, we can
view the point x as a function on the vector space of functions.
If x V and f V

then the above argument allows us to consider vectors of the space V as

linear functions on the dual space V

. Thus we can dene a map I : V V

by the formula
x I(x) (V

, where I(x)(l) = l(x) for any l V

.
Exercise 1.6. Prove that if V is nite-dimensional then I is an isomorphism. What can go wrong
in the innite-dimensional case?
1.3 The map /

Given a map / : V W one can dene a dual map /

: W

as follows. For any linear

function l W

we dene the function /

(l) V

by the formula /

(l)(x) = l(A(x)), x V . In
other words, /

(l) = l /.
2
2
In fact, the above formula makes sense in much more general situation. Given any map : X Y between two
sets X and Y the formula

(h) = h denes a map

: F(Y ) F(X) between the spaces of functions on Y and

X. Notice that this map goes in the direction opposite to the direction of the map .
11
Given bases B
v
= v
1
, . . . , v
n
and B
w
= w
1
, . . . , w
k
in the vector spaces V and W one can
associate with the map / a matrix A = M
BvBw
(/). Its columns are coordinates of the vectors
/(v
j
), j = 1, . . . , n, in the basis B
w
. Dual spaces V

and W

have dual bases X = x

1
, . . . x
n

and Y = y
1
, . . . , y
k
which consist of coordinate functions corresponding to the basis B
v
and B
w
.
Let us denote by A

the matrix of the dual map /

with respect to the bases Y and X, i. e.

A

= M
Y X
(/

).
Proposition 1.7. The matrices A and A

are transpose to each other, i.e. A

= A
T
.
Proof. By the denition of the matrix of a linear map we should take vectors of the basis Y =
y
1
, . . . , y
k
, apply to them the map /

, expand the images in the basis X = x

1
, . . . x
n
and write
the components of these vectors as columns of the matrix /

. Set y
i
= /

(y
i
), i = 1, . . . , k. For
any vector u =
n

j=i
u
j
v
j
V, we have y
i
(u) = y
i
(/(u)). The coordinates of the vector /(u) in the
basis w
1
, . . . , w
k
may be obtained by multiplying the matrix A by the column
_
_
_
_
_
u
1
.
.
.
u
n
_
_
_
_
_
. Hence,
y
i
(u) = y
i
(/(u)) =
n

j=1
a
ij
u
j
.
But we also have
n

j=1
a
ij
x
j
(u) =
n

j=1
a
ij
u
j
.
Hence, the linear function y
i
V

has an expansion
n

j=1
a
ij
x
j
in the basis X = x
1
, . . . x
n
of the
space V

. Hence the i-th column of the matrix A

equals
_
_
_
_
_
a
i1
.
.
.
a
in
_
_
_
_
_
, so that the whole matrix A

has the form

A

=
_
_
_
_
_
a
11
a
k1
. . .
a
1n
a
kn
_
_
_
_
_
= A
T
.

12
Exercise 1.8. Given a linear map / : V W with a matrix A, nd /

(y
i
).
Answer. The map /

sends the coordinate function y

i
on W to the function
n

j=1
a
ij
x
j
on V , i.e. to
its expression in coordinates x
i
.
Proposition 1.9. Consider linear maps
U
A
V
B
W .
Then (B /)

= /

.
Proof. For any linear function l W

we have
(B /)

(l)(x) = l(B(/(x)) = /

(B

(l))(x)
for any x U.
Exercise 1.10. Suppose that V is a Euclidean space and / is a linear map V V . Prove that
for any two vectors X, Y V we have
/(X), Y ) = X, T
1
/

T(Y )). (1.3.1)

Solution. By denition of the operator T we have
X, T
1
(Z)) = Z(X)
for any vector Z V

. Applying this to Z = /

T(Y ) we see that the right-hand send of (1.3.1) is

equal to to /

T(Y )(X). On the other hand, the left-hand side can be rewritten as T(Y )(/(X)).
But /

T(Y )(X) = T(Y )(/(X)).

Let us recall that if V is an Euclidean space, then operator B : V V is called adjoint to
/ : V V if for any two vectors X, Y V one has
/(X), Y ) = X, B(Y )).
The adjoint operator always exist and unique. It is denoted by /

. Clearly, (/

= /. In any
orthonormal basis the matrices of an operator and its adjoint are transpose to each other. An
13
operator / : V V is called self-adjoint if /

= /, or equivalently, if for any two vectors

X, Y V one has
/(X), Y ) = X, /(Y )).
The statement of Exercise 1.10 can be otherwise expressed by saying that the adjoint operator
/

is equal to T
1
/

T : V V . In particular, an operator / : V V is self-adjoint if and

only if /

T = T /.
Remark 1.11. As it follows from Proposition 1.7 and Exercise 1.10 the matrix of a self-adjoint
operator in any orthonormal basis is symmetric. This is not true in an arbitrrary basis.
1.4 Multilinear functions
A function l(X
1
, X
2
, . . . , X
k
) of k vector arguments X
1
, . . . , X
k
V is called k-linear (or multi-
linear) if it is linear with respect to each argument when all other arguments are xed. We say
bilinear instead of 2-linear. Multilinear functions are also called tensors. Sometimes, one may also
say a k-linear form, or simply k-form instead of a k-linear functions. However, we will reserve
the term k-form for a skew-symmetric tensors which will be dened in Section 2.3 below.
If one xes a basis v
1
. . . v
n
in the space V then with each bilinear function f(X, Y ) one can
associate a square n n matrix as follows. Set a
ij
= f(v
i
, v
j
). Then A = (a
ij
)
i,j=1,...,n
is called the
matrix of the function f in the basis v
1
, . . . , v
n
. For any 2 vectors
X =
n

1
x
i
v
i
, Y =
n

1
y
j
v
j
we have
f(X, Y ) = f
_
_
n

i=1
x
i
v
i
,
n

j=1
y
j
v
j
_
_
=
n

i,j=1
x
i
y
j
f(v
i
, v
j
) =
n

i,j=1
a
ij
x
i
y
j
= X
T
AY .
Exercise 1.12. How does the matrix of a bilinear function depend on the choice of a basis?
Answer. The matrices A and

A of the bilinear form f(x, y) in the bases v
1
, . . . , v
n
and v
1
, . . . , v
n
are related by the formula

A = C
T
AC, where C is the matrix of transition from the basis v
1
. . . v
n
14
to the basis v
1
. . . v
n
, i.e the matrix whose columns are the coordinates of the basis v
1
. . . v
n
in the
basis v
1
. . . v
n
.
Similarly, with a k-linear function f(X
1
, . . . , X
k
) on V and a basis v
1
, . . . , v
n
one can associate
a k-dimensional matrix
A = a
i
1
i
2
...i
k
; 1 i
1
, . . . , i
k
n,
where
a
i
1
i
2
...i
k
= f(v
i
1
, . . . , v
i
k
) .
If X
i
=
n

j=1
x
ij
v
j
, i = 1, . . . , k , then
f(X
1
, . . . , X
k
) =
n

i
1
,i
2
,...i
k
=1
a
i
1
i
2
...i
k
x
1i
1
x
2i
2
. . . x
ki
k
,
see Proposition 2.1 below.
1.5 Symmetric bilinear functions and quadratic forms
A function Q : V R on a vector space V is called quadratic if there exists a bilinear function
f(X, Y ) such that
Q(X) = f(X, X), X V. (1.5.1)
One also uses the term quadratic form. The bilinear function f(X, Y ) is not uniquely determined
by the equation (1.5.1). For instance, all the bilinear functions x
1
y
2
, x
2
y
1
and
1
2
(x
1
y
2
+ x
2
y
1
) on
R
2
dene the same quadratic form x
1
x
2
.
On the other hand, there is a 1-1 corerspondence between quadratic form and symmetric bilinear
functions. A bilinear function f(X, Y ) is called symmetric if f(X, Y ) = f(Y, X) for all X, Y V .
Lemma 1.13. Given a quadratic form Q : V R there exists a unique symmetric bilinear form
f(X, Y ) such that Q(X) = f(X, X), X V .
Proof. If Q(X) = f(X, X) for a symmetric f then
Q(X +Y ) = f(X +Y, X +Y ) = f(X, X) +f(X, Y ) +f(Y, X) +f(Y, Y )
= Q(X) + 2f(X, Y ) +Q(Y ),
15
and hence
f(X, Y ) =
1
2
_
Q(X +Y ) Q(X) Q(Y )
_
. (1.5.2)
I leave it as an exercise to check that the formula (1.5.2) always denes a symmetric bilinear
function.
Let S = v
1
, . . . , v
n
is a basis of V . The matrix A = (a
ij
) of a symmetric bilinear form f(X, Y )
in the basis S is called also the matrix of the corresponding quadratic form Q(X) = f(X, X). This
matrix is symmetric, and
Q(X) =

ij
a
ij
x
i
x
j
= a
11
x
2
1
+ +a
nn
x
2
n
+ 2

i<j
a
ij
x
i
x
j
.
Thus the matrix A is diagonal if and only if the quadratic form Q is the sum of squares (with
coecients). Let us recall that if one changes the basis S to a basis

S = v
1
, . . . , v
n
then the
matrix of a bilinear form f changes to

C = C
T
AC.
Exercise 1.14. (Sylvesters inertia law) Prove that there is always exists a basis

S = v
1
, . . . , v
n

in which a quadratic form Q is reduced to a sum of squares. The number of positive, negative and
zero coecients with the squares is independent of the choice of the basis.
Thus, in some coordinate system a quadratic form can always be written as

1
x
2
i
+
k+l

k+1
x
2
j
, k +l n.
The number k of negative squares is called the negative index or simply the index of the quadratic
form Q, the total number of k + l of non-zero squares is called the rank of the form. It coincides
with the rank of the matrix of the form in any basis. A bilinear (and quadratic) form is called
non-degenerate if its rank is maximal possible, i.e. equal to n. For a non-degenerate quadratic for
Q the dierence l k between the number of positive and negative squares is called the signature.
A quadratic form Q is called positive denite if Q(X) 0 and if Q(X) = 0 then X = 0.
Equivalently, one can say that a form is positive denite if it is non-degenerate and its negative
index is equal to 0.
16
Chapter 2
Tensor and exterior products
2.1 Tensor product
Given a k-linear function and a l-linear function , one can form a (k +l)-linear function, which
will be denoted by and called the tensor product of the functions and . By denition
(X
1
, . . . , X
k
, X
k+1
, . . . , X
k+l
) := (X
1
, . . . , X
k
) (X
k+1
, . . . , X
k+l
).
For instance, the tensor product two linear functions l
1
and l
2
is a bilinear function l
1
l
2
dened
by the formula
l
1
l
2
(U, V ) = l
1
(U)l
2
(V ).
Let v
1
. . . v
n
be a basis in V and x
1
, . . . , x
n
a dual basis in V

, i.e. x
1
, . . . , x
n
are coordinates of a
vector with respect to the basis v
1
, . . . , v
n
.
The tensor product x
i
x
j
is a bilinear function x
i
x
j
(Y, Z) = y
i
z
j
. Thus a bilinear function
f with a matrix A can be written as a linear combination of the functions x
i
x
j
as follows:
f =
n

i,j=1
a
ij
x
i
x
j
,
where a
ij
is the matrix of the form f in the basis v
1
. . . v
n
. Similarly any k-linear function f with
a k-dimensional matrix A = a
i
1
i
2
...i
k
can be written (see 2.1 below) as a linear combination of
functions
x
i
1
x
i
2
x
i
k
, 1 i
1
, i
2
, . . . , i
k
n.
17
Namely, we have
f =
n

i
1
,i
2
,...i
k
=1
a
i
1
i
2
...i
k
x
i
1
x
i
2
x
i
k
.
2.2 Spaces of multilinear functions
All k-linear functions, or k-tensors, on a given n-dimensional vector space V themselves form a
vector space, which will be denoted by V
k
. The space V
1
is, of course, just the dual space V

.
Proposition 2.1. Let v
1
, . . . v
n
be a basis of V , and x
1
, . . . , x
k
be the dual basis of V

formed
by coordinate functions with respect to the basis V . Then n
k
k-linear functions x
i
1
x
i
k
,
1 i
1
, . . . , i
k
n, form a basis of the space V
k
.
Proof. Take a k-linear function F from V
k
and evaluate it on vectors v
i
1
, . . . , v
i
k
:
F(v
i
1
, . . . , v
i
k
) = a
i
1
...i
k
.
We claim that we have
F =

1i
1
,...,i
k
n
a
i
1
...i
k
x
i
1
x
i
k
.
Indeed, the functions on the both sides of this equality being evaluated on any set of k basic vectors
v
i
1
, . . . , v
i
k
, give the same value a
i
1
...i
k
. The same argument shows that if

1i
1
,...,i
k
n
a
i
1
...i
k
x
i
1

x
i
k
= 0 , then all coecients a
i
1
...i
k
should be equal to 0. Hence the functions x
i
1
x
i
k
,
1 i
1
, . . . , i
k
n, are linearly independent, and therefore form a basis of the space V
k

Similar to the case of spaces of linear functions, a linear map / : V W induces a linear map
/

: W
k
V
k
, which sends a k-linear function F W
k
to a k-linear function /

(F) V
k
,
dened by the formula
/

(F)(X
1
, . . . , X
k
) = F(/(X
1
), . . . , /(X
k
)) for any vectors X
1
, . . . X
k
V .
Exercise 2.2. Suppose V is provided with a basis v
1
, . . . , v
n
and x
i
1
x
i
k
, 1 i
1
, . . . , i
k
n, is
the corresponding basis of the space V
k
. Suppose that the map / : V V has a matrix A = (a
ij
)
in the basis v
1
, . . . , v
n
. Find the matrix of the map /

: V
k
V
k
in the basis x
i
1
x
i
k
.
18
2.3 Symmetric and skew-symmetric tensors
A multilinear function (tensor) is called symmetric if it remains unchanged under the transposition
of any two of its arguments:
f(X
1
, . . . , X
i
, . . . , X
j
, . . . , X
k
) = f(X
1
, . . . , X
j
, . . . , X
i
, . . . , X
k
)
Equivalently, one can say that a k-tensor f is symmetric if
f(X
i
1
, . . . , X
i
k
) = f(X
1
, . . . , X
k
)
for any permutation i
1
, . . . , i
k
of indices 1, . . . , k.
Exercise 2.3. Show that a bilinear function f(X, Y ) is symmetric if and only if its matrix (in any
basis) is symmetric.
Notice that the tensor product of two symmetric tensors usually is not symmetric.
Example 2.4. Any linear function is (trivially) symmetric. However, the tensor product of two
functions l
1
l
2
is not a symmetric bilinear function unless l
1
= l
2
. On the other hand, the function
l
1
l
2
+l
2
l
1
, is symmetric.
A tensor is called skew-symmetric (or anti-symmetric) if it changes its sign when one transposes
any two of its arguments:
f(X
1
, . . . , X
i
, . . . , X
j
, . . . , X
k
) = f(X
1
, . . . , X
j
, . . . , X
i
, . . . , X
k
).
Equivalently, one can say that a k-tensor f is anti-symmetric if
f(X
i
1
, . . . , X
i
k
) = (1)
inv(i
1
...i
k
)
f(X
1
, . . . , X
k
)
for any permutation i
1
, . . . , i
k
of indices 1, . . . , k, where inv(i
1
. . . i
k
) is the number of inversions in
the permutation i
1
, . . . , i
k
. Recall that two indices i
k
, i
l
form an inversion if k < l but i
k
> i
l
.
The matrix A of a bilinear skew-symmetric function is skew-symmetric, i.e.
A
T
= A.
19
Example 2.5. The determinant det(X
1
, . . . , X
n
) (considered as a function of columns X
1
, . . . , X
n
of a matrix) is a skew-symmetric n-linear function.
Exercise 2.6. Prove that any n-linear skew-symmetric function on R
n
is proportional to the de-
terminant.
Linear functions are trivially anti-symmetric (as well as symmetric).
As in the symmetric case, the tensor product of two skew-symmetric functions is not skew-
symmetric. We will dene below in Section 2.5 a new product, called an exterior product of skew-
symmetric functions, which will again be a skew-symmetric function.
2.4 Symmetrization and anti-symmetrization
The following constructions allow us to create symmetric or anti-symmetric tensors from arbitrary
tensors. Let f(X
1
, . . . , X
k
) be a k-tensor. Set
f
sym
(X
1
, . . . , X
k
) :=

(i
1
...i
k
)
f(X
i
1
, . . . , X
i
k
)
and
f
asym
(X
1
, . . . , X
k
) :=

(i
1
...i
k
)
(1)
inv(i
1
,...,i
k
)
f(X
i
1
, . . . , X
i
k
)
where the sums are taken over all permutations i
1
, . . . , i
k
of indices 1, . . . , k. The tensors f
sym
and f
asym
are called, respectively, symmetrization and anti-symmetrization of the tensor f. It is
now easy to see that
Proposition 2.7. The function f
sym
is symmetric. The function f
asym
is skew-symmetric. If f
is symmetric then f
sym
= k!f and f
asym
= 0. Similarly, if f is anti-symmetric then f
asym
= k!f,
f
sym
= 0.
Exercise 2.8. Let x
1
, . . . , x
n
be coordinate function in R
n
. Find (x
1
x
2
. . . x
n
)
asym
.
Answer. The determinant.
20
2.5 Exterior product
For our purposes skew-symmetric functions will be more important. Thus we will concentrate on
studying operations on them.
Skew-symmetric k-linear functions are also called exterior k-forms. Let be an exterior k-form
and an exterior l-form. We dene an exterior (k + l)-form , the exterior product of and
, as
:=
1
k!l!
( )
asym
.
In other words,
(X
1
, . . . , X
k
, X
k+1
, . . . , X
k+l
) =
1
k!l!

i
1
,...i
k+l
(1)
inv(i
1
,...,i
k+l
)
(X
i
1
. . . , X
i
k
)(X
i
k+1
, . . . , X
i
k+l
),
where the sum is taken over all permutations of indices 1, . . . , k +l.
Note that because the anti-symmetrization of an anti-symmetric k-tensor amounts to its mul-
tiplication by k!, we can also write
(X
1
, . . . , X
k+l
) =

i
1
<...<i
k
,i
k+1
<...<i
k+l
(1)
inv(i
1
,...,i
k+l
)
(X
i
1
, . . . , X
i
k
)(X
i
k+1
, . . . , X
i
k+l
),
where the sum is taken over all permutations i
1
, . . . , i
k+l
of indices 1, . . . , k +l.
Exercise 2.9. The exterior product operation has the following properties:
For any exterior k-form and exterior l-form we have = (1)
kl
.
Exterior product is linear with respect to each factor:
(
1
+
2
) =
1
+
2

() = ( )
for k-forms ,
1
,
2
, l-form and a real number .
Exterior product is associative: ( ) = ( ).
21
First two properties are fairly obvious. To prove associativity one can check that both sides of
the equality ( ) = ( ) are equal to
1
k!l!m!
( )
asym
.
In particular, if , and are 1-forms, i.e. if k = l = m = 1 then
= ( )
asym
.
This formula can be generalized for computing the exterior product of any number of 1-forms:

1

k
= (
1

k
)
asym
. (2.5.1)
Example 2.10. x
1
x
2
= x
1
x
2
x
2
x
1
. For 2 vectors, U =
_
_
_
_
_
u
1
.
.
.
u
n
_
_
_
_
_
, V =
_
_
_
_
_
v
1
.
.
.
v
n
_
_
_
_
_
, we have
x
1
x
2
(U, V ) = u
1
v
2
u
2
v
1
=

u
1
v
1
u
2
v
2

.
For 3 vectors U, V, W we have
x
1
x
2
x
3
(U, V, W) =
= x
1
x
2
(U, V )x
3
(W) +x
1
x
2
(V, W)x
3
(U) +x
1
x
2
(W, U)x
3
(V ) =
(u
1
v
2
u
2
v
1
)w
3
+ (v
1
w
2
v
2
w
1
)u
3
+ (w
1
u
2
w
2
u
1
)v
3
=

u
1
v
1
w
1
u
2
v
2
w
2
u
3
v
3
w
3

.
The last equality is just the expansion formula of the determinant according to the last row.
Proposition 2.11. Any exterior 2-form f can be written as
f =

1i<jn
a
ij
x
i
x
j
22
Proof. We had seen above that any bilinear form can be written as f =

ij
a
i,j
x
i
x
j
. If f is
skew-symmetric then the matrix A = (a
ij
) is skew-symmetric, i.e. a
ii
= 0, a
ij
= a
ji
for i ,= j.
Thus, f =

1i<jn
a
ij
(x
i
x
j
x
j
x
i
) =

1i<jn
a
ij
x
i
x
j
.
Exercise 2.12. Prove that any exterior k-form f can be written as
f =

1i,<...<i
k
n
a
i
1
...i
k
x
i
1
x
i
2
. . . x
i
k
.
The following proposition can be proven by induction over k, similar to what has been done in
Example 2.10 for the case k = 3.
Proposition 2.13. For any k 1-forms l
1
, . . . , l
k
and k vectors X
1
, . . . , X
k
we have
l
1
l
k
(X
1
, . . . , X
k
) =

l
1
(X
1
) . . . l
1
(X
k
)
. . . . . . . . .
l
k
(X
1
) . . . l
k
(X
k
)

. (2.5.2)
Corollary 2.14. The 1-forms l
1
, . . . , l
k
are linearly dependent as vectors of V

if and only if
l
1
. . . l
k
= 0. In particular, l
1
. . . l
k
= 0 if k > n = dimV .
Proof. If l
1
, . . . , l
k
are dependent then for any vectors X
1
, . . . , X
k
V the rows of the determinant
in the equation (2.13) are linearly dependent. Therefore, this determinant is equal to 0, and hence
l
1
. . . l
k
= 0. In particular, when k > n then the forms l
1
, . . . , l
k
are dependent (because
dimV

= dimV = n).
On the other hand, if l
1
, . . . , l
k
are linearly independent, then the vectors l
1
, . . . , l
k
V

can be
completed to form a basis l
1
, . . . , l
k
, l
k+1
, . . . , l
n
of V

. According to Exercise 1.2 there exists a basis

w
1
, . . . , w
n
of V that is dual to the basis l
1
, . . . , l
n
of V

. In other words, l
1
, . . . , l
n
can be viewed as
coordinate functions with respect to the basis w
1
, . . . , w
n
. In particular, we have l
i
(w
j
) = 0 if i ,= j
and l
i
(w
i
) = 1 for all i, j = 1, . . . , n. Hence we have
l
1
l
k
(w
1
, . . . , w
k
) =

l
1
(w
1
) . . . l
1
(w
k
)
. . . . . . . . .
l
k
(w
1
) . . . l
k
(w
k
)

1 . . . 0
. . . 1 . . .
0 . . . 1

= 1 ,
23
i.e. l
1
l
k
,= 0.
Proposition 2.13 can be also deduced from formula (2.5.1).
Corollary 2.14 and Exercise 2.12 imply that there are no non-zero k-forms on an n-dimensional
space for k > n.
2.6 Spaces of symmetric and skew-symmetric tensors
As was mentioned above, k-tensors on a vector space V form a vector space under the operation of
addition of functions and multiplication by real numbers. We denoted this space by V
k
. Symmet-
ric and skew-symmetric tensors form subspaces of this space V
k
, which we denote, respectively,
by S
k
(V

) and
k
(V

). In particular, we have
V

= S
1
(V

) =
1
(V

)
.
Exercise 2.15. What is the dimension of the spaces S
k
(V

) and
k
(V

)?
Answer.
dim
k
(V

) =
_
_
n
k
_
_
=
n!
k!(n k)!
dimS
k
(V

) =
(n +k 1)!
k!(n 1)!
.
The basis of
k
(V

) is formed by exterior k-forms x

i
1
x
i
2
. . . x
i
k
, 1 i
1
< i
2
< . . . < i
k
n.
2.7 Operator /

on spaces of tensors
For any linear operator / : V W we introduced above in Section 1.3 the notion of a dual linear
operator /

: W

. Namely /

(l) = l / for any element l V

, which is just a linear

function on V . In this section we extend this construction to k-tensors for k 1, i.e. we will dene
a map /

: W
k
V
k
.
24
Given a k-tensor W
k
and k vectors X
1
, . . . , X
k
V we dene
/

()(X
1
, . . . , X
k
) = (/(X
1
), . . . , /(X
k
)) .
Note that if is symmetric, or anti-symmetric, so is /

(). Hence, the map /

also induces the

maps S
k
(W

) S
k
(V

) and
k
(W

)
k
(V

). We will keep the same notation /

for both of
these maps as well.
Proposition 2.16. Let / : V W be a linear map. Then
1. /

( ) = /

() /

() for any W
k
, W
l
;
2. /

(
asym
) = (/

())
asym
, /

(
sym
) = (/

())
sym
;
3. /

( ) = /

() /

() for any
k
(W

),
l
(W

).
If B : W U is another linear map then (B /)

= /

.
Proof.
1. Take any k +l vectors X
1
, . . . , X
k+l
V . Then by denition of the operator /

we have
/

( )(X
1
, . . . , X
k+l
) = (/(X
1
), . . . , /(X
n+k
) =
(/(X
1
), . . . , /(X
k
)(/(X
k+1
), . . . , /(X
n+k
)) =
/

()(X
1
, . . . , X
k
)/

()(X
k+1
, . . . , X
k+n
) =
/

() /

()(X
1
, . . . , X
k+l
).
2. Given k vectors X
1
, . . . , X
k
V we get
/

(
asym
)(X
1
, . . . , X
k
) =
asym
(/(X
1
), . . . , /(X
k
)) =

(i
1
...i
k
)
(1)
inv(i
1
,...,i
k
)
(/(X
i
1
), . . . , /(X
i
k
)) =

(i
1
...i
k
)
(1)
inv(i
1
,...,i
k
)
/

()(X
i
1
, . . . , X
i
k
) = (/

())
asym
(X
1
, . . . , X
k
),
where the sum is taken over all permutations i
1
, . . . , i
k
of indices 1, . . . , k. Similarly one proves that
/

(
sym
) = (/

())
sym
.
3. /

( ) =
1
k!l!
/

(( )
asym
)) =
1
k!l!
(/

( ))
asym
= /

() /

().
The last statement of Proposition 2.16 is straightforward and its proof is left to the reader.
25
Let us now discuss how to compute /

() in coordinates. Let us x bases v

1
, . . . , v
m
and
w
1
, . . . , w
n
in spaces V and W. Let x
1
, . . . , x
m
and y
1
, . . . , y
n
be coordinates and
A =
_
_
_
_
_
a
11
. . . a
1m
.
.
.
.
.
.
.
.
.
a
n1
. . . a
nm
_
_
_
_
_
be the matrix of a linear map / : V W in these bases. Note that the map / in these coordinates
is given by n linear coordinate functions:
y
1
= l
1
(x
1
, . . . , x
m
) = a
11
x
1
+a
12
x
2
+. . . +a
1m
x
m
y
2
= l
2
(x
1
, . . . , x
m
) = a
21
x
1
+a
22
x
2
+. . . +a
2m
x
m
. . .
y
n
= l
n
(x
1
, . . . , x
k
) = a
n1
x
1
+a
n2
x
2
+. . . +a
nm
x
n
We have already computed in Section 1.3 that /

(y
k
) = l
k
=
m

j=1
a
kj
x
j
, k = 1, . . . , n. Indeed, the
coecients of the function l
k
= /

(y
k
) form the k-th column of the transpose matrix A
T
. Hence,
using Proposition 2.16 we compute:
/

(y
j
1
y
j
k
) = l
j
1
l
j
k
and
/

(y
j
1
y
j
k
) = l
j
1
l
j
k
.
Consider now the case when V = W, n = m, and we use the same basis v
1
, . . . , v
n
in the source
and target spaces.
Proposition 2.17.
/

(x
1
x
n
) = det Ax
1
x
n
.
Note that the determinant det A is independent of the choice of the basis. Indeed, the matrix
of a linear map changes to a similar matrix C
1
AC in a dierent basis, and det C
1
AC = det A.
Hence, we can write det / instead of det A, i.e. attribute the determinant to the linear operator /
rather than to its matrix A.
26
Proof. We have
/

(x
1
x
n
) = l
1
l
n
=
n

i
1
=1
a
1i
1
x
i
1

n

in=1
a
nin
x
in
=
n

i
1
,...,in=1
a
1i
1
. . . a
nin
x
i
1
x
in
.
Note that the in the latter sum all terms with repeating indices vanish, and hence we can replace
this sum by a sum over all permutations of indices 1, . . . , n. Thus, we can continue
/

(x
1
x
n
) =

i
1
,...,in
a
1i
1
. . . a
nin
x
i
1
x
in
=
_
_

i
1
,...,in
(1)
inv(i
1
,...,in)
a
1i
1
. . . a
nin
_
_
x
1
x
n
= det Ax
1
x
n
.
Exercise 2.18. Apply the equality
/

(x
1
x
k
x
k+1
x
n
) = /

(x
1
x
k
) /

(x
k+1
x
n
)
for a map / : R
n
R
n
to deduce the formula for expansion of a determinant according to its rst
k rows:
det A =

i
1
<<i
k
, j
1
<<j
nk
; im=j
l
(1)
inv(i
1
,...,j
nk
)

a
1,i
1
. . . a
1,i
k
.
.
.
.
.
.
.
.
.
a
k,i
1
. . . a
k,i
k

a
k+1,j
1
. . . a
k+1,j
nk
.
.
.
.
.
.
.
.
.
a
n,j
1
. . . a
n,j
nk

.
27
28
Chapter 3
Orientation and Volume
3.1 Orientation
We say that two bases v
1
, . . . , v
k
and w
1
, . . . , w
k
of a vector space V dene the same orientation of
V if the matrix of transition from one of these bases to the other has a positive determinant. Clearly,
if we have 3 bases, and the rst and the second dene the same orientation, and the second and the
third dene the same orientation then the rst and the third also dene the same orientation. Thus,
one can subdivide the set of all bases of V into the two classes. All bases in each of these classes
dene the same orientation; two bases chosen from dierent classes dene opposite orientation of
the space. To choose an orientation of the space simply means to choose one of these two classes of
bases.
There is no way to say which orientation is positive or which is negativeit is a question
of convention. For instance, the so-called counter-clockwise orientation of the plane depends from
which side we look at the plane. The positive orientation of our physical 3-space is a physical, not
mathematical, notion.
Suppose we are given two oriented spaces V, W of the same dimension. An invertible linear map
(= isomorphism) / : V W is called orientation preserving if it maps a basis which denes the
given orientation of V to a basis which denes the given orientation of W.
Any non-zero exterior n-form on V induces an orientation of the space V . Indeed, the preferred
set of bases is characterized by the property (v
1
, . . . , v
n
) > 0.
29
3.2 Orthogonal transformations
Let V be a Euclidean vector space. Recall that a linear operator | : V V is called orthogonal if
it preserves the scalar product, i.e. if
|(X), |(Y )) = X, Y ), (3.2.1)
for any vectors X, Y V . Recall that we have
|(X), |(Y )) = X, |

(|(Y ))),
where |

: V V is the adjoint operator to |, see Section 1.3 above.

Hence, the orthogonality of an operator | is equivalent to the identity |

| = Id, or |

= |
1
.
Here we denoted by Id the identity operator, i.e. Id(X) = X for any X V .
Let us recall, see Exercise 1.10, that the adjoint operator U

is related to the dual operator

U

: V

by the formula
U

= T
1
U

T.
Hence, for an orthogonal operator |, we have T
1
U

T = |
1
, i.e.
|

T = T |
1
. (3.2.2)
Let v
1
, . . . , v
n
be an orthonormal basis in V and U be the matrix of | in this basis. The matrix
of the adjoint operator in an orthonormal basis is the transpose of the matrix of this operator.
Hence, the equation |

| = Id translates into the equation U

T
U = E, or equivalently UU
T
= E,
or U
1
= U
T
for its matrix. Matrices, which satisfy this equation are called orthogonal. If we write
U =
_
_
_
_
_
u
11
. . . u
1n
. . . . . . . . .
u
n1
. . . u
nn
_
_
_
_
_
,
then the equation U
T
U = E can be rewritten as

i
u
ki
u
ji
=
_

_
1, if k = j,
0, if k ,= j,
.
30
Similarly, the equation UU
T
= E can be rewritten as

i
u
ik
u
ij
=
_

_
1, if k = j,
0, if k ,= j,
.
The above identities mean that columns (and rows) of an orthogonal matrix U form an or-
thonormal basis of R
n
with respect to the dot-product.
In particular, we have
1 = det(U
T
U) = det(U
T
) det U = (det U)
2
,
and hence det U = 1. In other words, the determinant of any orthogonal matrix is equal 1. We
can also say that the determinant of an orthogonal operator is equal to 1 because the determinant
of the matrix of an operator is independent of the choice of a basis. Orthogonal transformations
with det = 1 preserve the orientation of the space, while those with det = 1 reverse it.
Composition of two orthogonal transformations, or the inverse of an orthogonal transformation
is again an orthogonal transformation. The set of all orthogonal transformations of an n-dimensional
Euclidean space is denoted by O(n). Orientation preserving orthogonal transformations sometimes
called special, and the set of special orthogonal transformations is denoted by SO(n). For instance
O(1) consists of two elements and SO(1) of one: O(1) = 1, 1, SO(1) = 1. SO(2) consists of
rotations of the plane, while O(2) consists of rotations and reections with respect to lines.
3.3 Determinant and Volume
We begin by recalling some facts from Linear Algebra. Let V be an n-dimensional Euclidean space
with an inner product , ). Given a linear subspace L V and a point x V , the projection
proj
L
(x) is a vector y L which is uniquely characterized by the property x y L, i.e.
x y, z) = 0 for any z L. The length [[x proj
L
(x)[[ is called the distance from x to L; we
denote it by dist(x, L).
Let U
1
, . . . , U
k
V be linearly independent vectors. The k-dimensional parallelepiped spanned
by vectors U
1
, . . . , U
k
is, by denition, the set
P(U
1
, . . . , U
k
) =
_
k

j
U
j
; 0
1
, . . . ,
k
1
_
Span(U
1
, . . . , U
k
).
31
Given a k-dimensional parallelepiped P = P(U
1
, . . . , U
k
) we will dene its k-dimensional volume
by the formula
VolP = [[U
1
[[dist(U
2
, Span(U
1
))dist(U
3
, Span(U
1
, U
2
)) . . . dist(U
k
, Span(U
1
, . . . , U
k1
)). (3.3.1)
Of course we can write dist(U
1
, 0) instead of [[U
1
[[. This denition agrees with the denition of the
area of a parallelogram, or the volume of a 3-dimensional parallelepiped in the elementary geometry.
Proposition 3.1. Let v
1
, . . . , v
n
be an orthonormal basis in V . Given n vectors U
1
, . . . , U
n
let us
denote by U the matrix whose columns are coordinates of these vectors in the basis v
1
, . . . , v
n
:
U :=
_
_
_
_
_
u
11
. . . u
1n
.
.
.
u
n1
. . . u
nn
_
_
_
_
_
Then
Vol P(U
1
, . . . , U
n
) = [ det U[.
Proof. If the vectors U
1
, . . . , U
n
are linearly dependent then Vol P(U
1
, . . . , U
n
) = det U = 0.
Suppose now that the vectors U
1
, . . . , U
n
are linearly independent, i.e. form a basis. Consider rst
the case where this basis is orthonormal. Then the matrix U is orthogonal. i.e. UU
T
= E, and
hence det U = 1. But in this case Vol P(U
1
, . . . , U
n
) = 1, and hence Vol P(U
1
, . . . , U
n
) = [ det U[.
Now let the basis U
1
, . . . , U
n
be arbitrary. Let us apply to it the Gram-Schmidt orthonormaliza-
tion process. Recall that this process consists of the following steps. First, we normalize the vector
U
1
, then subtract from U
2
its projection to Span(U
1
), Next, we normalize the new vector U
2
, then
subtract from U
3
its projection to Span(U
1
, U
2
), and so on. At the end of this process we obtain
an orthonormal basis. It remains to notice that each of these steps aected Vol P(U
1
, . . . , U
n
) and
[ det U[ in a similar way. Indeed, when we multiplied the vectors by a positive number, both the
volume and the determinant were multiplied by the same number. When we subtracted from a vec-
tor U
k
its projection to Span(U
1
, . . . , U
k1
), this aected neither the volume nor the determinant.

Corollary 3.2. 1. Let x

1
, . . . , x
n
be a Cartesian coordinate system.
1
Then
Vol P(U
1
, . . . , U
n
) = [x
1
. . . x
n
(U
1
, . . . , U
n
)[.
1
i.e. a coordinate system with respect to an orthonormal basis
32
2. Let / : V V be a linear map. Then
Vol P(/(U
1
), . . . , /(U
n
)) = [ det /[Vol P(U
1
, . . . , U
n
).
Proof.
1. According to 2.13, x
1
. . . x
n
(U
1
, . . . , U
n
) = det U.
2. x
1
. . . x
n
(/(U
1
), . . . , /(U
n
)) = /

(x
1
x
n
)(U
1
, . . . , U
n
) = det /x
1
. . . x
n
(U
1
, . . . , U
n
).
In view of Proposition 3.1 and the rst part of Corollary 3.2 the value
x
1
. . . x
n
(U
1
, . . . , U
n
) = det U
is called sometimes the signed volume of the parallelepiped P(U
1
, . . . , U
n
). It is positive when the
basis U
1
, . . . , U
n
denes the given orientation of the space V , and it is negative otherwise.
Note that x
1
. . . x
k
(U
1
, . . . , U
k
) for 0 k n is the signed k-dimensional volume of the
orthogonal projection of the parallelepiped P(U
1
, . . . , U
k
) to the coordinate subspace x
k+1
=
= x
n
= 0.
For instance, let be the 2-form x
1
x
2
+x
3
x
4
on R
4
. Then for any two vectors U
1
, U
2
R
4
the value (U
1
, U
2
) is the sum of signed areas of projections of the parallelogram P(U
1
, U
2
) to the
coordinate planes spanned by the two rst and two last basic vectors.
3.4 Volume and Gram matrix
In this section we will compute the Vol
k
P(v
1
, . . . , v
k
) in the case when the number k of vectors is
less than the dimension n of the space.
Let V be an Euclidean space. Given vectors v
1
, . . . , v
k
V we can form a k k-matrix
G(v
1
, . . . , v
k
) =
_
_
_
_
_
v
1
, v
1
) . . . v
1
, v
k
)
. . . . . . . . .
v
k
, v
1
) . . . v
k
, v
k
)
_
_
_
_
_
, (3.4.1)
which is called the Gram matrix of vectors v
1
, . . . , v
k
.
Suppose we are given Cartesian coordinate system in V and let us form a matrix C whose
columns are coordinates of vectors v
1
, . . . , v
k
. Thus the matrix C has n rows and k columns. Then
G(v
1
, . . . , v
k
) = C
T
C,
33
because in Cartesian coordinates the scalar product looks like the dot-product.
We also point out that if k = n and vectors v
1
, . . . , v
n
form a basis of V , then G(v
1
, . . . , v
k
) is
just the matrix of the bilinear function X, Y ) in the basis v
1
, . . . , v
n
. It is important to point out
that while the matrix C depends on the choice of the basis, the matrix G does not.
Proposition 3.3. Given any k vectors v
1
, . . . , v
k
in an Euclidean space V the volume Vol
k
P(v
1
, . . . , v
k
)
can be computed by the formula
Vol
k
P(v
1
, . . . , v
k
)
2
= det G(v
1
, . . . , v
k
) = det C
T
C, (3.4.2)
where G(v
1
, . . . , v
k
) is the Gram matrix and C is the matrix whose columns are coordinates of
vectors v
1
, . . . , v
k
in some orthonormal basis.
Proof. Suppose rst that k = n. Then according to Proposition 3.1 we have Vol
k
P(v
1
, . . . , v
k
) =
[ det C[. But det C
T
C = det C
2
, and the claim follows.
Let us denote vectors of our orthonormal basis by w
1
, . . . , w
n
. Consider now the case when
Span(v
1
, . . . , v
k
) Span(w
1
, . . . , w
k
). (3.4.3)
In this case the elements in the j-th row of the matrix C are zero if j > k. Hence, if we denote
by

C the square k k matrix formed by the rst k rows of the matrix C, then C
T
C =

C
T

C and
thus det C
T
C = det

C
T

C. But det

C
T

C = Vol
k
P(v
1
, . . . , v
k
) in view of our above argument in the
equi-dimensional case applied to the subspace Span(w
1
, . . . , w
k
) V , and hence
Vol
2
k
P(v
1
, . . . , v
k
) = det C
T
C = det G(v
1
, . . . , v
k
).
But neither Vol
k
P(v
1
, . . . , v
k
), nor the Gram matrix G(v
1
, . . . , v
k
) depends on the choice of an
orthonormal basis. On the other hand, using Gram-Schmidt process one can always nd an or-
thonormal basis which satises condition (3.4.3).
Remark 3.4. Note that det G(v
1
, . . . , v
k
) 0 and det G(v
1
, . . . , v
k
) = 0 if an and only if the
vectors v
1
, . . . , v
k
are linearly dependent.
34
Chapter 4
Dualities
4.1 Duality between k-forms and (nk)-forms on a n-dimensional
Euclidean space V
Let V be an n-dimensional vector space. As we have seen above, the space
k
(V

) of k-forms, and
the space
nk
(V

) of (n k)-forms have the same dimension

n!
k!(nk)!
; these spaces are therefore
isomorphic. Suppose that V is an oriented Euclidean space, i.e. it is supplied with an orientation
and an inner product , ). It turns out that in this case there is a canonical way to establish this
isomorphism which will be denoted by
:
k
(V

)
nk
(V

) .
Denition 4.1. Let be a k-form. Then given any vectors U
1
, . . . , U
nk
, the value (U
1
, . . . , U
nk
)
can be computed as follows. If U
1
, . . . , U
nk
are linearly dependent then (U
1
, . . . , U
nk
) = 0. Oth-
erwise, let S

denote the orthogonal complement to the space S = Span(U

1
, . . . , U
nk
). Choose a
basis Z
1
, . . . , Z
k
of S

such that
Vol
k
(Z
1
, . . . , Z
k
) = Vol
nk
(U
1
, . . . , U
nk
)
and the basis Z
1
, . . . , Z
k
, U
1
, . . . , U
nk
denes the given orientation of the space V . Then
(U
1
, . . . , U
nk
) = (Z
1
, . . . , Z
k
). (4.1.1)
35
Let us rst show that
Lemma 4.2. is a (n k)-form, i.e. is skew-symmetric and multilinear.
Proof. To verify that is skew-symmetric we note that for any 1 i < j n q the bases
Z
1
, Z
2
, . . . , Z
k
, U
1
, . . . , U
i
, . . . , U
j
, . . . , U
nk
and
Z
1
, Z
2
, . . . , Z
k
, U
1
, . . . , U
j
, . . . , U
i
, . . . , U
nk
dene the same orientation of the space V , and hence
(U
1
, . . . , U
j
, . . . , U
i
, . . . , U
nk
) = (Z
1
, Z
2
, . . . , Z
k
)
=(Z
1
, Z
2
, . . . , Z
k
) = (U
1
, . . . , U
i
, . . . , U
j
, . . . , U
nk
).
Hence, in order to check the multi-linearity it is sucient to prove the linearity of with respect
to the rst argument only. It is also clear that
(U
1
, . . . , U
nk
) = (U
1
, . . . , U
nk
). (4.1.2)
Indeed, multiplication by ,= 0 does not change the span of the vectors U
1
, . . . , U
nq
, and hence
if (U
1
, . . . , U
nk
) = (Z
1
, . . . , Z
k
) then (U
1
, . . . , U
nk
) = (Z
1
, . . . , Z
k
) = (Z
1
, . . . , Z
k
).
Thus it remains to check that
(U
1
+

U
1
, U
2
, . . . , U
nk
) = (U
1
, U
2
, . . . , U
nk
) +(

U
1
, U
2
, . . . , U
nk
)).
Let us denote L := Span(U
2
, . . . , U
nk
) and observe that proj
L
(U
1
+

U
1
) = proj
L
(U
1
) +
proj
L
(

U
1
). Denote N := U
1
proj
L
(U
1
) and

N :=

U
1
proj
L
(

U
1
). The vectors N and

N are
normal components of U
1
and

U
1
with respect to the subspace L, and the vector N +

N is the
normal component of U
1
+

U
1
with respect to L. Hence, we have
(U
1
, . . . , U
nk
) = (N, . . . , U
nk
), (

U
1
, . . . , U
nk
) = (

N, . . . , U
nk
),
and
(U
1
+

U
1
, . . . , U
nk
) = (N +

N, . . . , U
nk
).
Indeed, in each of these three cases,
36
- vectors on both side of the equality span the same space;
- the parallelepiped which they generate have the same volume, and
- the orientation which these vectors dene together with a basis of the complementary space
remains unchanged.
Hence, it is sucient to prove that
(N +

N, U
2
, . . . , U
nk
) = (N, U
2
, . . . , U
nk
) +(

N, U
2
, . . . , U
nk
). (4.1.3)
If the vectors N and

N are linearly dependent, i.e. one of them is a multiple of the other, then
(4.1.3) follows from (4.1.2).
Suppose now that N and

N are linearly independent. Let L

denote the orthogonal complement

of L = Span(U
2
, . . . , U
nk
). Then dimL

= k + 1 and we have N,

N L

. Let us denote by M
the plane in L

spanned by the vectors N and

N, and by M

its orthogonal complement in L

.
Note that dimM

= k 1.
Choose any orientation of M so that we can talk about counter-clockwise rotation of this plane.
Let Y,

Y M be vectors obtained by rotating N and

N in M counter-clockwise by the angle

2
.
Then Y +

Y can be obtained by rotating N +

N in M counter-clockwise by the same angle

2
. Let
us choose in M

a basis Z
2
, . . . , Z
k
such that
Vol
k1
P(Z
2
, . . . , Z
k
) = Vol
nk1
P(U
2
, . . . , U
nk
).
Note that the orthogonal complements to Span(N, U
2
, . . . , U
nk
), Span(

N, U
2
, . . . , U
nk
), and to
Span(N+

N, U
2
, . . . , U
nk
) in V coincide, respectively, with the orthogonal complements to the the
vectors N,

N and to N +

N in L

. In other words, we have

(Span(N, U
2
, . . . , U
nk
))

V
= Span(Y, Z
2
, . . . , Z
k
),
_
Span(

N, U
2
, . . . , U
nk
)
_

V
= Span(

Y , Z
2
, . . . , Z
k
) and
_
Span(N +

N, U
2
, . . . , U
nk
)
_

V
= Span(Y +

Y , Z
2
, . . . , Z
k
).
Next, we observe that
Vol
nk
P(N, U
2
, . . . , U
nk
) = Vol
k
P(Y, Z
2
, . . . , Z
k
),
37
Vol
nk
P(

N, U
2
, . . . , U
nk
) = Vol
k
P(

Y , Z
2
, . . . , Z
k
) and
Vol
nk
P(N +

N, U
2
, . . . , U
nk
) = Vol
k
P(Y +

Y , Z
2
, . . . , Z
k
).
Consider the following 3 bases of V :
Y, Z
2
, . . . , Z
k
, N, U
2
, . . . , U
nk
,

Y , Z
2
, . . . , Z
k
,

N, U
2
, . . . , U
nk
,
Y +

Y , Z
2
, . . . , Z
k
, N +

N, U
2
, . . . , U
nk
,
and observe that all three of them dene the same a priori given orientation of V . Thus, by denition
of the operator we have:
(N +

N, U
2
, . . . , U
nk
) = (Y +

Y , Z
2
, . . . , Z
k
)
= (Y, Z
2
, . . . , Z
k
) +(

Y , Z
2
, . . . , Z
k
) = (N, U
2
, . . . , U
nk
) +(

N, U
2
, . . . , U
nk
).
This completes the proof that is an (n k)-form.
Thus the map denes a map :
k
(V

)
nk
(V

). Clearly, this map is linear. In

order to check that is an isomorphism let us choose an orthonormal basis in V and consider the
coordinates x
1
, . . . , x
n
V

corresponding to that basis.

Let us recall that the forms x
i
1
x
i
2
x
i
k
, 1 i
1
< i
2
< < i
k
n, form a basis of the
space
k
(V

).
Lemma 4.3.
x
i
1
x
i
2
x
i
k
= (1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)
x
j
1
x
j
2
x
j
nk
, (4.1.4)
where j
1
< < j
nk
is the set of indices, complementary to i
1
, . . . , i
k
. In other words, i
1
, . . . , i
k
, j
1
, . . . , j
nk
is a permutation of indices 1, . . . , n.
Proof. Evaluating (x
i
1
x
i
k
) on basic vectors v
j
1
, . . . , v
j
nk
, 1 j
1
< < j
nq
n, we get
0 unless all the indices j
1
, . . . , j
nk
are all dierent from i
1
, . . . , i
k
, while in the latter case we get
(x
i
1
x
i
k
)(v
j
1
, . . . , v
j
nk
) = (1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)
.
38
Hence,
x
i
1
x
i
2
x
i
k
= (1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)
x
j
1
x
j
2
x
j
nk
.

Thus establishes a 1 to 1 correspondence between the bases of the spaces

k
(V

) and
the space
nk
(V

), and hence it is an isomorphism. Note that by linearity for any form =

1i
1
<<i
k
n
a
i
1
...i
k
x
i
1
x
i
k
) we have
=

1i
1
<<i
k
n
a
i
1
...i
k
(x
i
1
x
i
k
) .
Examples.
1. C = Cx
1
x
n
; in other words the isomorphism acts on constants (= 0-forms) by
multiplying them by the volume form.
2. In R
3
we have
x
1
= x
2
x
3
, x
2
= x
1
x
3
= x
3
x
1
, x
3
= x
1
x
2
,
(x
1
x
2
) = x
3
, (x
3
x
1
) = x
2
, (x
2
x
3
) = x
1
.
3. More generally, given a 1-form l = a
1
x
1
+ +a
n
x
n
we have
l = a
1
x
2
x
n
a
2
x
1
x
3
x
n
+ + (1)
n1
a
n
x
1
x
n1
.
In particular for n = 3 we have
(a
1
x
1
+a
2
x
2
+a
3
x
3
) = a
1
x
2
x
3
+a
2
x
3
x
1
+a
3
x
1
x
2
.
Proposition 4.4.

2
= (1)
k(nk)
Id, i.e. () = (1)
k(nk)
for any k-form .
In particular, if dimension n = dimV is odd then
2
= Id. If n is even and is a k-form then
() = if k is even, and () = if k is odd.
39
Proof. It is sucient to verify the equality
() = (1)
k(nk)

for the case when is a basic form, i.e.

= x
i
1
x
i
k
, 1 i
1
< < i
k
n.
We have
(x
i
1
x
i
k
) = (1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)
x
j
1
x
j
2
x
j
nk
and
(x
j
1
x
j
2
x
j
nk
) = (1)
inv(j
1
,...,j
nk
,i
1
,...,i
k
)
x
i
1
x
i
k
.
But the permutations i
1
. . . i
k
j
1
. . . j
nk
and j
1
. . . j
nk
i
1
. . . i
k
dier by k(n k) transpositions of
pairs of its elements. Hence, we get
(1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)
= (1)
k(nk)
(1)
inv(j
1
,...,j
nk
,i
1
,...,i
k
)
,
and, therefore,

_
(x
i
1
x
i
k
)
_
=
_
(1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)
x
j
1
x
j
2
x
j
nk
_
= (1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)
(x
j
1
x
j
2
x
j
nk
)
= (1)
inv(i
1
,...,i
k
,j
1
,...,j
nk
)+inv(j
1
,...,j
nk
,i
1
,...,i
k
)
x
i
1
x
i
k
= (1)
k(nk)
x
i
1
x
i
k
.

Exercise 4.5. (a) For any special orthogonal operator / the operators /

and commute, i.e.

/

= /

.
(b) Let A be an orthogonal matrix of order n with det A = 1. Prove that for any k 1, . . . , n
the absolute value of each k-minor M of A is equal to the absolute value of its complementary
minor of order (n k). (Hint: Apply (a) to the form x
i
1
x
i
k
).
40
(c) Let V be an oriented 3-dimensional Euclidean space. Prove that for any two vectors X, Y V ,
their cross-product can be written in the form
X Y = T
1
((T(X) T(Y ))) .
4.2 Euclidean structure on the space of exterior forms
Suppose that the space V is oriented and Euclidean, i.e. it is endowed with an inner product , )
and an orientation.
Given two forms ,
k
(V

), k = 0, . . . , n, let us dene
, )) = ( ).
Note that is an n-form for every k, and hence, , )) is a 0-form, i.e. a real number.
Proposition 4.6. 1. The operation , )) denes an inner product on
k
(V

) for each k =
0, . . . , n.
2. If / : V V is a special orthogonal operator then the operator /

:
k
(V

)
k
(V

) is
orthogonal with respect to the inner product , )).
Proof. 1. We need to check that , )) is a symmetric bilinear function on
k
(V

) and , )) > 0
unless = 0. Bilinearity is straightforward. Hence, it is sucient to verify the remaining properties
for basic vectors = x
i
1
x
i
k
, = x
j
1
x
j
k
, where 1 i
1
< < i
k
n, 1 j
1
< <
j
k
n. Here (x
1
, . . . , x
n
) is any Cartersian coordinates in V which dene its given orientation.
Note that , )) = 0 = , )) unless i
m
= j
m
for all m = 1, . . . , k, and in the the latter case
we have = . Furthermore, we have
, )) = ( ) = (x
1
x
n
) = 1 > 0.
2. The inner product , )) is dened only in terms of the Euclidean structure and the orientation
of V . Hence, for any special orthogonal operator / (which preserves these structures) the induced
operator /

:
k
(V

)
k
(V

) preserves the inner product , )).

Note that we also proved that the basis of k-forms x
i
1
x
i
k
, 1 i
1
< < i
k
n, is
orthonormal with respect to the scalar product , )). Hence, we get
41
Corollary 4.7. Suppose that a k-form can be written in Cartesian coordinates as
=

1i
1
<<i
k
n
a
i
1
...i
k
x
i
1
x
i
k
.
Then
[[ [[
2
= , )) =

1i
1
<<i
k
n
a
2
i
1
...i
k
.
Exercise 4.8. Show that if , are 1-forms on an Euclidean space V . Then
, )) = T
1
(), T
1
()),
i.e the scalar product , )) on V

is the push-forward by T of the scalar product , ) on V .

Corollary 4.9. Let V be a Euclidean n-dimensional space. Choose an orthonormal basis e
1
, . . . , e
n
in V . Then for any vectors Z
1
= (z
11
, . . . , z
n1
), . . . , Z
k
= (z
1k
, . . . , z
nk
) V we have
(Vol
k
P(Z
1
, . . . , Z
k
))
2
=

1i
1
<<i
k
n
Z
2
i
1
,...,i
k
, (4.2.1)
where
Z
i
1
,...,i
k
=

z
i
1
1
. . . z
i
1
k
. . . . . . . . .
z
i
k
1
. . . z
i
k
k

.
Proof. Consider linear functions l
j
= T(Z
j
) =
n

i=1
z
ij
x
i
V

, j = 1, . . . , k. Then
l
1
l
k
=
n

i
1
=1
z
i
1
j
x
i
1

n

i
k
=1
z
i
k
j
x
i
k
=

i
1
,...,i
k
z
i
1
. . . z
i
k
x
i
1
x
i
k
=

1i
1
<<i
k
n
Z
i
1
,...,i
k
x
i
1
. . . x
i
k
. (4.2.2)
In particular, if one has Z
1
, . . . Z
k
Span(e
1
, . . . , e
k
) then Z
1...k
= Vol P(Z
1
, . . . , Z
k
) and hence
l
1
l
k
= Z
1...k
x
1
x
k
= Vol P(Z
1
, . . . , Z
k
)x
1
x
k
,
which yields the claim in this case.
42
In the general case, according to Proposition 4.7 we have
[[ l
1
l
k
[[
2
=

1i
1
<<i
k
n
Z
2
i
1
,...,i
k
, (4.2.3)
which coincides with the right-hand side of (4.2.1). Thus it remains to check to that
[[ l
1
l
k
[[= Vol
k
P(Z
1
, . . . , Z
k
).
Given any orthogonal transformation / : V V we have, according to Proposition 4.6, the equality
[[ l
1
l
k
[[=[[ /

l
1
/

l
k
[[ . (4.2.4)
We also note that any orthogonal transformation B : V V preserves k-dimensional volume of all
k-dimensional parallelepipeds:
[Vol
k
P(Z
1
, . . . , Z
k
)[ = [Vol
k
P(B(Z
1
), . . . , B(Z
k
))[. (4.2.5)
On the other hand, there exists an orthogonal transformation / : V V such that /
1
(Z
1
), . . . , /
1
(Z
k
)
Span(e
1
, . . . , e
k
). Denote

Z
j
:= /
1
(Z
j
), j = 1, . . . , k. Then, according to (3.2.2) we have

l
j
:= T(

Z
j
) = T(/
1
(Z
j
)) = /

(T(Z
j
)) = /

l
j
.
As was pointed out above we then have
[Vol
k
P(

Z
1
, . . . ,

Z
k
)[ =[[

l
1

l
k
[[=[[ /

l
1
/

l
k
[[, (4.2.6)
and hence, the claim follows from (4.2.4) and (4.2.5) applied to B = /
1
.
We recall that an alternative formula for computing Vol
k
P(Z
1
, . . . , Z
k
) was given earlier in
Proposition 3.3.
4.3 Contraction
Let V be a vector space and
k
(V

) a k-form. Dene a (k 1)-form = v by the formula

(X
1
, . . . , X
k1
) = (v, X
1
, . . . , X
k1
)
for any vectors X
1
, . . . , X
k1
V . We say that the form is obtained by a contraction of with
the vector v. Sometimes, this operation is called also an interior product of with v and denoted
by i(v) instead of v . In these notes we will not use this notation.
43
Proposition 4.10. Contraction is a bilinear operation, i.e.
(v
1
+v
2
) = v
1
+v
2

(v) = (v )
v (
1
+
2
) = v
1
+v
2
v () = (v ).
Here v, v
1
, v
2
V ; ,
1
,
2

k
(V

); R.
The proof is straightforward.
Let be a non-zero n-form. Then we have
Proposition 4.11. The map : V
n1
(V

), dened by the formula (v) = v is an isomor-

phism between the vector spaces V and
n1
(V

).
Proof. Take a basis v
1
, . . . , v
n
. Let x
1
, . . . , x
n
V

be the dual basis, i.e. the corresponding coor-

dinate system. Then = ax
1
. . . x
n
, where a ,= 0. To simplify the notation let us assume that
a = 1, so that
= x
1
. . . x
n
.
Let us compute the images v
i
, i = 1, . . . , k of the basic vectors. Let us write
v
i
=
n

1
a
j
x
1
x
j1
x
j+1
x
n
.
Then
v
i
(v
1
, . . . , v
l1
, v
l+1
, . . . , v
n
)
=
n

1
a
j
x
1
x
j1
x
j+1
x
n
(v
1
, . . . , v
l1
, v
l+1
, . . . , v
n
) = a
l
, (4.3.1)
44
but on the other hand,
v
i
(v
1
, . . . , v
l1
, v
l+1
, . . . , v
n
) = (v
i
, v
1
, . . . , v
l1
, v
l+1
, . . . , v
n
)
= (1)
i1
(v
1
, . . . , v
l1
, v
i
, v
l+1
, . . . , v
n
) =
_

_
(1)
i1
, l = i ;
0 , otherwise
(4.3.2)
Thus,
v
i
= (1)
i1
x
1
x
i1
x
i+1
x
n
.
Hence, the map sends a basis of V

into a basis of
n1
(V

), and therefore it is an isomorphism.

Take a vector v =
n

1
a
j
v
j
. Then we have
v (x
1
x
n
) =
n

1
(1)
i1
a
i
x
1
x
i1
x
i+1
x
n
. (4.3.3)
This formula can be interpreted as the formula of expansion of a determinant according to the rst
column (or the rst row). Indeed, for any vectors U
1
, . . . , U
n1
we have
v (U
1
, . . . , U
n1
) = det(v, U
1
, . . . , U
n1
) =

a
1
u
1,1
. . . u
1,n1
. . . . . . . . . . . .
a
n
u
n,1
. . . u
n,n1

,
where
_
_
_
_
_
u
1,i
.
.
.
u
n,i
_
_
_
_
_
are coordinates of the vector U
i
V in the basis v
1
, . . . , v
n
. On the other hand,
v (U
1
, . . . , U
n1
) =
n

1
(1)
i1
a
i
x
1
x
i1
x
i+1
x
n
(U
1
, . . . , U
n1
)
= a
1

u
2,1
. . . u
2,n1
u
3,1
. . . u
3,n1
. . . . . . . . .
u
n,1
. . . u
n,n1

+ + (1)
n1
a
n

u
1,1
. . . u
1,n1
u
2,1
. . . u
2,n1
. . . . . . . . .
u
n1,1
. . . u
n1,n1

. (4.3.4)
45
Suppose that dimV = 3. Then the formula (4.3.3) can be rewritten as
v (x
1
x
2
x
3
) = a
1
x
2
x
3
+a
2
x
3
x
1
+a
3
x
1
x
2
,
where
_
_
_
_
_
a
1
a
2
a
3
_
_
_
_
_
are coordinates of the vector V . Let us describe the geometric meaning of the
operation . Set = v (x
1
x
2
x
3
). Then (U
1
, U
2
) is the volume of the parallelogram dened
by the vectors U
1
, U
2
and v. Let be the unit normal vector to the plane L(U
1
, U
2
) V . Then we
have
(U
1
, U
2
) = Area P(U
1
, U
2
) v, ) .
If we interpret v as the velocity of a uid ow in the space V then (U
1
, U
2
) is just an amount
of uid own through the parallelogram generated by vectors U
1
and U
2
for the unit time. It is
called the ux of v through the parallelogram .
Let us return back to the case dimV = n.
Exercise 4.12. Let
= x
i
1
x
i
k
, 1 i
1
< < i
k
n
and v = (a
1
, . . . , a
n
). Show that
v =
k

j=1
(1)
j+1
a
i
j
x
i
1
. . . x
i
j1
x
i
j+1
. . . x
i
k
.
The next proposition establishes a relation between the isomorphisms , and T.
Proposition 4.13. Let V be a Euclidean space, and x
1
, . . . , x
n
be coordinates in an orthonormal
basis. Then for any vector v V we have
Tv = v (x
1
x
n
) .
Proof. Let v = (a
1
, . . . , a
n
). Then Tv = a
1
x
1
+ +a
n
x
n
and
Tv = a
1
x
2
x
n
a
2
x
1
x
3
x
n
x
1
+ + (1)
n1
a
n
x
1
x
n1
.
But according to Proposition 4.11 the (n1)-form v (x
1
x
n
is dened by the same formula.

46
We nish this section by the proposition which shows how the contraction operation interacts
with the exterior product.
Proposition 4.14. Let , be forms of order k and l, respectively, and v a vector. Then
v ( ) = (v ) + (1)
k
(v ).
Proof. Note that given any indices k
1
, . . . k
m
(not necessarily ordered) we have e
i
x
k
1
x
km
= 0
if i / k
1
, . . . , k
m
and e
i
x
k
1
x
km
= (1)
J
x
k
1
. . .
i
x
km
, where J = J(i; k
1
, . . . , k
m
)
is the number of variables ahead of x
i
.
By linearity it is sucient to consider the case when v, , are basic vector and forms, i.e.
v = e
i
, = x
i
1
x
i
k
, = x
j
1
. . . x
j
l
.
We have v ,= 0 if and only if the index i is among the indices i
1
, . . . , i
k
. In that case
v = (1)
J
x
i
1
. . .
i
x
i
k
and if v ,= 0 then
v = (1)
J

x
j
1
. . .
i
x
j
l
,
where J = J(i; i
1
, . . . , i
k
), J

= J(i; j
1
, . . . , j
l
).
If it enters bot not then
v ( ) = (1)
J
x
i
1
. . .
i
x
i
k
x
j
1
x
j
l
= (v ) ,
while (v ) = 0.
Similarly, if it enters but not then
v ( ) = (1)
J

+m
x
i
1
x
i
k
x
j
1
. . .
i
x
j
l
= (1)
k
(v ),
while v = 0. Hence, in both these cases the formula holds.
If x
i
enters both products then = 0, and hence v ( = 0.
On the other hand,
(v ) + (1)
k
(v ) = (1)
J
x
i
1
. . .
i
x
i
k
x
j
1
x
j
l
+ (1)
k+J

x
i
1
x
i
k
x
j
1
. . .
i
x
j
l
= 0,
47
because the products x
i
1
. . .
i
x
i
k
x
j
1
x
j
l
and x
i
1
x
i
k
x
j
1
. . .
i
x
j
l
dier
only in the position of x
i
. In the rst product it is at the (k + J

)-s position, and in the second

at (J + 1)-st position. Hence, the dierence in signs is (1)
J+J

+k+1
, which leads to the required
cancellation.

48
Chapter 5
Complex vector spaces
5.1 Complex numbers
The space R
2
can be endowed with an associative and commutative multiplication operation. This
operation is uniquely determined by three properties:
it is a bilinear operation;
the vector (1, 0) is the unit;
the vector (0, 1) satises (0, 1)
2
= (0, 1).
The vector (0, 1) is usually denoted by i, and we will simply write 1 instead of the vector (1, 0).
Hence, any point (a, b) R
2
can be written as a +bi, where a, b R, and the product of a +bi and
c +di is given by the formula
(a +bi)(c +di) = ac bd + (ad +bc)i.
The plane R
2
endowed with this multiplication is denoted by C and called the set of complex
numbers. The real line generated by 1 is called the real axis, the line generated by i is called the
imaginary axis. The set of real numbers R can be viewed as embedded into C as the real axis.
Given a complex number z = x +iy, the numbers x and y are called its real and imaginary parts,
respectively, and denoted by Rez and Imz, so that z = Rez +iImz.
49
For any non-zero complex number z = a + bi there exists an inverse z
1
such that z
1
z = 1.
Indeed, we can set
z
1
=
a
a
2
+b
2

b
a
2
+b
2
i.
The commutativity, associativity and existence of the inverse is easy to check, but it should not
be taken for granted: it is impossible to dene a similar operation any R
n
for n > 2.
Given z = a +bi C its conjugate is dened as z = a bi. The conjugation operation z z is
the reection of C with respect to the real axis R C. Note that
Rez =
1
2
(z + z), Imz =
1
2i
(z z).
Let us introduce the polar coordinates (r, ) in R
2
= C. Then a complex number z = x + yi
can be written as r cos + ir sin = r(cos + i sin). This form of writing a complex number
is called, sometimes, ttrigonometric. The number r =
_
x
2
+y
2
is called the modulus of z and
denoted by [z[ and is called the argument of and denoted by arg z. Note that the argument is
dened only mod 2. The value of the argument in [0, 2) is sometimes called the principal value
of the argument. When z is real than its modulus [z[ is just the absolute value. We also not that
[z[ =

z z.
An important role plays the triangle inequality

[z
1
[ [z
2
[

[z
1
+z
2
[ [z
1
[ +[z
2
[.
Exponential function of a complex variable
Recall that the exponential function e
x
has a Taylor expansion
e
x
=

0
x
n
n!
= 1 +x +
x
2
2
+
x
3
6
+. . . .
We then dene for a complex the exponential function by the same formula
e
z
:= 1 +z +
z
2
2!
+ +
z
n
n!
+. . . .
One can check that this power series absolutely converging for all z and satises the formula
e
z
1
+z+2
= e
z
1
e
z
2
.
50
Figure 5.1: Leonhard Euler (1707-1783)
In particular, we have
e
iy
= 1 +iy
y
2
2!
i
y
3
3!
+
y
4
4!
+ +. . . (5.1.1)
=

k=0
(1)
k
y
2k
2k!
+i

k=0
(1)
k
y
2k+1
(2k + 1)!
. (5.1.2)
But

k=0
(1)
k y
2k
2k!
= cos y and

k=0
(1)
k y
2k+1
(2k+1)!
= siny, and hence we get Eulers formula
e
iy
= cos y +i siny,
and furthermore,
e
x+iy
= e
x
e
iy
= e
x
(cos y +i siny),
i.e. [e
x+iy
[ = e
x
, arg(e
z
) = y. In particular, any complex number z = r(cos + i sin) can be
rewritten in the form z = re
i
. This is called the exponential form of the complex number z.
Note that
_
e
i
_
n
= e
in
,
and hence if z = re
i
then z
n
= r
n
e
in
= r
n
(cos n +i sinn).
Note that the operation z iz is the rotation of C counterclockwise by the angle

2
. More
generally a multiplication operation z zw, where w = e
i
is the composition of a rotation by
the angle and a radial dilatation (homothety) in times.
51
Exercise 5.1. 1. Compute
n

0
cos k and
n

1
sink.
2. Compute 1 +
_
n
4
_
+
_
n
8
_
+
_
n
12
_
+. . . .
5.2 Complex vector space
In a real vector space one knows how to multiply a vector by a real number. In a complex vector
space there is dened an operation of multiplication by a complex number. Example is the space
C
n
whose vectors are n-tuples z = (z
1
, . . . , z
n
) of complex numbers, and multiplication by any
complex number = + i is dened component-wise: (z
1
, . . . , z
n
) = (z
1
, . . . , z
n
). Complex
vector space can be viewed as an upgrade of a real vector space, or better to say as a real vector
space with an additional structure.
In order to make a real vector space V into a complex vector space, one just needs to dene how
to multiply a vector by i. This operation must be a linear map : V V which should satisfy
the condition
2
= Id, i.e ((v)) = i(iv) = v.
Example 5.2. Consider R
2n
with coordinates (x
1
, y
1
, . . . , x
n
, y
n
). Consider a 2n 2n-matrix
J =
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
0 1 0 0
1 0 0 0
0 0 0 1
0 0 1 0
. . .
0 1
1 0
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
_
Then J
2
= I. Consider a linear operator : R
2n
R
2n
with this matrix, i.e. (Z) = JZ for
any vector Z R
2n
which we view as a column-vector. Then
2
= Id, and hence we can dene
on R
2n
a complex structure (i.e.the multiplication by i by the formula
iZ = (Z), Z R
2n
.
This complex vector space is canonically isomorphic to C
n
, where we identify the real vector
(x
1
, y
1
, . . . , x
n
, y
n
) R
2n
with a complex vector (z
1
= x
1
+iy
1
, . . . , z
n
x
n
+iy
n
).
52
On the other hand, any complex vector space can be viewed as a real vector space. In order to
do that we just need to forget how to multiply by i. This procedure is called the realication of
a complex vector space. For example, the realication of C
n
is R
2n
. Sometimes to emphasize the
realication operation we will denote the realication of a complex vector space V by V
R
. As the
sets these to objects coincide.
Given a complex vector space V we can dene linear combinations

i
v
i
, where
i
C are
complex numbers, and thus similarly to the real case talk about really dependent, really independent
vectors. Given vectors v
1
, . . . v
n
V we dene its complex span Span
C
(V
1
, . . . , v
n
) by the formula
Span
C
(v
1
, . . . , v
n
) =
_
n

i
v
i
,
i
C.
_
. A basis of a complex vector space is a system of complex linear independent vectors v
1
, . . . , v
n
such that Span
C
(v
1
, . . . , v
n
) = V . The number of vectors in a complex basis is called the complex
dimension of V and denoted dim
C
V .
For instance dimC
n
= n. On the other hand, its realication R
2n
has real dimension 2n. In
particular, C is a complex vector space of dimension 1, and therefore it is called a complex line
rather than a plane.
Exercise 5.3. Let v
1
, . . . , v
n
be a complex basis of a complex vector space V . Find the real basis of
its realication V
R
.
Answer. v
1
, iv
1
, v
2
, iv
2
, . . . , v
n
, iv
n
.
There is another important operation which associates with a real vector space V of real dimen-
sion n a complex vector space V
C
of complex dimension n. It is done in a way similar to how we
made complex numbers out of real numbers. As a real vector space the space V
C
is just the direct
sum V V = (v, w); v, w V . This is a real space of dimension 2n. We then make V V into
a complex vector space by dening the multiplication by i by the formula:
i(v, w) = (w, v).
We will write vector (v, 0) simply by v and (0, v) = i(v, 0) by iv. Hence, every vector of V
C
can be
written as v + iw, where v, w V . If v
1
, . . . , v
n
is a real basis of V , then the same vectors form a
complex basis of V
C
.
53
5.3 Complex linear maps
Complex linear maps and their realications
Given two complex vector spaces V, W a map / : V W is called complex linear (or C-linear)
if /(X + Y ) = /(X) + /(Y ) and /(X) = /(X) for any vectors X, Y V and any complex
number C. Thus complex linearity is stronger condition than the real linearity. The dierence is
in the additional requirement that /(iX) = i/(X). In other words, the operator / must commute
with the operation of multiplication by i.
Any linear map / : C C is a multiplication by a complex number a = c + id. If we view C
as R
2
and right the real matrix of this map in the standard basis 1, i we get the matrix
_
_
c d
d c
_
_
.
Indeed, /(1) = a = c + di and /(i) = ai = d + ci, so the rst column of the matrix is equal to
_
_
c
d
_
_
and the second one is equal to
_
_
d
c
_
_
.
If we have bases v
1
, . . . , v
n
of V and w
1
, . . . w
m
of W then one can associate with / an m n
complex matrix A by the same rule as in the real case.
Recall (see Exercise 5.3) that vectors v
1
, v

1
= iv
1
, v
2
, v

2
= iv
2
, . . . , v
n
, v

n
= iv
n
and w
1
, w

1
=
iw
1
, w
2
, w

2
= iw
2
, . . . w
m
, w

m
= iw
m
) form real bases of the realications V
R
and W
R
of the spaces
V and W. if
A =
_
_
_
_
_
a
11
. . . a
1n
. . .
a
m1
. . . a
mn
_
_
_
_
_
is the complex matrix of / then the real matrix A
R
of the map / is the real basis has order 2n2n
and is obtained from A by replacing each complex element a
kl
= c
kl
+ id
kl
by a 2 2 matrix
_
_
c
kl
d
kl
d
kl
c
kl
_
_
.
Exercise 5.4. Prove that
det A
R
= [ det A[
2
.
54
Complexication of real linear maps
Given a real linear map / : V W one can dene a complex linear map /
C
: V
C
W
C
by the
formula
/
C
(v +iw) = /(v) +i/(w).
. If A is the matrix of / in a basis v
1
, . . . , v
n
then /
C
has the same matrix in the same basis viewed
as a complex basis of V
C
. The operator /
C
is called the complexication of the operator /.
In particular, one can consider C-linear functions V C on a complex vector space V . Complex
coordinates z
1
, . . . , z
n
in a complex basis are examples of C-linear functions, and any other C-linear
function on V has a form c
1
z
1
+. . . c
n
z
n
, where c
1
, . . . , c
n
C are complex numbers.
Complex-valued R-linear functions
It is sometimes useful to consider also C-valued R-linear functions on a complex vector space V ,
i.e. R-linear maps V C (i.e. a linear map V
R
R
2
). Such a C-valued function has the form
= + i, where , are usual real linear functions. For instance the function z on C is a
C-valued R-linear function which is not C-linear.
If z
1
, . . . , z
n
are complex coordinates on a complex vector space V then any R-linear complex-
valued function can be written as
n

1
a
i
z
i
+b
i
z
i
, where a
i
, b
i
C are complex numbers.
We can furthermore consider complex-valued tensors and, in particular complex-valued exterior
forms. A C-valued k-form can be written as + i where and are usual R-valued k-forms.
For instance, we can consider on C
n
the 2-form =
i
2
n

1
z
k
z
k
. It can be rewritten as =
i
2
n

1
(x
k
+iy
k
) (x
k
iy
k
) =
n

1
x
k
y
k
.
55
56
Part II
Calculus of dierential forms
57
Chapter 6
Topological preliminaries
6.1 Elements of topology in a vector space
We recall in this section some basic topological notions in a nite-dimensional vector space and
elements of the theory of continuous functions. The proofs of most statements are straightforward
and we omit them.
Let us choose in V a scalar product.
Notation B
r
(p) := x V ; [[x p[[ < r, D
r
(p) := x V ; [[x p[[ r and S
r
(p) := x
V ; [[x p[[ = r stand for open, closed balls and the sphere of radius r centered at a point p V .
Open and closed sets
A set U V is called open if for any x U there exists > 0 such that B

(x) U.
A set A V is called closed if its complement V A is open. Equivalently,
Lemma 6.1. The set A is closed if and only if for any sequence x
n
A, n = 1, 2, . . . which
converges to a point a V , the limit point a belongs to A.
Remark 6.2. It is important to note that the notion of open and closed sets are independent of
the choice of the auxiliary Euclidean structure in the space V .
Points which appear as limits of sequences of points x
n
A are called limit points of A.
There are only two subsets of V which are simultaneously open and closed: V and .
59
Figure 6.1: Bernard Bolzano (1781-1848)
Lemma 6.3. 1. For any family U

, of open sets the union

is open.
2. For any family A

, of closed sets the intersection

is closed.
3. The union
n

1
A
i
of a nite family of closed sets is closed.
4. The intersection
n

1
U
i
of a nite family of open sets is open.
By a neighborhood of a point a V we understand any open set U p.
Given a subset X V
- a subset Y X is called relatively open in Y if there exists an open set U V such that
Y = X U.
- a subset Y X is called relatively closed in Y if there exists a closed set A V such that
Y = X A.
One also call relatively and open and closed subsets of X just open and closed in X.
Given any subset X V a point a V is called
interior point for A if there is a neighborhood U p such that U A;
boundary point if it is not an interior point neither for A nor for its complement V A.
60
We emphasize that a boundary point of A may or may not belong to A. Equivalently, a point a V
is a boundary point of A if it is a limit point both for A and V A.
The set of all interior points of A is called the interior of A and denoted Int A. The set of all
boundary points of A is called the boundary of A and denoted A. The union of all limit points of
A is called the closure of A and denoted A.
Lemma 6.4. 1. We have A = A A, Int A = A A.
2. A is equal to the intersection of all closed sets containg A
3. Int U is the union of all open sets contained in A.
6.2 Everywhere and nowhere dense sets
A closed set A is called nowhere dense if Int A = . For instance any nite set is nowhere dense.
Any linear subspace L V is nowhere dense in V if dimL < dimV . Here is a more interesting
example of a nowhere dense set.
Fix some number
1
3
. For any interval = [a, b] we denote by

the open interval centered

at the point c =
a+b
2
, the middle point of , of the total length equal to (b 1a). We denote by
C() :=

. Thus C() consists of two disjoint smaller closed intervals. Let I = [0, 1]. Take
C(I) = I
1
I
2
. Take again C(I
1
) C(I
2
) then again apply the operation C to four new closed
intervals. Continue the process, and take the intersection of all sets arising on all steps of this
construction. The resulted closed set K

I is nowhere dense. It is called a Cantor set.

A subset B A is called everywhere dense in A if B A. For instance the the set Q I of
rational points in the interval I = [0, 1] is everywhere dense in I.
6.3 Compactness and connectedness
A set A V is called compact if one of the following equivalent conditions is satised:
COMP1. A is closed and bounded.
COMP2. from any innite sequence of points x
n
A one can choose a subsequence x
n
k
converging
to a point a A.
61
Figure 6.2: Karl Weierstrass (1815-1897)
COMP3. from any family U

, of open sets covering A, i.e.

A, one can choose

nitely many sets U

1
, . . . , U

k
which cover A, i.e.
k

1
U

k
A.
The equivalence of these denitions is a combination of theorems of Bolzano-Weierstrass and

Emile Borel.
A set A is called path-connected if for any two points a
0
, a
1
A there is a continuous path
: [0, 1] A such that (0) = a
0
and (1) = a
1
.
A set A is called connected if one cannot present A as a union A = A
1
A
2
such that A
1
A
2
= ,
A
1
, A
2
,= and both A
1
and A
2
are simultaneously relatively closed and open in A.
Lemma 6.5. Any path-connected set is connected. If a connected set A is open then it is path
connected.
Proof. Suppose that A is disconnected. Then it can be presented as a union A = A
0
A
1
of two
non-empty relatively open (and hence relatively closed) subsets. Consider the function : A R
dened by the formula
(x) =
_

_
0, x A
0
,
1, x A
1
.
We claim that the function is is continuous. Indeed, For each i = 0, 1 and any point a A
i
there exists > 0 such that B

(x) A A
i
. Hence the function is constant on B

(x) A,
62
Figure 6.3:

Emile Borel (1871-1956)
and hence continuous at the point x. Now take points x
0
A
0
and x
1
A
1
and connect them
by a path : [0, 1] A (this path exists because A is path-connected). Consider the function
:= : [0, 1] R. This function is continuous (as a composition of two continuous maps).
Furthermore, (0) = 0, (1) = 1. Hence, by an intermediate value theorem of Cauchy the function
must take all values in the interval [0, 1]. But this is a contradiction because by construction the
function takes no other values except 0 and 1.
Lemma 6.6. Any open connected subset U R
n
is path connected.
Proof. Take any point a U. Denote by C
a
the set of all points in U which can be connected with
a by a path. We need to prove that C
a
= U.
First, we note that C
a
is open. Indeed, if b C
a
then using openness of U we can nd > 0
such the ball B

(b) U. Any point of c B

(b) can be connected by a straight interval I

bc
B

(b)
with b, and hence it can be connected by a path with a, i.e. c C
a
. Thus B

(b) C
1
, and hence
C
a
is open. Similarly we prove that the complement U C
a
is open. Indeed, take b / C
a
. As above,
there exists an open ball B

(b) U. Then B

(b) U C
a
. Indeed, if it were possible to connect a
point c B

(b) with a by a path, then the same would be true for b, because b and c are connected
by the interval I
bc
. Thus, we have U = C
a
(U C
a
), both sets C
a
and U C
a
are open and C
a
is
non-empty. Hence, U C
a
is to be empty in view of connectedness of U. Thus, C
a
= U, i.e. U is
path-connected.
63
Exercise 6.7. Prove that any non-empty connected (= path-connected) open subset of R is equal
to an interval (a, b) (we allow here a = and b = ). If one drops the condition of openness,
then one needs to add a closed and semi-closed intervals and a point.
Remark 6.8. One of the corollaries of this exercise is that in R any connected set is path-connected.
Solution. Let A R be a non-empty connected subset. Let a < b be two points of A. Suppose
that a point c (a, b) does not belong to A. Then we can write A = A
0
A
1
, where A
0
=
A (, 0), A
1
= A (0, ). Both sets A
0
and A
1
are relatively open and non-empty, which
contradicts connectedness of A. Hence if two points a and b, a < b, are in A, then the whole
interval [a, b] is also contained in A. Denote m := inf A and M := supA (we assume that m =
if A is unbounded from below and M = + if A is unbounded from above). Then the above
argument shows that the open interval (m, M) is contained in A. Thus, there could be 5 cases:
m, M / A; in this case A = (m, M);
m A, M / A; in this case A = [m, M);
m / A, M A; in this case A = (m, M];
m, M A and m < M; in this case A = [m, M];
m, M A and m = M; in this case A consists of one point.

6.4 Connected and path-connected components

Lemma 6.9. Let A

, be any family of connected (resp. path-connected) subsets of a vector

space V . Suppose

,= . Then

is also connected (resp. path-connected)

Proof. Pick a point a

. Consider rst the case when A

are path connected. Pick a point

a

. Then a can be connected by path with all points in A

for any points in . Hence,

all points of A

and A

can be connected with each other for any ,

.
Suppose now that A

are connected. Denote A :=

. Suppose A can be presented as a

union A = U U

of disjoint relatively open subsets, where we denoted by U the set which contains
64
the point a

. Then for each the intersections U

:= U A

and U

:= U

are
relatively open in A

. We have A

= U

. By assumption, U

a, and hence U

,= . Hence,
connectedness of A

implies that U

= . But then U

= , and therefore A is connected.

Given any set A V and a point a A the connected component (resp. path-connected com-
ponent C
a
A of the point a A is the union of all connected (resp. path-connected) subsets
of A which contains the point a. Due to Lemma 6.9 the (path-)connected component C
a
is itself
(path-)connected, and hence it is the biggest (path-)connected subset of A which contains the point
a. The path-connected component of a can be equivalently dened as the set of all points of A one
can connect with a by a path in A.
Note that (path-)connected components of dierent points either coincide or do not intersect,
and hence the set A can be presented as a disjoint union of (path-)-connected components.
Lemma 6.6 shows that for open sets in a vector space V the notions of connected and path-
connected components coincide, and due to Exercise 6.7 the same is true for any subsets in R. In
particular, any open set U R can be presented as a union of disjoint open intervals, which are its
connected (= path-connected) components. Note that the number of these intervals can be innite,
but always countable.
6.5 Continuous maps and functions
Let V, W be two Euclidean spaces and A is a subset of V . A map f : A W is called continuous
if one of the three equivalent properties hold:
1. For any > 0 and any point x A there exists > 0 such that f(B

(x) A) B

(f(x)).
2. If for a sequence x
n
A there exists limx
n
= x A then the sequence f(x
n
) W converges
to f(x).
3. For any open set U W the pre-image f
1
(U) is relatively open in A.
4. For any closed set B W the pre-image f
1
(B) is relatively closed in A.
65
Let us verify equivalence of 3 and 4. For any open set U W its complement B = W U is closed
and we have f
1
(U) = A f
1
(B). Hence, if f
1
(U) is relatively open, i.e. f
1
(U) = U

A for
an open set U

V , then f
1
(B) = A (V U

), i.e. f
1
(B) is relatively closed. The converse is
similar.
Let us deduce 1 from 3. The ball B

(f(x)) is open. Hence f

1
(B

(f(x))) is relatively open in

A. Hence, there exists > 0 such that B

(x) A f
1
(B

(f(x))), i.e. f(B

(x) A) B

(f(x)).
We leave the converse and the equivalence of denition 2 to the reader.
Remark 6.10. Consider a map f : A W and denote B := f(A). Then denition 3 can be
equivalently stated as follows:
3

. For any set U B relatively open in B its pre-image f

1
(U) is relatively open in A.
Denition 4 can be reformulated in a similar way.
Indeed, we have U = U

A for an open set U

W, while f
1
(U) = f
1
(U

).
The following theorem summarize properties of continuous maps.
Theorem 6.11. Let f : A W be a continuous map. Then
1. if A is compact then f(A) is compact;
2. if A is connected then f(A) is connected;
3. if A is path connected then f(A) is path-connected.
Proof. 1. Take any innite sequence y
n
f(A). Then there exist points x
n
A such that y
n
=
f(x
n
), n = 1, . . . . Then there exists a converging subsequence x
n
k
a A. Then by continuity
limk f(x
n
k
) = f(a) f(A), i.e. f(A) is compact.
2. Suppose that f(A) can be presented as a union B
1
B
2
of simultaneously relatively open
and closed disjoint non-empty sets. Then f
1
(B
1
), f
1
(B
2
) A are simultaneously relatively open
and closed in A, disjoint and non-empty. We also have f
1
(B
1
) f
1
(B
2
) = f
1
(B
1
B
2
) =
f
1
(f(A)) = A. Hence A is disconnected which is a contradiction.
3. Take any two points y
0
, y
1
f(A). Then there exist x
0
, x
1
A such that f(x
0
) = y
0
, f(x
1
) =
y
1
. But A is path-connected. Hence the points x
0
, x
1
can be connected by a path : [0, 1] A.
Then the path f : [0, 1] f(A) connects y
0
and y
1
, i.e. f(A) is path-connected.
66
Figure 6.4: George Cantor (1845-1918)
Note that in the case W = R Theorem 6.11.1 is just the Weierstrass theorem: a continuos
function on a compact set is bounded and achieves its maximal and minimal values.
We nish this section by a theorem of George Cantor about uniform continuity.
Theorem 6.12. Let A be compact and f : A W a continuous map. Then for any > 0 there
exists > 0 such that for any x A we have f(B

(x)) B

(f(x)).
Proof. Choose > 0. By continuity of f for every point x A there exists (x) > 0 such that
f(B
(x)
(x)) B

4
(f(x)).
We need to prove that inf
xA
(x) > 0. Note that for any point in y B(x)
2
(x) we have B(x)
2
(y)
B
(x)
(x), and hence f(B(x)
2
(y)) B

(f(y)). By compactness, from the covering

xA
B(x)
2
(x) we
can choose a nite number of balls B(x
j
)
2
(x
j
), j = 1, . . . , N which still cover A. Then = min
k
(x
j
)
2
satisfy the condition of the theorem, i.e. f(B

(x)) B

(f(x)) for any x A.

67
68
Chapter 7
Vector elds and dierential forms
7.1 Dierential and gradient
Given a vector space V we will denote by V
x
the vector space V with the origin translated to the
point x V . One can think of V
x
as that tangent space to V at the point x. Though the parallel
transport allows one to identify spaces V and V
x
it will be important for us to think about them
as dierent spaces.
Let f : U R be a function on a domain U V in a vector space V . The function f is called
dierentiable at a point x U if there exists a linear function l : V
x
R such that
f(x +h) f(x) = l(h) +o([[h[[)
for any suciently small vector h, where the notation o(t) stands for any function such that
o(t)
t

t0
0. The linear function l is called the dierential of the function f at the point x and
is denoted by d
x
f. In other words, f is dierentiable at x U if for any h V
x
there exists a limit
l(h) = lim
t0
f(x +th) f(x)
t
,
and the limit l(h) linearly depends on h. The value l(h) = d
x
f(h) is called the directional derivative
of f at the point x in the direction h. The function f is called dierentiable on the whole domain
U if it is dierentiable at each point of U.
Simply speaking, the dierentiability of a function means that at a small scale near a point x
the function behaves approximately like a linear function, the dierential of the function at the
69
point x. However this linear function varies from point to point, and we call the family d
x
f
xU
of
all these linear functions the dierential of the function f, and denote it by df (without a reference
to a particular point x).
Let us summarize the above discussion. Let f : U R be a dierentiable function. Then for
each point x U there exists a linear function d
x
f : V
x
R, the dierential of f at the point x
dened by the formula
d
x
f(h) = lim
t0
f(x +th) f(x)
t
, x U, h V
x
.
We recall that existence of partial derivatives at a point a U does not guarantee the dieren-
tiability of f at the point a. On the other hand if partial derivatives exists in a neighborhood of a
and continuous at the point a then f is dierentiable at this point. The functions whose rst partial
derivatives are continuous in u are called C
1
-smooth, or sometimes just smooth. Equivalently, we
can say that f is smooth if the dierential d
x
f continuously depends on the point x U.
If v
1
, . . . , v
n
are vectors of a basis of V , parallel transported to the point x, then we have
d
x
f(v
i
) =
f
x
i
(x), x U, i = 1, . . . , n,
where x
1
, . . . , x
n
are coordinates with respect to the chosen basis v
1
, . . . , v
n
.
Notice that if f is a linear function,
f(x) = a
1
x
1
+ +a
n
x
n
,
then for each x V we have
d
x
f(h) = a
1
h
1
+ +a
n
h
n
, h = (h
1
, . . . , h
n
) V
x
.
Thus the dierential of a linear function f at any point x V coincides with this function, parallel
transported to the space V
x
. This observation, in particular, can be applied to linear coordinate
functions x
1
, . . . , x
n
with respect to a chosen basis of V .
In Section 7.7 below we will dene the dierential for maps f : U W, where W is a vector
space and not just the real line R.
70
7.2 Smooth functions
We recall that existence of partial derivatives at a point a U does not guarantee the dierentia-
bility of f at the point a. On the other hand if partial derivatives exists in a neighborhood of a and
continuous at the point a then f is dierentiable at this point. The functions whose rst partial
derivatives are continuous in u are called C
1
-smooth. Equivalently, we can say that f is smooth if
the dierential d
x
f continuously depends on the point x U.
More generally, for k 1 a function f : U R is called C
k
-smooth all its partial derivatives
up to order k are continuous in U. The space of C
k
-smooth functions is denoted by C
k
(U). We will
also use the notation C
0
(U) and C

(U) which stands, respectively, for the spaces of continuous

functions and functions with continuous derivatives of all orders. In this notes we will often speak
of smooth functions without specifying the class of smoothness, assuming that functions have as
many continuous derivatives as necessary to justify our computations.
Remark 7.1. We will often need to consider smooth maps, functions, vectors elds, dierential
forms, etc. dened on a closed subset A of a vector space V . We will always mean by that the these
objects are dened on some open neighborhood U A. It will be not important for us how exactly
these objects are extended to U but to make sense of dierentiability we need to assume that they
are extended. In fact, one can dene what dierentiability means without any extension, but this
would go beyond the goals of these lecture notes.
Moreover, a theorem of Hassler Whitney asserts that any function smooth on a closed subset
A V can be extended to a smooth function to a neighborhood U A.
7.3 Gradient vector eld
If V is an Euclidean space, i.e. a vector space with an inner product , ), then there exists a
canonical isomorphism T : V V

, dened by the formula T(v)(x) = v, x) for v, x V . Of

course, T denes an isomorphism V
x
V

x
for each x V . Set
f(x) = T
1
(d
x
f).
71
The vector f(x) is called the gradient of the function f at the point x U. We will also use
the notation gradf(x).
By denition we have
f(x), h) = d
x
f(h) for any vector h V .
If [[h[[ = 1 then d
x
f(h) = [[f(x)[[ cos , where is the angle between the vectors f(x) and h. In
particular, the directional derivative d
x
f(h) has its maximal value when = 0. Thus the direction
of the gradient is the direction of the maximal growth of the function and the length of the gradient
equals this maximal value.
As in the case of a dierential, the gradient varies from point to point, and the family of vectors
f(x)
xU
is called the gradient vector eld f.
We discuss the general notion of a vector eld in Section 7.4 below.
7.4 Vector elds
A vector eld v on a domain U V is a function which associates to each point x U a vector
v(x) V
x
, i.e. a vector originated at the point x.
A gradient vector eld f of a function f provides us with an example of a vector eld, but as
we shall see, gradient vector elds form only a small very special class of vector elds.
Let v be a vector eld on a domain U V . If we x a basis in V , and parallel transport this
basis to all spaces V
x
, x V , then for any point x V the vector v(x) V
x
is described by its
coordinates (v
1
(x), v
2
(x), . . . , v
n
(x)). Thus to dene a vector eld on U is the same as to dene n
functions v
1
, . . . , v
n
on U, i.e. to dene a map (v
1
, . . . , v
n
) : U R
n
.
Thus, if a basis of V is xed, then the dierence between the maps U R
n
and vector elds
on U is just a matter of geometric interpretation. When we speak about a vector eld v we view
v(x) as a vector in V
x
, i.e. originated at the point x U. When we speak about a map v : U R
n
we view v(x) as a point of the space V , or as a vector, with its origin at 0 V .
Vector elds naturally arise in a context of Physics, Mechanics, Hydrodynamics, etc. as force,
velocity and other physical elds.
72
There is another very important interpretation of vector elds as rst order dierential opera-
tors.
Let C

(U) denote the vector space of innitely dierentiable functions on a domain U V .

Let v be a vector eld on V . Let us associate with v a linear operator
D
v
: C

(U) C

(U),
given by the formula
D
v
(f) = df(v), f C

(U).
In other words, we compute at any point x U the directional derivative of f in the direction of
the vector v(x). Clearly, the operator D
v
is linear: D
v
(af +bg) = aD
v
(f)+bD
v
(g) for any functions
f, g C

(U) and any real numbers a, b R. It also satises the Leibniz rule:
D
v
(fg) = D
v
(f)g +fD
v
(g).
In view of the above correspondence between vector elds and rst order dierential operatopts
, it is sometimes convenient just to view a vector eld as a dierential operator. Hence, when it will
not be confusing we may drop the notation D
v
and just directly apply the vector v to a function
f.
Let v
1
, . . . , v
n
be a basis of V , and x
1
, . . . , x
n
be the coordinate functions in this basis. We
would like to introduce the notation for the vector eld obtained from vectors v
1
, . . . , v
n
by parallel
transporting them to all points of the domain U. To motivate the notation which we are going
to introduce, let us temporarily denote these vector elds by v
1
, . . . , v
n
. Observe that D
v
i
(f) =
f
x
i
, i = 1, . . . , n. Thus the operator D
v
i
is just the operator

x
i
of taking i-th partial derivative.
Hence, viewing the vector eld v
i
as a dierential operator we will just use the notation

x
i
instead
of v
i
. Given any vector eld v with coordinate functions a
1
, a
2
, . . . , a
n
: U R we have
D
v
(f)(x) =
n

i=1
a
i
(x)
f
x
i
(x), for any f C

(U),
and hence we can write v =
n

i=1
a
i

x
i
. Note that the coecients a
i
here are functions and not
constants.
73
Suppose that V, , ) is a Euclidean vector space. Choose a (not necessarily orthonormal) basis
v
1
, . . . , v
n
. Let us nd the coordinate description of the gradient vector eld f, i.e. nd the
coecients a
j
in the expansion f(x) =
n

1
a
i
(x)

x
i
. By denition we have
f(x), h) = d
x
f(h) =
n

1
f
x
j
(x)h
j
(7.4.1)
for any vector h V
x
with coordinates (h
1
, . . . , h
n
) in the basis v
1
, . . . , v
n
parallel transported to
V
x
. Let us denote g
ij
= v
i
, v
j
). Thus G = (g
ij
) is a symmetric n n matrix, which is called the
Gram matrix of the basis v
1
, . . . , v
n
. Then the equation (7.4.1) can be rewritten as
n

i,j=1
g
ji
a
i
h
j
=
n

1
f
x
j
(x)h
j
.
Because h
j
are arbitrarily numbers it implies that the coecients with h
j
in the right and left sides
should coincide for all j = 1, . . . , n. Hence we get the following system of linear equations:
n

i=1
g
ij
a
i
=
f
x
j
(x), j = 1, . . . , n, (7.4.2)
or in matrix form
G
_
_
_
_
_
a
1
.
.
.
a
n
_
_
_
_
_
=
_
_
_
_
_
f
x
1
(x)
.
.
.
f
xn
(x)
_
_
_
_
_
,
and thus
_
_
_
_
_
a
1
.
.
.
a
n
_
_
_
_
_
= G
1
_
_
_
_
_
f
x
1
(x)
.
.
.
f
xn
(x)
_
_
_
_
_
, (7.4.3)
i.e.
f =
n

i,j=1
g
ij
f
x
i
(x)

x
j
, (7.4.4)
where we denote by g
ij
the entries of the inverse matrix G
1
= (g
ij
)
1
If the basis v
1
, . . . , v
n
is orthonormal then G is the unit matrix, and thus in this case
f =
n

1
f
x
j
(x)

x
j
, (7.4.5)
74
i.e. f has coordinates (
f
x
1
, . . . ,
f
xn
). However, simple expression (7.4.5) for the gradient holds
only in the orthonormal basis. In the general case one has a more complicated expression
(7.4.4).
7.5 Dierential forms
Similarly to vector elds, we can consider elds of exterior forms, i.e. functions on U V which
associate to each point x U a k-form from
k
(V

x
). These elds of exterior k-forms are called
dierential k-forms.
Thus the relation between k-forms and dierential k-forms is exactly the same as the relation
between vectors and vector-elds. For instance, a dierential 1-form associates with each point
x U a linear function (x) on the space V
x
. Sometimes we will write
x
instead of (x) to leave
space for the arguments of the function (x).
Example 7.2. 1. Let f : V R be a smooth function. Then the dierential df is a dierential
1-form. Indeed, with each point x V it associates a linear function d
x
f on the space V
x
.
As we shall see, most dierential 1-form are not dierentials of functions (just as most vector
elds are not gradient vector elds).
2. A dierential 0-form f on U associates with each point x U a 0-form on V
x
, i. e. a number
f(x) R. Thus dierential 0-forms on U are just functions U R.
7.6 Coordinate description of dierential forms
Let x
1
, . . . , x
n
be coordinate linear functions on V , which form the basis of V

dual to a chosen
basis v
1
, . . . , v
n
of V . For each i = 1, . . . , n the dierential dx
i
denes a linear function on each
space V
x
, x V . Namely, if h = (h
1
, . . . , h
n
) V
x
then dx
i
(h) = h
i
. Indeed
d
x
x
i
(h) = lim
t0
x
i
+th
i
x
i
t
= h
i
,
independently of the base point x V . Thus dierentials dx
1
, . . . , dx
n
form a basis of the space V

x
for each x V . In particular, any dierential 1-form on v can be written as
75
= f
1
dx
1
+. . . +f
n
dx
n
,
where f
1
, . . . , f
n
are functions on V . In particular,
df =
f
x
1
dx
1
+. . . +
f
x
n
dx
n
. (7.6.1)
Let us point out that this simple expression of the dierential of a function holds in an arbitrary
coordinate system, while an analogous simple expression (7.4.5) for the gradient vector eld
is valid only in the case of Cartesian coordinates. This reects the fact that while the notion of
dierential is intrinsic and independent of any extra choices, one needs to have a background inner
product to dene the gradient.
Similarly, any dierential 2-form w on a 3-dimensional space can be written as
= b
1
(x)dx
2
dx
3
+b
2
(x)dx
3
dx
1
+b
3
(x)dx
1
dx
2
where b
1
, b
2
, and b
3
are functions on V . Any dierential 3-form on a 3-dimensional space V has
the form
= c(x)dx
1
dx
2
dx
3
for a function c on V .
More generally, any dierential k-form can be expressed as
=

1i
1
<i
2
<i
k
n
a
i
1
...i
k
dx
i
1
dx
i
k
for some functions a
i
1
...i
k
on V .
7.7 Smooth maps and their dierentials
Let V, W be two vector spaces of arbitary (not, necessarily, equal) dimensions and U V be an
open domain in V .
Recall that a map f : U W is called dierentiable if for each x U there exists a linear map
l : V
x
W
f(x)
76
such that
l(h) = lim
t0
f(x +th) f(x)
t
for any h V
x
. In other words,
f(x +th) f(x) = tl(h) +o(t), where
o(t)
t

t0
0.
The map l is denoted by d
x
f and is called the dierential of the map f at the point x U. Thus,
d
x
f is a linear map V
x
W
f(x)
.
The space W
f(x)
can be identied with W via a parallel transport, and hence sometimes it is
convenient to think about the dierential as a map V
x
W, In particular, in the case of a linear
function. i.e. when W = R it is customary to do that, and hence we dened earlier in Section 7.1
the dierential of a function f : U R at a point x U as a linear function V
x
R, i.e. an
element of V

x
, rather than a linear map V
x
W
f(x)
.
Let us pick bases in V and W and let (x
1
, . . . , x
k
) and (y
1
, . . . , y
n
) be the corresponding coor-
dinate functions. Then each of the spaces V
x
and W
y
, x V, y W inherits a basis obtained by
parallel transport of the bases of V and W. In terms of these bases, the dierential d
x
f is given by
the Jacobi matrix
_
_
_
_
_
f
1
x
1
. . .
f
1
x
k
. . . . . . . . .
fn
x
1
. . .
fn
x
k
_
_
_
_
_
In what follows we will consider only suciently smooth maps, i.e. we assume that all maps
and their coordinate functions are dierentiable as many times as we need it.
7.8 Operator f

Let U be a domain in a vector space V and f : U W a smooth map. Then the dierential df
denes a linear map
d
x
f : V
x
W
f(x)
77
for each x V .
Let be a dierential k-form on W. Thus denes an exterior k-form on the space W
y
for
each y W.
Let us dene the dierential k-form f

on U by the formula
(f

)[
Vx
= (d
x
f)

([
W
f(x)
).
Here the notation [
Wy
stands for the exterior k-form dened by the dierential form on the
space W
y
.
In other words, for any k vectors, H
1
, . . . , H
k
V
x
we have
f

(H
1
, . . . , H
k
) = (d
x
f(H
1
), . . . , d
x
f(H
k
)).
We say that the dierential form f

is induced from by the map f, or that f

is the pull-back
of by f.
Example 7.3. Let = h(x)dx
1
dx
n
. Then formula (3.2) implies
f

= h f det Dfdx
1
dx
n
.
Here
det Df =

f
1
x
1
. . .
f
1
xn
. . . . . . . . .
fn
x
1
. . .
fn
xn

is the determinant of the Jacobian matrix of f = (f

1
, . . . , f
n
).
Similarly to Proposition 1.9 we get
Proposition 7.4. Given 2 maps
U
1
f
U
2
g
U
3
and a dierential k form on U
3
we have
(g f)

() = f

(g

) .
78
An important special case of the pull-back operator f

is the restriction operator. Namely Let

L V be an ane subspace. Let j : L V be the inclusion map. Then given a dierential k-form
on a domain U V we can consider the form j

on the domain U

:= LU. This form is called

the restriction of the form to U

and it is usually denoted by [

U
. Thus the restricted form [

U
is the same form but viewed as function of a point a U

and vectors T
1
, . . . , T
k
L
a
.
7.9 Coordinate description of the operator f

Consider rst the linear case. Let / be a linear map V W and

p
(W

). Let us x coordinate
systems x
1
, . . . , x
k
in V and y
1
, . . . , y
n
in W. If A is the matrix of the map / then we already have
seen in Section 2.7 that
/

y
j
= l
j
(x
1
, . . . , x
k
) = a
j1
x
1
+a
j2
x
2
+. . . +a
jk
x
k
, j = 1, . . . , n,
and that for any exterior k-form
=

1i
1
<...<ipn
A
i
1
,...,ip
y
i
1
. . . y
ip
we have
/

1i,<...<ipn
A
i
1
...ip
l
i
1
. . . l
ip
.
Now consider the non-linear situation. Let be a dierential p-form on W. Thus it can be
written in the form
=

A
i
1
...ip
(y)dy
i
1
. . . dy
ip
for some functions A
i
1
...ip
on W .
Let U be a domain in V and f : U W a smooth map.
Proposition 7.5. f

A
i
1
...,ip
(f(x))df
i
1
. . . df
ip
, where f
1
, . . . , f
n
are coordinate functions
of the map f.
79
Proof. For each point x U we have, by denition,
f

[
Vx
= (d
x
f)

([
W
fx
)
But the coordinate functions of the linear map d
x
f are just the dierentials d
x
f
i
of the coordinate
functions of the map f. Hence the desired formula follows from the linear case proven in the previous
proposition.
7.10 Examples
1. Consider the domain U = r > 0, 0 < 2 on the plane V = R
2
with cartesian coordinates
(r, ). Let W = R
2
be another copy of R
2
with cartesian coordinates (x, y). Consider a map
P : V W given by the formula
P(r, ) = (r cos , r sin).
This map introduces (r, ) as polar coordinates on the plane W. Set = dx dy. It is called
the area form on W. Then
P

= d(r cos ) d(r sin) = (cos dr +rd(cos )) (sindr +rd(sin) =

(cos dr r sind) (sindr +r cos d) =
cos sindr dr r sin
2
d dr +r cos
2
dr d r
2
sincos d d =
r cos
2
dr d +r sin
2
dr d = rdr d.
2. Let f : R
2
R be a smooth function and the map F : R
2
R
3
be given by the formula
F(x, y) = (x, y, f(x, y))
Let
= P(x, y, z)dy dz +Q(x, y, z)dz dx +R(x, y, z)dx dy
80
Figure 7.1: Johann Friedrich Pfa (17651825)
be a dierential 2-form on R
3
. Then
F

= P(x, y, f(x, y))dy df +

+ Q(x, y, f(x, y))df dx +R(x, y, f(x, y))dx dy
= P(x, y, f(x, y))dy (f
x
dx +f
y
dy) +
+ Q(x, y, f(x, y))(f
x
dx +f
y
dy) dx +
+ R(x, y, f(x, y))dx dy =
= (R(x, y, f(x, y)) P(x, y, f(x, y))f
x
Q(x, y, f(x, y))f
y
)dx dy
where f
x
, f
y
are partial derivatives of f.
3. If p > k then the pull-back f

of a p-form on U to a k-dimensional space V is equal to 0.

7.11 Pfaan equations
Given a non-zero linear function l on an n-dimensional vector space V the equation l = 0 denes
a hyperplane, i.e. an (n 1)-dimensional subspace of V .
81
y
x
z
Figure 7.2: Contact structure
Suppose we are given a dierential 1-form on a domain U V . Suppose that
x
,= 0 for each
x U.
Then the equation
= 0 (7.11.1)
denes a hyperplane eld on U, i.e. a family of of hyperplanes
x
=
x
= 0 V
x
, x U.
The equation of this type is called Pfaan in honor of a German mathematician Johann
Friedrich Pfa (17651825).
Example 7.6. Let V = R
3
with coordinates (x, y, z)
1. Let = dz. Then = dz = 0 is the horizontal plane eld which is equal to Span(

x
,

y
).
2. Let = dzydx. Then the plane eld dzydx is shown on Fig. 7.11. This plane is non-integrable
in the following sense. There are no surfaces in R
3
tangent to . This plane eld is called a contact
structure. It plays an important role in symplectic and contact geometry, which is, in turn, the
geometric language for Mechanics and Geometric Optics.
82
Chapter 8
Exterior dierential
8.1 Coordinate denition of the exterior dierential
Let us denote by by
k
(U) the space of all dierential k-forms on U. When we will need to specify
the class of smoothness of coecients of the form we will use the notation
k
l
(U) for the space of
dierential k-forms with C
l
-smooth coecients, i.e. which depend C
l
-smoothly on a point of U.
We will dene a map
d :
k
(U)
k+1
(U),
or more precisely
d :
k
l
(U)
k+1
l1
(U),
(thus assuming that l 1), which is called the exterior dierential. In the current form it was
introduced by

Elie Cartan, but essentially it was known already to Henri Poincare.
We rst dene it in coordinates and then prove that the result is independent of the choice
of the coordinate system. Let us x a coordinate system x
1
, . . . , x
n
in V U. As a reminder, a
dierential k-form w
k
(U) has the form w =

i
1
<...<i
k
a
i
1
...,i
k
dx
i
1
. . . dx
i
k
where a
i
1
...i
k
are
functions on the domain U. Dene
dw :=

i
1
<...<i
k
da
i
1
...i
k
dx
i
1
. . . dx
i
k
.
Examples. 1. Let w
1
(U), i.e. w =
n

i=1
a
i
dx
i
. Then
83
Figure 8.1:

Elie Cartan (18691951)
dw =
n

i=1
da
i
dx
i
=
n

i=1
_
_
n

j=1
a
i
x
j
dx
j
_
_
dx
i
=

1i<jn
_
a
j
x
i

a
i
x
j
_
dx
i
dx
j
.
For instance, when n = 2 we have
d(a
1
dx
1
+a
2
dx
2
) =
_
a
2
x
1

a
1
x
2
_
dx
1
dx
2
For n = 3, we get
d(a
1
dx
1
+a
2
dx
2
+a
3
dx
3
) =
_
a
3
x
2

a
2
x
3
_
dx
2
dx
3
+
_
a
1
x
3

a
3
x
1
_
dx
3
dx
1
+
_
a
2
x
1

a
1
x
2
_
dx
1
dx
2
.
2. Let n = 3 and w
2
(U). Then
w = a
1
dx
2
dx
3
+a
2
dx
3
dx
1
+a
3
dx
1
dx
2
and
dw = da
1
dx
2
dx
3
+da
2
dx
3
dx
1
+da
3
dx
1
dx
2
=
_
a
1
x
1
+
a
2
x
2
+
a
3
x
3
_
dx
1
dx
2
dx
3
84
Figure 8.2: Henri Poincare (18541912)
3. For 0-forms, i.e. functions the exterior dierential coincides with the usual dierential of a
function.
8.2 Properties of the operator d
Proposition 8.1. For any two dierential forms,
k
(U),
l
(U) we have
d( ) = d + (1)
k
d.
Proof. We have
=

i
1
<...<i
k
a
i
1
...i
k
dx
i
1
. . . dx
i
k
=

j
1
<...<j
l
b
j
i
...j
l
dx
j
1
. . . dx
j
l
=
_
_

i
1
<...<i
k
a
i
1
...i
k
dx
i
1
. . . dx
i
k
_
_

_
_

j
i
<...<j
l
b
j
1
...j
l
dx
j
1
. . . dx
j
l
_
_
=

i
1
<...<i
k
j
1
<...<j
l
a
i
1
...i
k
b
j
1
...j
l
dx
i
l
. . . dx
i
k
dx
j
l
. . . dx
j
l
85
d( ) =

i
1
<...<i
k
j
1
<...<j
l
(b
j
1
...j
l
da
i
1
...i
k
+a
i
1
...i
k
db
j
1
...j
l
) dx
i
1
. . . dx
i
k
dx
j
1
. . . dx
j
l
=

i
1
<...<i
k
j
1
<...<j
l
b
j
1
...j
l
da
i
1
...i
k
dx
i
1
. . . dx
i
k
dx
j
l
. . . dx
j
l
+

i
1
<...<i
k
j
1
<...<j
l
a
i
1
...i
k
db
j
1
...j
l
dx
i
1
. . . dx
i
k
dx
j
1
. . . dx
j
l
=
_
_

i
1
<...<i
k
da
i
1
...i
k
dx
i
1
...
dx
i
k
_
_

b
j
1
...j
l
dx
j
1
. . . dx
j
l
_
+ (1)
k
_
_

i
1
<...<i
k
a
i
1
...i
k
dx
i
1
. . . dx
i
k
_
_

_
_

j
i
<...<j
l
db
j
1
...j
l
dx
j
1
. . . dx
j
l
_
_
= d + (1)
k
d.
Notice that the sign (1)
k
appeared because we had to make k transposition to move db
j
1
...j
l
to its
place.
Proposition 8.2. For any dierential k-form w we have
ddw = 0.
Proof. Let w =

i
1
<...<i
k
a
i
1
...i
k
dx
i
1
. . . dx
i
k
. Then we have
dw =

i
1
<...<i
k
da
i
1
...i
k
dx
i
1
. . . dx
i
k
.
Applying Proposition 8.1 we get
ddw =

i
1
<...<i
k
dda
i
1
...i
k
dx
i
1
. . . dx
i
k
da
i
1
...i
k
ddx
i
1
. . . dx
i
k
+. . .
+ (1)
k
da
i
1
...i
k
dx
i
1
. . . ddx
i
k
.
86
But ddf = 0 for any function f as was shown above. Hence all terms in this sum are equal to 0, i.e.
ddw = 0.
Denition. A k-form is called closed if d = 0. It is called exact if there exists a (k 1)-form
such that d = . The form is called the primitive of the form . The previous theorem can be
reformulated as follows:
Corollary 8.3. Every exact form is closed.
The converse is not true in general. For instance, take a dierential 1-form
=
xdy ydx
x
2
+y
2
on the punctured plane U = R
2
0 (i.e the plane R
2
with the deleted origin). It is easy to calculate
that d = 0, i.e is closed. On the other hand it is not exact. Indeed, let us write down this form
in polar coordinates (r, ). We have
x = r cos , y = r sin.
Hence,
=
1
r
2
(r cos (sindr +r cos d) r sin(cosdr r sind)) = d .
If there were a function H on U such that dH = , then we would have to have H = + const,
but this is impossible because the polar coordinate is not a continuous univalent function on U.
Hence is not exact.
However, a closed form is exact if it is dened on the whole vector space V .
Proposition 8.4. Operators f

and d commute, i.e. for any dierential k-form w

k
(W), and
a smooth map f : U W we have
df

w = f

dw
.
Proof. Suppose rst that k = 0, i.e. w is a function : W R. Then f

= f. Then
d( f) = f

d. Indeed, for any point x U and a vector X V

x
we have
87
d( f)(X) = d(d
x
f(X))
(chain rule)
But d(d
x
f(X)) = f

(d(X)).
Consider now the case of arbitrary k-form w,
w =

i
1
<...<i
k
a
i
1
...i
k
dx
i
1
. . . dx
i
k
.
Then
f

w =

i
1
<...<i
k
a
i
1
...i
k
fdf
i
1
. . . df
i
k
where f
1
, . . . , f
n
are coordinate functions of the map f. Using the previous theorem and taking into
account that d(df
i
) = 0, we get
d(f

w) =

i
1
<...<i
k
d(a
i
1
...i
k
f) df
i
1
. . . df
i
k
.
On the other hand
dw =

i
1
<...<i
k
da
i
1
...i
k
dx
i
1
. . . dx
i
k
and therefore
f

dw =

i
1
<...<i
k
f

(da
i
1
...i
k
) df
i
1
. . . df
i
k
.
But according to what is proven above, we have
f

da
i
1
...i
k
= d(a
i
1
...i
k
f)
Thus,
f

dw =

i
1
<...<i
k
d(a
i
1
...i
k
f) df
i
1
. . . df
i
k
= df

The above theorem shows, in particular, that the denition of the exterior dierential is in-
dependent of the choice of the coordinate. Moreover, one can even use non-linear (curvilinear)
coordinate systems, like polar coordinates on the plane.
88
8.3 Curvilinear coordinate systems
A (non-linear) coordinate system on a domain U in an n-dimensional space V is a smooth map
f = (f
1
, . . . , f
n
) : U R
n
such that
1. For each point x U the dierentials d
x
f
1
, . . . , d
x
f
n
(V
x
)

are linearly independent.

2. f is injective, i.e. f(x) ,= f(y) for x ,= y.
Thus a coordinate map f associates n coordinates y
1
= f
1
(x), . . ., y
n
= f
n
(x) with each point
x U. The inverse map f
1
: U

U is called the parameterization. Here U

= f(U) R
n
is the image of U under the map f. If one already has another set of coordinates x
1
. . . x
n
on
U, then the coordinate map f expresses new coordinates y
1
. . . y
n
through the old one, while the
parametrization map expresses the old coordinate through the new one. Thus the statement
g

dw = dg

w
applied to the parametrization map g just tells us that the formula for the exterior dierential is
the same in the new coordinates and in the old one.
Consider a space R
n
with coordinates (u
1
, . . . , u
n
). The j-th coordinate line is given by equations
u
i
= c
i
, i = 1, . . . , n; i ,= j. Given a domain U

R
n
consider a parameterization map g : U

U V . The images gu
i
= c
i
, i ,= j) U of coordinates lines u
i
= c
i
, i ,= j U

are called
coordinate lines in U with respect to the curvilinear coordinate system (u
1
, . . . , u
n
). For instance,
coordinate lines for polar coordinates in R
2
are concentric circles and rays, while coordinate lines
for spherical coordinates in R
3
are rays from the origin, and latitudes and meridians on concentric
spheres.
8.4 Geometric denition of the exterior dierential
We will show later (see Lemma 10.2) that one can give another equivalent denition of the operator
d) without using any coordinates at all. But as a rst step we give below an equivalent denition
of the exterior dierential which is manifestly invariant of a choice of ane coordinates.
89
Given a point a V and vectors h
1
, . . . , h
k
V
a
we will denote by P
a
(h
1
, . . . , h
k
) the k-
dimensional parallelepiped
_
a +
1
2
k

1
u
j
h
j
; [u
1
[, . . . , [u
k
[ 1
_
centered at the point a with its sides parallel to the vectors h
1
, . . . , h
k
, see Fig. ??. For instance,
for k = 1 the parallelepiped P
a
(h) is an interval centered at a of length [a[.
As it was explained above in Section 8.1 for a 0-form f
0
(U), i.e. a function f : U R its
exterior dierential is just its usual dierential, and hence it can be dened by the formula
(df)
a
(h) = d
a
f(h) = lim
t0
1
t
_
f
_
a +
th
2
_
f
_
a
th
2
__
, t R, h V
a
, a U
Proposition 8.5. Let
k1
1
(U) be a dierential (k 1)-form on a domain U R
n
. Then for
any point a U and any vectors h
1
, . . . , h
k
V
a
we have
(d)
a
(h
1
, . . . , h
k
) = lim
t0
1
t
k
k

j=1
(1)
k1
_

a+
th
j
2
(th
1
, . . . ,
j
, . . . , th
k
)
a
th
j
2
(th
1
, . . . ,
j
, . . . , th
k
)
_
= lim
t0
1
t
k

j=1
(1)
k1
_

a+
th
j
2
(h
1
, . . . ,
j
, . . . , h
k
)
a
th
j
2
(h
1
, . . . ,
j
, . . . , h
k
)
_
(8.4.1)
For instance, for a 1-form
1
(U), we have
(d)
a
(h
1
, h
2
) := lim
t0
1
t
2
_

a+
th
1
2
(th
2
) +
a
th
1
2
(th
2
) +
a+
th
2
2
(th
1
) +
a
th
2
2
(th
1
)
_
= lim
t0
1
t
_

a+
th
1
2
(h
2
)
a
th
1
2
(h
2
)
a+
th
2
2
(h
1
) +
a
th
2
2
(h
1
)
_
Proof. First we observe that for a C
1
-smooth form the limit in the right-hand side of formula
(8.4.1) exists and dene a k-form (i.e. for each a it is a multilinear skewsymmetric function of
vectors h
1
, . . . , h
k
R
n
a
). We can also assume that the vectors h
1
, . . . , h
k
are linearly independent.
Otherwise, both denitions give 0. Denote L
a
= Span(h
1
, . . . , h
k
) V
a
and consider the restriction

:= [
LU
. Let (y
1
, . . . , y
k
) be coordinates in L
a
corresponding to the basis h
1
, . . . , h
k
. Then we
90
can write

:=

P
i
dy
1
. . .
i
dy
k
and thus
lim
t0
1
t
_
_
k

j=1

a+
th
j
2
(h
1
, . . . ,
i
, . . . , h
k
)
a+
th
j
2
(h
1
, . . . ,
i
, . . . , h
k
)
_
_
=
k

j=1
k

i=1
lim
t0
P
j
(a +
th
j
2
) P
j
(a
th
j
2
)
t
dy
1
. . .
i
dy
k
(h
1
, . . . ,
j
, . . . , h
k
)
=
k

j=1
k

i=1
P
j
y
j
(a)dy
1
. . .
i
dy
k
(h
1
, . . . ,
i
, . . . , h
k
).
(8.4.2)
But
dy
1
. . .
i
dy
k
(h
1
, . . . ,
j
, . . . , h
k
) =
_

_
(1)
i1
, if i = j
0, if i ,= j.
Hence, the expression in (8.4.2) is equal to
k

1
(1)
i1 P
i
y
i
(a). But
d

=
_
k

1
(1)
i1
P
i
y
i
_
dy
1
dy
k
,
and therefore (d

)
a
(h
1
, . . . , h
k
) =
k

1
(1)
i1 P
i
y
i
(a), i.e. the two denitions of the exterior dieren-
tial coincide.
8.5 More about vector elds
Similarly to the case of linear coordinates, given any curvilinear coordinate system (u
1
, . . . , u
n
) in
U, one denotes by

u
1
, . . . ,

u
n
the vector elds which correspond to the partial derivatives with respect to the coordinates u
1
, . . . , u
n
.
In other words, the vector eld

u
i
is tangent to the u
i
-coordinate lines and represents the the ve-
locity vector of the curves u
1
= const
1
, . . . , u
i1
= const
i1
, u
i+1
= const
i+1
, . . . , u
n
= const
n
,
parameterized by the coordinate u
i
.
For instance, suppose we are given spherical coordinates (r, , ), r 0, [0, ], [0, 2) in
91
R
3
. The spherical coordinates are related to the cartesian coordinates (x, y, z) by the formulas
x = r sincos , (8.5.1)
y = r sinsin, (8.5.2)
z = r cos . (8.5.3)
Then the vector elds

r
,

, and

are mutually orthogonal and dene the same orientation of the space as the standard Cartesian
coordinates in R
3
. We also have [[

r
[[ = 1. However the length of vector elds

and

vary. When
r and are xed and varies, then the corresponding point (r, , ) is moving along a meridian of
radius r with a constant angular speed 1. Hence,
[[

[[ = r .
When r and are xed and varies, then the point (r, , ) is moving along a latitude of radius
r sin with a constant angular speed 1. Hence,
[[

[[ = r sin.
Note that it is customary to introduce unit vector elds in the direction of

r
,

and

:
e
r
:=

r
, e

:=
1
r

, e

=
1
r sin

,
which form an orthonormal basis at every point. The vector elds e
r
, e

and e

are not dened at

the origin and at the poles = 0, .
The chain rule allows us to express the vector elds

u
1
, . . . ,

un
through the vector elds

x
1
, . . . ,

xn
. Indeed, for any function f : U R we have
f
u
i
=
n

j=1
f
x
j
x
j
u
i
,
and, therefore,

u
i
=
n

j=1
x
j
u
i

x
j
,
92
For instance, spherical coordinates (r, , ) are related to the cartesian coordinates (x, y, z) by
the formulas (8.5.1), and hence we derive the following expression of the vector elds

r
,

through the vector elds

x
,

y
,

z
:

r
= sincos

x
+ sinsin

y
+ cos

z
,

= r sinsin

x
+r sincos

y
,

= r cos cos

x
+r cos sin

y
r sin

z
.
8.6 Case n = 3. Summary of isomorphisms
Let U be a domain in the 3-dimensional space V . We will consider 5 spaces associated with U.

0
(U) = C

(U)the space of 0-forms, i.e. the space of smooth functions;

k
(U) for k = 1, 2, 3the spaces of dierential k-forms on U;
Vect(U)the space of vector elds on U.
Let us x a volume form w
3
(U) that is any nowhere vanishing dierential 3-form. In coordinates
w can be written as
w = f(x)dx
1
dx
2
dx
3
where the function f : U R is never equal to 0. The choice of the form w allows us to dene the
following isomorphisms.
1.
w
: C

(U)
3
(U),
w
(h) = hw for any function h C

(U).
2.
w
: Vect(U)
2
(U)
w
(v) = v w.
Sometimes we will omit the subscript w an write just and .
Our third isomorphism depends on a choice of a scalar product <, > in V . Let us x a scalar
product. This enables us to dene an isomorphism
T = T
<,>
: Vect(U)
1
(U)
93
which associates with a vector eld v on U a dierential 1-form T(v) = v, .). Let us write down the
coordinate expressions for all these isomorphisms. Fix a cartesian coordinate system (x
1
, x
2
, x
3
) in
V so that the scalar product x, y) in these coordinates equals x
1
y
1
+ x
2
y
2
+ x
3
y
3
. Suppose also
that w = dx
1
dx
2
dx
3
. Then (h) = hdx
1
dx
2
dx
3
.
(v) = v
1
dx
2
dx
3
+v
2
dx
3
dx
1
+v
3
dx
1
dx
2
where v
1
, v
2
, v
3
are coordinate functions of the vector eld v.
T(v) = v
1
dx
1
+v
2
dx
2
+v
3
dx
3
.
If V is an oriented Euclidean space then one also has isomorphisms
:
k
(V )
3k
(V ), k = 0, 1, 2, 3.
If w is the volume form on V for which the unit cube has volume 1 and which dene the given
orientation of V (equivalently, if w = x
1
x
2
x
3
for any Cartesian positive coordinate system on
V ), then
w
(v) = T(v), and
w
= :
0
(V )
3
(V ) .
8.7 Gradient, curl and divergence of a vector eld
The above isomorphism, combined with the operation of exterior dierentiation, allows us to dene
the following operations on the vector elds. First recall that for a function f C

(U),
gradf = T
1
(df).
Now let v Vect (U) be a vector eld. Then its divergence div v is the function dened by the
formula
div v =
1
(d( v))
In other words, we take the 2-form v w (w is the volume form) and compute its exterior dierential
d(v w). The result is a 3-form, and, therefore is proportional to the volume form w, i.e. d(v w) = hw.
94
This proportionality coecient (which is a function; it varies from point to point) is simply the
divergence: div v = h.
Given a vector eld v, its curl is as another vector eld curl v dened by the formula
curl v :=
1
d(Tv) = T
1
d(Tv).
If one xes a cartesian coordinate system in V such that w = dx
1
dx
2
dx
3
and x, y) =
x
1
y
1
+x
2
y
2
+x
3
y
3
then we get the following formulas
grad f =
_
f
x
1
,
f
x
2
,
f
x
3
_
div v =
v
1
x
1
+
v
2
x
2
+
v
3
x
3
curl v =
_
v
3
x
2

v
2
x
3
,
v
1
x
3

v
3
x
1
,
v
2
x
1

v
1
x
2
_
where v = (v
1
, v
2
, v
3
).
We will discuss the geometric meaning of these operations later in Section ??.
8.8 Example: expressing vector analysis operations in spherical
coordinates
Given spherical coordinates (r, , ) in R
3
, consider orthonormal vector elds
e
r
=

r
, e

=
1
r

and e

=
1
r sin

tangent to the coordinate lines and which form orthonormal basis at every point.
Operator T.
Given a vector eld v = a
1
e
r
+ a
2
e

+ a
3
e

we compute Tv. For any vector eld h = h

1
e
r
+
h
2
e

+h
3
e

we have v, h) = Tv(h). Let us write Tv = c

1
dr +c
2
d +c
3
d. Then
v, h) = a
1
h
1
+a
2
h
2
+a
3
h
3
95
and
Tv(h) = (c
1
dr +c
2
d +c
3
d)(h
1
e
r
+h
2
e

+h
3
e

)
= c
1
h
1
dr(e
r
) +c
2
h
2
d(e

) +c
3
h
3
d(e

) = c
1
h
1
+
c
2
r
h
2
+
c
3
r sin
h
3
.
Hence, c
1
= a
1
, c
2
= ra
2
, c
3
= ra
3
sin, i.e.
Tv = a
1
dr +ra
2
d +ra
3
sind. (8.8.1)
Operator . First we express the volume form = dx dy dz in spherical coordinates. One
way to do that is just to plug into the form the expression of x, y, z through spherical coordinates.
But we can also argue as follows. We have = cdr d d. Let us evaluate both sides of this
equality on vectors e
r
, e

, e

. Then (e
r
, e

, e

) = 1 because these vectors elds are at every point

orthonormal and dene the standard orientation of the space. On the other hand,
(dr d d)(e
r
, e

, e

)
=
1
r
2
sin
(dr d d)(

r
,

) =
1
r
2
sin
,
and therefore c = r
2
sin, i.e.
= r
2
sindr d d.
In particular, given a 3-form = adr d d we get =
a
r
2
sin
.
Let us now compute dr, d and d. We argue that
dr = Ad d, d = Bdr d, d = Cd dr.
Indeed, suppose that dr = Ad d +A

d dr +A

dr d and compute the value of both

sides on vectors e

, e
r
:
dr(e

, e
r
) = dr(e

) = 0,
because the 2-dimensional volume (i.e the area) of the 2-dimensional parallelepiped P(e

, e
r
) and
the 1-dimensional volume (i.e. the length) of the 1-dimensional parallelepiped P(e) are both equal
to 1. On the other hand,
(Ad d +A

d dr +A

dr d)(e

, e
r
) =
1
r sin
A

,
96
and therefore A

= 0. Similarly, A

= 0. The argument for d and d is also similar.

It remains to compute the coecients A, B, C. We have
dr(e, e

) = dr(e
r
) = 1 = Ad d(e, e

) =
A
r
2
sin
,
d(e
r
, e

) = d(e

) =
1
r sin
= Bdr d(e
r
, e

) =
B
r
,
d(e

, e
r
) = d(e

) =
1
r
= Cd dr(e

, e
r
) =
C
r sin
.
Thus, i.e.
A = r
2
sin, B =
1
sin
, C = sin,
dr = r
2
sind d, d =
1
sin
dr d, d = sind dr . (8.8.2)
Hence we also have
d d =
1
r
2
dr,
dr d = sind,
d dr =
1
sin
d.
(8.8.3)
Now we are ready to express the vector analysis operations in spherical coordinates.
Exercise 8.6. (Gradient) Given a function f(r, , ) expressed in cylindrical coordinates, com-
pute f.
Solution. We have
f = D
1
(f
r
dr +f

d +f

d) = f
r
e
r
+
f

r
e

+
f

r sin
e

.
Here we denote by f
r
, f

and f

the respective partial derivatives of the function f.

Exercise 8.7. Given a vector eld v = ae
r
+be

+ce

compute div v and curl v.

Solution.
97
a) Divergence. We have
v = (ae
r
+be

+ce

) r
2
sindr d d
= ar
2
sind d +cr dr d +br sind dr
and
d( v ) =
_
(r
2
a
r
+ 2ra) sin +rc

+rb

sin +rb cos

_
dr d d
= (a
r
+
2a
r
+
c

r sin
+
b

r
+
b
r
cot ) .
Hence,
div v = a
r
+
2a
r
+
c

r sin
+
b

r
+
b
r
cot . (8.8.4)
b) Curl. We have
curl v = T
1
dTv = T
1
d(adr +rbd +rc sind)
= T
1

_
(a

b rb
r
)dr d + (c sin +rc
r
sin a

)d dr
+ (rb

rc

sin rc cos )d d
_
= T
1
_
rb

rc

sin rc cos
r
2
dr +
c sin +rc
r
sin a

sin
d
+ (a

b rb
r
) sind
_
=
rb

rc

sin rc cos
r
2
e
r
+
c sin +rc
r
sin a

r sin
e

+
a

b rb
r
r
e

.
(8.8.5)

8.9 Complex-valued dierential k-forms

One can consider complex-valued dierential k-forms. A C-valued dierential 1-form is a eld of
C-valued k-forms, or simply it is an expression +i, where , are usual real-valued k-forms. All
operations on complex valued k-forms ( exterior multiplication, pull-back and exterior dierential)
are dened in a natural way:
(
1
+i
1
) (
2
+i
2
) =
1

2

1

2
+i(
1

2
+
1

2
),
98
f

( +i) = f

+if

, d( +i) = d +id.
We will be in particular interested in complex valued on C. Note that a complex-valued function
(or 0-form) is on a domain U C is just a map f = u+iv : U C. Its dierential df is the same as
the dierntial of this map, but it also can be viewed as a C-valued dierential 1-form df = du+idv.
Example 8.8.
dz = dx +idy, d z = dx idy, zdz = (x +iy)(dx +idy) = xdx ydy +i(xdy +ydx),
dz d z = (dx +idy) (dx idy) = 2idx dy.
Exercise 8.9. Prove that d(z
n
) = nz
n1
dz for any integer n ,= 0.
Solution. Let us do the computation in polar coordinates. Then z
n
= r
n
e
in
and assuming
that n ,= 0 we have
d(z
n
) = nr
n1
e
in
dr +inr
n
e
in
d = nr
n1
e
in
(dr +id).
On the other hand,
nz
n1
dz = nr
n1
e
i(n1)
d(re
i
) = nr
n1
e
i(n1)
_
e
i
dr +ie
i
d
_
= nr
n1
e
in
(dr +id).
Comparing the two expressions we conclude that d(z
n
) = nz
n1
dz.
It follows that the 1-form
dz
z
n
is exact on C 0 for n > 1. Indeed,
dz
z
n
= d
_
1
(1 n)z
n1
_
.
On the other hand the form
dz
z
is closed on C 0 but not exact. Indeed,
dz
z
=
d(re
i
)
re
i
=
e
i
dr +ire
i
d
re
i
=
dr
r
+id,
and hence d
_
dz
z
_
= 0.
On the other hand, we already had seen that the form d, and hence
dz
z
, is not exact.
99
100
Chapter 9
Integration of dierential forms and
functions
9.1 Useful technical tools: partition of unity and cut-o functions
Let us recall that the support of a function is the closure of the set of points where it is not
equal to 0. We denote the support by Supp(). We say that is supported in an open set U if
Supp() U.
Lemma 9.1. There exists a C

function : R [0, ) with the following properties:

(x) 0, [x[ 1;
(x) = (x);
(x) > 0 for [x[ < 1.
Proof. There are a lot of functions with this property. For instance, one can be constructed as
follows. Take the function
h(x) =
_

_
e

1
x
2
, x > 0
0 , x 0.
(9.1.1)
The function e

1
x
2
has the property that all its derivatives at 0 are equal to 0, and hence the
function h is C

-smooth. Then the function (x) := h(1 +x)h(1 x) has the required properties.
101

Lemma 9.2. Existence of cut-off functions Let C V be compact set and U C its open
neighborhood. Then there exists a C

-smooth function
C,U
: V [0, ) with its support in U
which is equal to 1 on C
Proof. Let us x a Euclidean structure in V and a Cartesian coordinate system. Thus we can
identify V with R
n
with the standard dot-product. Given a point a V and > 0 let us denote
by
a,
the bump function on V dened by

a,
(x) :=
_
[[x a[[
2

2
_
, (9.1.2)
where : R

[0, ) is the function constructed in Lemma 9.1. Note that

a,
(x) is a C

-function
with Supp(
a,
) = D

:= B

(a) and such that

a,
(x) > 0 for x B

(a).
Let us denote by U

(C) the -neighborhood of C, i.e.

U

(C) = x V ; y C, [[y x[[ < .

There exists > 0 such that U

(C) U. Using compactness of C we can nd nitely many points

z
1
, . . . , z
N
C such that the balls B

(z
1
), . . . , B

(z
N
) U cover U

2
(C), i.e. U

2
(C)
N

1
B

(z
j
).
Consider a function

1
:=
N

z
i
,

2
: V R.
The function
1
is positive on U

2
(C) and has Supp(
1
) U.
The complement E = V U

2
(C) is a closed but unbounded set. Take a large R > 0 such that
B
R
(0) U. Then E
R
= D
R
(0) U

2
(C) is compact. Choose nitely many points x
1
, . . . , x
M
E
R
such that
M

1
B

4
(x
i
) E
R
. Notice that
M

1
B

4
(x
i
) C = . Denote

2
:=
M

x
i
,

4
.
Then the function
2
is positive on V
R
and vanishes on C. Note that the function
1
+
2
is positive
on B
R
(0) and it coincides with
1
on C. Finally, dene the function
C,U
by the formula

C,U
:=

1

1
+
2
102
on B
(
R)(0) and extend it to the whole space V as equal to 0 outside the ball B
R
(0). Then
C,U
= 1
on C and Supp(
C,U
) U, as required. .
Let C V be a compact set. Consider its nite covering by open sets U
1
, . . . , U
N
, i.e.
N
_
1
U
j
C.
We say that a nite sequence
1
, . . . ,
K
of C

-functions dened on some open neighborhood U of

C in V forms a partition of unity over C subordinated to the covering U
j

j=1,...,N
if

j
(x) = 1 for all x C;
Each function
j
, j = 1, . . . , K is supported in one of the sets U
i
, i = 1, . . . , K.
Lemma 9.3. For any compact set C and its open covering U
j

j=1,...,N
there exists a partition of
unity over C subordinated to this covering.
Proof. In view of compactness of there exists > 0 and nitely many balls B

(z
j
) centered at
points z
j
C, j = 1, . . . , K, such that
K

1
B

(z
j
) C and each of these balls is contained in one of
the open sets U
j
, j = 1, . . . , N. Consider the functions
z
j
,
dened in (9.1.2). We have
K

z
j
,
> 0
on some neighborhood U C. Let
C,U
be the cut-o function constructed in Lemma 9.2. For
j = 1, . . . , K we dene

j
(x) =
_

_
z
j
,(x)
C,U
(x)
K
P
1
z
j
,(x)
, if x U,
0, otherwise
.
Each of the functions is supported in one of the open sets U
j
, j = 1, . . . , N, and we have for every
x C
K

j
(x) =
K

z
j
,
(x)
K

z
j
,
(x)
= 1.

103
Figure 9.1: Bernhard Riemann (1826-1866)
9.2 One-dimensional Riemann integral for functions and dieren-
tial 1-forms
A partition T of an interval [a, b] is a nite sequence a = t
0
< t
1
< < t
N
= b. We will
denote by T
j
, j = 0, . . . , N the vector t
j+1
t
j
R
t
j
and by
j
the interval [t
j
, t
j+1
]. The length
t
j+1
t
j
= [[T
j
[[ of the interval
j
will be denoted by
j
. The number max
j=1,...,N

j
is called the
neness or the size of the partition T and will be denoted by (T). Let us rst recall the denition
of (Riemann) integral of a function of one variable. Given a function f : [a, b] R we will form a
lower and upper integral sums corresponding to the partition T:
L(f; T) =
N1

0
( inf
[t
j
,t
j+1
]
f)(t
j+1
t
j
) ,
U(f; T) =
N1

0
( sup
[t
j
,t
j+1
]
f)(t
j+1
t
j
) , (9.2.1)
The function is called Riemann integrable if
sup
P
L(f; T) = inf
P
U(f; T),
and in this case this number is called the (Riemann) integral of the function f over the interval
[a, b]. The integrability of f can be equivalently reformulated as follows. Let us choose a set C =
104
c
1
, . . . , c
N1
, c
j

j
, and consider an integral sum
I(f; T, C) =
N1

0
f(c
j
)(t
j+1
t
j
), c
j

j
. (9.2.2)
Then the function f is integrable if there exists a limit lim
(P)0
I(f, T, C). In this case this limit is
equal to the integral of f over the interval [a, b]. Let us emphasize that if we already know that the
function is integrable, then to compute the integral one can choose any sequence of integral sum,
provided that their neness goes to 0. In particular, sometimes it is convenient to choose c
j
= t
j
,
and in this case we will write I(f; T) instead of I(f; T, C).
The integral has dierent notations. It can be denoted sometimes by
_
[a,b]
f, but the most common
notation for this integral is
b
_
a
f(x)dx. This notation hints that we are integrating here the dierential
form f(x)dx rather than a function f. Indeed, given a dierential form = f(x)dx we have
f(c
j
)(t
j+1
t
j
) =
c
j
(T
j
),
1
and hence
I(; T, C) = I(f; T, C) =
N1

c
j
(T
j
), c
j

j
. (9.2.3)
We say that a dierential 1-form is integrable if there exists a limit lim
(P)0
I(, T, C), which is
called in this case the integral of the dierential 1-form over the oriented interval [a, b] and will
be denoted by
_

[a,b]
, or simply
b
_
a
. By denition, we say that
_

[a,b]
=
_

[a,b]
. This agrees with the
denition
_

[a,b]
= lim
(P)0
N1

c
j
(T
j
), and with the standard calculus rule
a
_
b
f(x)dx =
b
_
a
f(x)dx.
Let us recall that a map : [a, b] [c, d] is called a dieomorphism if it is smooth and has a
smooth inverse map
1
: [c, d] [a, b]. This is equivalent to one of the following:
(a) = c; (b) = d and

> 0 everywhere on [a, b]. In this case we say that preserves

orientation.
(a) = d; (b) = c and

< 0 everywhere on [a, b]. In this case we say that reverses

orientation.
1
Here we parallel transported the vector Tj from the point tj to the point cj [tj, tj+1].
105
Theorem 9.4. Let : [a, b] [c, d] be a dieomorphism. Then if a 1-form = f(x)dx is integrable
over [c, d] then its pull-back f

is integrable over [a, b], and we have

_

[a,b]

=
_

[c,d]
, (9.2.4)
if preserves the orientation and
_

[a,b]

=
_

[c,d]
=
_

[c,d]
,
if reverses the orientation.
Remark 9.5. We will show later a stronger result:
_

[a,b]

=
_

[c,d]

for any : [a, b] [c, d] with (a) = c, (b) = d, which is not necessarily a dieomorphism.
Proof. We consider only the orientation preserving case, and leave the orientation reversing one
to the reader. Choose any partition T = a = t
0
< < t
N1
< t
N
= b of the interval [a, b] and
choose any set C = c
0
, . . . , c
N1
such that c
j

j
. Then the points

t
j
= (t
j
) [c, d], j = 0, . . . N
form a partition of [c, d]. Denote this partition by

T, and denote

j
:= [

t
j
,

t
j+1
] [c, d], c
j
= (c
j
),

C = (C) = c
0
, . . . , c
N1
. Then we have
I(

, T, C) =
N1

c
j
(T
j
) =
N1

ec
j
(d(T
j
)) =
N1

ec
j
(

(c
j
)
j
). (9.2.5)
Recall that according to the mean value theorem there exists a point d
j

j
, such that

T
j
=

t
j+1

t
j
= (t
j+1
) (t
j
) =

(d
j
)(t
j+1
t
j
).
Note also that the function

is uniformly continuous, i.e. for any > 0 there exists > 0 such
that for any t, t

[a, b] such that [t t

[ < we have [

(t)

(t

)[ < . Besides, the function

106
is bounded above and below by some positive constants: m <

< M. Hence m
j
<

j
< M
j
for
all j = 1, . . . , N 1. Hence, if (T) < then we have

I(

, T, C) I(;

T,

C)

N1

ec
j
_
(

(c
j
)

(d
j
))
j
_

N1

1
f(c
j
)

=

m

I(;

T,

C)

. (9.2.6)
When (T) 0 we have

(T) = 0, and hence by assumption I(;

T,

C)
d
_
c
, but this implies
that I(

, T, C) I(;

T,

C) 0, and thus

is integrable over [a, b] and

b
_
a

= lim
(P)0
I(

, T, C) = lim
(
e
P)0
I(,

T,

C) =
d
_
c
.

If we write = f(x)dx, then

= f((t))

(t)dt and the formula (9.4) takes a familiar form

of the change of variables formula from the 1-variable calculus:
d
_
c
f(x)dx =
b
_
a
f((t))

(t)dt.
9.3 Integration of dierential 1-forms along curves
Curves as paths
A path, or parametrically given curve in a domain U in a vector space V is a map : [a, b] U. We
will assume in what follows that all considered paths are dierentiable. Given a dierential 1-form
in U we dene the integral of over by the formula
_

=
_
[a,b]

.
Example 9.6. Consider the form = dz ydx +xdy on R
3
. Let : [0, 2] R
3
be a helix given
by parametric equations x = Rcos t, y = Rsint, z = Ct. Then
_

=
2
_
0
(Cdt +R
2
(sin
2
tdt + cos
2
tdt)) =
2
_
0
(C +R
2
)dt = 2(C +R
2
).
107
Note that
_

= 0 when C = R
2
. One can observe that in this case the curve is tangent to the
plane eld given by the Pfaan equation = 0.
Proposition 9.7. Let a path be obtained from : [a, b] U by a reparameterization, i.e.
= , where : [c, d] [a, b] is an orientation preserving dieomorphism. Then
_
e
=
_

.
Indeed, applying Theorem 9.4 we get
_
e
=
d
_
c

=
d
_
c

)
b
_
a

=
_

.
A vector

(t) V
(t)
is called the velocity vector of the path .
Curves as 1-dimensional submanifolds
A subset U is called a 1-dimensional submanifold of U if for any point x there is a
neighborhood U
x
U and a dieomorphism
x
: U
x

x
R
n
, such that
x
(x) = 0 R
n
and

x
(U
x
) either coincides with x
2
= . . . x
n
= 0
x
, or with x
2
= . . . x
n
= 0, x
1
0
x
. In
the latter case the point x is called a boundary point of . In the former case it is called an interior
point of .
A 1-dimensional submanifold is called closed if it is compact and has no boundary. An example
of a closed 1-dimensional manifold is the circle S
1
= x
2
+y
2
= 1 R
2
.
WARNING. The word closed is used here in a dierent sense than when one speaks about closed
subsets. For instance, a circle in R
2
is both, a closed subset and a closed 1-dimensional submanifold,
while a closed interval is a closed subset but not a closed submanifold: it has 2 boundary points.
An open interval in R (or any R
n
) is a submanifold without boundary but it is not closed because
it is not compact. A line in a vector space is a 1-dimensional submanifold which is a closed subset
of the ambient vector space. However, it is not compact, and hence not a closed submanifold.
Proposition 9.8. 1. Suppose that a path : [a, b] U is an embedding. This means that

(t) ,= 0 for all t [a, b] and (t) ,= (t

) if t ,= t

.
2
Then = ([a, b]) is 1-dimensional
compact submanifold with boundary.
2
If only the former property is satised that is called an immersion.
108
2. Suppose U is given by equations F
1
= 0, . . . , F
n1
= 0 where F
1
, . . . , F
n1
: U R are
smooth functions such that for each point x the dierential d
x
F
1
, . . . , d
x
F
n1
are linearly
independent. Then is a 1-dimensional submanifold of U.
3. Any compact connected 1-dimensional submanifold U can be parameterized either by an
embedding : [a, b] U if it has non-empty boundary, or by an embedding : S
1

U if it is closed.
Proof. 1. Take a point c [a, b]. By assumption

(c) ,= 0. Let us choose an ane coordinate

system (y
1
, . . . , y
n
) in V centered at the point C = (c) such that the vector

(c) V
C
coincide
with the rst basic vector. In these coordinates the map gamma can be written as (
1
, . . . ,
n
)
where

1
(c) = 1,

j
(c) = 0 for j > 1 and
j
(c) = 0 for all j = 1, . . . , n. By the inverse function
theorem the function
1
is a dieomorphism of a neighborhood of c onto a neighborhood of 0 in
R (if c is one of the end points of the interval, then it is a dieomorphism onto the corresponding
one-sided neighborhood of 0). Let be the inverse function dened on the interval equal to
(, ), [0, ) and (, 0], respectively, depending on whether c is an interior point, c = a or c = b].
so that
1
((u)) = u for any u . Denote

= () [a, b]. Then (

) U can be given by
the equations:
y
2
=
2
(u) :=
2
((y
1
)), . . . , y
n
=
n
(u) :=
n
((y
1
)); y
1
.
Let us denote
= () := max
j=2,...,n
max
u
[
j
(u)[.
Denote
P

:= [y
1
[ , [y
j
[ ().
We have () P

.
We will show now that for a suciently small we have ([a, b]) P

= (). For every

point t [a, b] Int denote d(t) = [[(t) (c)[[. Recall now the condition that (t) ,= (t

)
for t ,= t

. Hence d(t) > 0 for all t [a, b] Int . The function d(t) is continuous and hence
achieve the minimum value on the compact set [a, b] Int . Denote d := min
t[a,b]\Int
d(t) > 0. Chose

< min(d, ) and such that (

) = max
j=2,...,n
max
|uc|<

[
j
(u)[ < d. Let

= [u[

[. Then
([a, b]) P

= y
2
=
2
(u), . . . , y
n
=
n
(u); y
1

.
109
2. Take a point c . The linear independent 1-forms d
c
F
1
, . . . , d
c
F
n1
V

c
can be completed
by a 1-form l V

c
to a basis of V

c
. We can choose an ane coordinate system in V with c as its
origin and such that the function x
n
coincides with l. Then the Jacobian matrix of the functions
F
1
, . . . , F
n1
, x
n
is non-degenerate at c = 0, and hence by the inverse function theorem the map
F = (F
1
, . . . , F
n1
, x
n
) : V R
n
is invertible in the neighborhood of c = 0, and hence these
functions can be chosen as new curvilinear coordinates y
1
= F
1
, . . . , y
n1
= F
n1
, y
n
= x
n
near the
point c = 0. In these coordinates the curve is given near c by the equations y
1
= = y
n1
= 0.
3. See Exercise ??.
In the case where is closed we will usually parameterize it by a path : [a, b] U
with (a) = (b). For instance, we parameterize the circle S
1
= x
2
+ y
2
= 1 R
2
by a path
[0, 2] (cos t, sint). Such , of course, cannot be an embedding, but we will require that

(t) ,= 0
and that for t ,= t

we have (t) ,= (t

) unless one of these points is a and the other one is b. We

will refer to 1-dimensional submanifolds simply as curves, respectively closed, with boundary etc.
Given a curve its tangent line at a point x is a subspace of V
x
generated by the velocity
vector

(t) for any local parameterization : [a, b] with (t) = x. If is given implicitly, as in
9.8.2, then the tangent line is dened in V
x
by the system of linear equations d
x
F
1
= 0, . . . , d
x
F
n1
=
0.
Orientation of a curve is the continuously depending on points orientation of all its tangent
lines. If the curve is given as a path : [a, b] U such that

(t) ,= 0 for all t [a, b] than

it is canonically oriented. Indeed, the orientation of its tangent line l
x
at a point x = (t) is
dened by the velocity vector

(t) l
x
.
It turns out that one can dene an integral of a dierential form over an oriented compact
curve directly without referring to its parameterization. For simplicity we will restrict our discussion
to the case when the form is continuous.
Let be a compact connected oriented curve. A partition of is a sequence of points T =
z
0
, z
1
, . . . , z
N
ordered according to the orientation of the curve and such that the boundary
points of the curve (if they exist) are included into this sequence. If is closed we assume that
z
N
= z
0
. The neness (T) of T is by denition is max
j=0,...,N1
dist(z
j
, z
j+1
) (we assume here that V
a Euclidean space).
110
Denition 9.9. Let be a dierential 1-form and a compact connected oriented curve. Let
T = z
0
, . . . , z
N
be its partition. Then we dene
_

= lim
(P)0
I(, T),
where I(, T) =
N1

z
j
(Z
j
), Z
j
= z
j+1
z
j
V
z
j
.
When is a closed submanifold then one sometimes uses the notation
_

instead of
_

.
Proposition 9.10. If one chooses a parameterization : [a, b] which respects the given
orientation of then
_

=
b
_
a

=
_

.
Proof. Indeed, let

T = t
0
, . . . , t
N
be a partition of [a, b] such that (t
j
) = z
j
, j = 0, . . . , N.
I(

,

T) =
N1

t
j
(T
j
) =
N1

z
j
(U
j
),
where U
j
= d
t
j
(T
j
)) V
z
j
is a tangent vector to at the point z
j
. Let us evaluate the dierence
U
j
Z
j
. Choosing some Cartesian coordinates in V we denote by
1
, . . . ,
n
the coordinate functions
of the path . Then using the mean value theorem for each of the coordinate functions we get

i
(t
j+1
)
i
(t
j
) =

i
(c
i
j
)
j
for some c
i
j

j
, i = 1, . . . , n; j = 0, . . . , N 1. Thus
Z
j
= (t
j+1
) (t
j
) = (

1
(c
1
j
), . . . ,

i
(c
n
j
))
j
.
On the other hand, U
j
= d
t
j
(T
j
)) V
z
j
=

(t
j
)
j
. Hence,
[[Z
j
U
j
[[ =
j

_
n

1
(

i
(c
i
j
)

i
(t
j
))
2
.
Note that if (T) 0 then we also have (

T) 0, and hence using smoothness of the path we

conclude that for any > 0 there exists > 0 such that [[Z
j
U
j
[[ <
j
for all j = 1, . . . , N. Thus
N1

z
j
(

T
j
)
N1

z
j
(Z
j
)
(P)
0,
111
and therefore
b
_
a

= lim
(
e
P)0
I(

,

T) = lim
(P)0
I(, T) =
_

9.4 Integrals of closed and exact dierential 1-forms

Theorem 9.11. Let = df be an exact 1-form in a domain U V . Then for any path : [a, b] U
which connects points A = (a) and B = (b) we have
_

= f(B) f(A).
In particular, if is a loop then
_

= 0.
Similarly for an oriented curve U with boundary = B A we have
_

= f(B) f(A).
Proof. We have
_

df =
b
_
a

df =
b
_
a
d(f ) = f((b)) f((a)) = f(B) f(A).
It turns out that closed forms are locally exact. A domain U V is called star-shaped with
respect to a point a V if with any point x U it contains the whole interval I
a,x
connecting a
and x, i.e. I
a,x
= a +t(x a); t [0, 1]. In particular, any convex domain is star-shaped.
Proposition 9.12. Let be a closed 1-form in a star-shaped domain U V . Then it is exact.
Proof. Dene a function F : U R by the formula
F(x) =
_

Ia,x
, x U,
where the intervals I
a,x
are oriented from 0 to x.
We claim that dF = . Let us identify V with the R
n
choosing a as the origin a = 0. Then
can be written as =
n

1
P
k
(x)dx
k
, and I
0,x
can be parameterized by
t tx, t [0, 1].
112
Hence,
F(x) =
_

I
0,x
=
1
_
0
n

1
P
k
(tx)x
k
dt. (9.4.1)
Dierentiating the integral over x
j
as parameters, we get
F
x
j
=
1
_
0
n

k=1
tx
k
P
k
x
j
(tx)dt +
1
_
0
P
j
(tx)dt.
But d = 0 implies that
P
k
x
j
=
P
j
x
k
, and using this we can further write
F
x
j
=
1
_
0
n

k=1
tx
k
P
j
x
k
(tx)dt +
1
_
0
P
j
(tx)dt =
1
_
0
t
dP
j
(tx)
dt
dt +
1
_
0
P
j
(tx)dt
= (tP
j
(tx))[
1
0

1
_
0
P
j
(tx)dt +
1
_
0
P
j
(tx)dt = P
j
(tx)
Thus
dF =
n

j=1
F
x
j
dx
j
=
n

j=1
P
j
(x)dx =
9.5 Integration of functions over domains in high-dimensional spaces
Riemann integral over a domain in R
n
.
In this section we will discuss integration of bounded functions over bounded sets in a vector space
V . We will x a basis e
1
, . . . , e
n
and the corresponding coordinate system x
1
, . . . , x
n
in the space
and thus will identify V with R
n
. Let denote the volume form x
1
. . . x
n
. As it will be clear
below, the denition of an integral will not depend on the choice of a coordinate system but only
on the background volume form, or rather its absolute value because the orientation of V will be
irrelevant.
We will need a special class of parallelepipeds in V , namely those which are generated by vectors
proportional to basic vectors, or in other words, parallelepipeds with edges parallel to the coordinate
axes. We will also allow these parallelepipeds to be parallel transported anywhere in the space. Let
us denote
P(a
1
, b
1
; a
2
, b
2
; . . . ; a
n
, b
n
) := a
i
x
i
b
i
; i = 1, . . . , n R
n
.
113
We will refer to P(a
1
, b
1
; a
2
, b
2
; . . . ; a
n
, b
n
) as a special parallelepiped, or rectangle.
Let us x one rectangle P := P(a
1
, b
1
; a
2
, b
2
; . . . ; a
n
, b
n
). Following the same scheme as we used
in the 1-dimensional case, we dene a partition T of P as a product of partitions a
1
= t
1
0
< <
t
1
N
1
= b
1
, . . . , a
n
= t
n
0
< < t
n
Nn
= b
n
, of intervals [a
1
, b
1
], . . . , [a
n
, b
n
]. For simplicity of notation
we will always assume that each of the coordinate intervals is partitioned into the same number
of intervals, i.e. N
1
= = N
n
= N. This denes a partition of P into N
n
smaller rectangles
P
j
= t
1
j
1
x
1
t
1
j
1
+1
, . . . , t
n
jn
x
n
t
n
jn+1
, where j = (j
1
, . . . , j
n
) and each index j
k
takes
values between 0 and N 1. Let us dene
Vol(P
j
) :=
n

k=1
(t
k
j
k
+1
t
k
j
k
). (9.5.1)
This agrees with the denition of the volume of a parallelepiped which we introduced earlier (see
formula (3.3.1) in Section 3.3). We will also denote
j
:= max
k=1,...,n
(t
k
j
k
+1
t
k
j
k
) and (T) := max
j
(
j
).
Let us x a point c
j
P
j
and denote by C the set of all such c
j
. Given a function f : P R we
form an integral sum
I(f; T, C) =

j
f(c
j
)Vol(P
j
) (9.5.2)
where the sum is taken over all elements of the partition. If there exists a limit lim
(P)0
I(f; T, C)
then the function f : P R is called integrable (in the sense of Riemann) over P, and this limit is
called the integral of f over P. There exist several dierent notations for this integral:
_
P
f,
_
P
fdV ,
_
P
fdVol, etc. In the particular case of n = 2 one often uses notation
_
P
fdA, or
__
P
fdA. Sometime, the
functions we integrate may depend on a parameter, and in these cases it is important to indicate with
respect to which variable we integrate. Hence, one also uses the notation like
_
P
f(x, y)dx
n
, where
the index n refers to the dimension of the space over which we integrate. One also use the notation
_
. . .
_
. .
P
f(x
1
, . . . x
n
)dx
1
. . . dx
n
, which is reminiscent both of the integral
_
P
f(x
1
, . . . x
n
)dx
1

dx
n
which will be dened later in Section 9.7 and the notation
bn
_
an
. . .
b
1
_
a
1
f(x
1
, . . . x
n
)dx
1
. . . dx
n
for
n interated integral which will be discussed in Section 9.6.
Alternatively and equivalently the integrability can be dened via upper and lower integral sum,
114
similar to the 1-dimensional case. Namely, we dene
U(f; T) =

j
M
j
(f)Vol(P
j
), L(f; T) =

j
m
j
(f)Vol(P
j
),
where M
j
(f) = sup
P
j
f, m
j
(f) = inf
P
j
f, and say that the function f is integrable over P if inf
P
U(f; T) =
sup
P
L(f, T).
Note that inf
P
U(f; T) and sup
P
L(f, T) are sometimes called upper and lower integrals, respec-
tively, and denoted by
_
P
f and
_
P
f. Thus a function f : P R is integrable i
_
P
f =
_
P
f.
Let us list some properties of Riemann integrable functions and integrals.
Proposition 9.13. Let f, g : P R be integrable functions. Then
1. af +bg, where a, b R, is integrable and
_
P
af +bg = a
_
P
f +b
_
P
g;
2. If f g then
_
P
f
_
P
g;
3. h = max(f, g) is integrable; in particular the functions f
+
:= max(f, 0) and f

:= max(f, 0)
and [f[ = f
+
+f

are integrable;
4. fg is integrable.
Proof. Parts 1 and 2 are straightforward and we leave them to the reader as an exercise. Let us
check properties 3 and 4.
3. Take any partition T of P. Note that
M
j
(h) m
j
(h) max(M
j
(f) m
j
(f), M
j
(g) m
j
(g)) . (9.5.3)
Indeed, we have M
j
(h) = max(M
j
(f), M
j
(g)) and m
j
(h) max(m
j
(f), m
j
(g)). Suppose for deter-
minacy that max(M
j
(f), M
j
(g)) = M
j
(f). We also have m
j
(h) m
j
(f). Thus
M
j
(h) m
j
(h) M
j
(f) m
j
(f) max(M
j
(f) m
j
(f), M
j
(g) m
j
(g)).
Then using (9.5.3) we have
U(h; T) L(h; T) =

j
(M
j
(h) m
j
(h))Vol(P
j
)

j
max (M
j
(f) m
j
(f), M
j
(g) m
j
(g)) Vol(P
j
) =
max (U(f; T) L(f; T), U(f; T) L(f; T)) .
115
By assumption the right-hand side can be made arbitrarily small for an appropriate choice of the
partition T, and hence h is integrable.
4. We have f = f
+
f

, g = g
+
g

and fg = f
+
g
+
+ f

f
+
g

g
+
. Hence, using 1
and 3 we can assume that the functions f, g are non-negative. Let us recall that the functions f, g
are by assumption bounded, i.e. there exists a constant C > 0 such that f, g C. We also have
M
j
(fg) M
j
(f)M
j
(g) and m
j
(fg) m
j
(f)m
j
(g). Hence
U(fg; T) L(fg; T) =

j
(M
j
(fg) m
j
(fg)) Vol(P
j
)

j
(M
j
(f)M
j
(g) m
j
(f)m
j
(g))Vol(P
j
) =

j
(M
j
(f)M
j
(g) m
j
(f)M
j
(g) +m
j
(f)M
j
(g) m
j
(f)m
j
(g)) Vol(P
j
)

j
((M
j
(f) m
j
(f))M
j
(g) +m
j
(f)(M
j
(g) m
j
(g))) Vol(P
j
)
C(U(f; T) L(f; T) +U(g; T) L(g; T)).
By assumption the right-hand side can be made arbitrarily small for an appropriate choice of the
partition T, and hence fg is integrable.
Consider now a bounded subset K R
n
and choose a rectangle P K. Given any function
f : K R one can always extend it to P as equal to 0. A function f : K R is called integrable
over K if this trivial extension f is integrable over P, and we dene
_
K
fdV :=
_
fdV . When this
will not be confusing we will usually keep the notation f for the above extension.
Volume
We further dene the volume
Vol(K) =
_
K
1dV =
_
P

K
dV,
provided that this integral exists. In this case we call the set K measurable in the sense of Riemann,
or just measurable.
3
Here
K
is the characteristic or indicator function of K, i.e. the function which
3
There exists a more general and more common notion of measurability in the sense of Lebesgue. Any Riemann
measurable set is also measurable in the sense of Lebesgue, but not the other way around. Historically an attribution
116
is equal to 1 on K and 0 elsewhere. In the 2-dimensional case the volume is called the area, and in
the 1-dimensional case the length.
Remark 9.14. For any bounded set A there is dened a lower and upper volumes,
Vol(A) =
_

A
dV Vol(A) =
_

A
dV.
The set is measurable i Vol(A) = Vol(A). If Vol(A) = 0 then Vol(A) = 0, and hence A is
measurable and Vol(A) = 0.
Exercise 9.15. Prove that for the rectangles this denition of the volume coincides with the one
given by the formula (9.5.1).
The next proposition lists some properties of the volume.
Proposition 9.16. 1. Volume is monotone, i.e. if A, B P are measurable and A B then
Vol(A) Vol(B).
2. If sets A, B P are measurable then AB, A B and AB are measurable as well and we
have
Vol(A B) = Vol(A) + Vol(B) Vol(A B).
3. If A can be covered by a measurable set of arbitrarily small total volume then Vol(A) = 0.
Conversely, if Vol(A) = 0 then for any > 0 there exists a > 0 such that for any partition
T with (T) < the elements of the partition which intersect A have arbitrarily small total
volume.
4. A is measurable i Vol (A) = 0.
Proof. The rst statement is obvious. To prove the second one, we observe that
AB
= max(
A
,
B
),

AB
=
A

B
, max(
A
,
B
) =
A
+
B

A

B
,
A\B
=
A

AB
and then apply Proposition
9.13. To prove 9.16.3 we rst observe that if a set B is measurable and VolB < then then for a
suciently ne partition T we have U(
B
; T) < VolB + < 2. Since A B then
A

B
, and
of this notion to Riemann is incorrect. It was dened by Camille Jordan and Giuseppe Peano before Riemann integral
was introduced. What we call in these notes volume is also known by the name Jordan content.
117
therefore U(
A
, T) U(
B
, T) < 2. Thus, inf
P
U(
A
, T) = 0 and therefore A is measurable and
Vol(A) = 0. Conversely, if Vol(A) = 0 then for any > 0 for a suciently ne partition T we have
U(
A
; T) < . But U(
A
; T) is equal to the sum of volumes of elements of the partition which have
non-empty intersection with A.
Finally, let us prove 9.16.4. Consider any partition T of P and form lower and upper integral
sums for
A
. Denote M
j
:= M
j
(
A
) and m
j
= m
j
(
A
). Then all numbers M
j
, m
j
are equal to either
0 or 1. We have M
j
= m
j
= 1 if P
j
A; M
j
= m
j
= 0 if P
j
A = and M
j
= 1, m
j
= 0 if P
j
has
non-empty intersection with both A and P A. In particular,
B(T) :=
_
j;M
j
m
j
=1
P
j
A.
Hence, we have
U(
A
; T) L(
A
; T) =

j
(M
j
m
j
)Vol(P
j
) = VolB(T).
Suppose that A is measurable. Then there exists a partition such that U(
A
; T) L(
A
; T) < ,
and hence A is can be covered by the set B(T) of volume < . Thus applying part 3 we conclude
that Vol(A) = 0. Conversely, we had seen below that if Vol(A) = 0 then there exists a partition
such that the total volume of the elements intersecting A is < . Hence, for this partition we have
L(
A
; T) U(
A
; T) < , which implies the integrability of
A
, and hence measurability of A.
Corollary 9.17. If a bounded set A V is measurable then its interior Int A and its closure A
are also measurable and we have in this case
VolA = Vol Int A = VolA.
Proof. 1. We have A A and (Int A) A. Therefore, Vol A = Vol Int A = 0, and
therefore the sets A and Int A are measurable. Also Int AA = A and Int AA = . Hence, the
additivity of the volume implies that VolA = Vol Int A+Vol A = Vol Int A. On the other hand,
Int A A A. and hence the monotonicity of the volume implies that Vol Int A Vol A Vol A.
Hence, Vol A = Vol Int A = Vol A.
Exercise 9.18. If Int A or A are measurable then this does not imply that A is measurable. For
instance, if A is the set of rational points in interval I = [0, 1] R then Int A = and A = I.
However, show that A is not Riemann measurable.
118
2. A set A is called nowhere dense if Int A = . Prove that if A is nowhere dense then either
VolA = 0, or A is not measurable in the sense of Riemann. Find an example of a non-measurable
nowhere dense set.
Volume and smooth maps
We will further need the following lemma. Let us recall that given a compact set C V , we say
that a map f : C W is smooth if it extends to a smooth map dened on an open neighborhood
U C. Here V, W are vector spaces.
Lemma 9.19. Let A V be a compact set of volume 0 and f : V W a C
1
-smooth map, where
dimW dimV . Then Volf(A) = 0.
Proof. The C
1
-smoothness of f and compactness of A imply that there exists a constant K such
that [[d
x
f(h)[[ K[[h[[ for any x A and h V
x
. In particular, the image d
x
f(P) of every
rectangle P of size in V
x
, x A, is contained in a cube of size K in W
f(x)
. C
1
-smoothness
of f also implies that for any > 0 there exists such > 0 that if x A and [[h[[ then
[[f(x+h) f(x) d
x
f(h)[[ [[h[[. This implies that if we view P as a subset of V , rather than V
x
,
then the image f(P) is contained in a cube in W of size 2K if is small enough, and if P A ,= .
Let us denote dimensions of V and W by n and m, respectively. By assumption A can be covered
by N cubes of size such that the total volume of these cubes is equal to N
n
. Hence, f(A)
can be covered by N cubes of size 2K of total volume
N(2K)
m
= N
n
(2K)
m

mn
= (2K)
m

mn

0
0,
because m n.
Corollary 9.20. Let A V be a compact domain and f : A W a C
1
-smooth map. Suppose that
n = dimV < m = dimW. Then Vol(f(A)) = 0.
Indeed, f can be extended to a smooth map dened on a neighborhood of A0 in V R (e.g.
as independent of the new coordinate t R). But Vol
n+1
(A 0) = 0 and m n + 1. Hence, the
required statement follows from Lemma 9.19.
119
Remark 9.21. The statement of Corollary 9.20 is wrong for continuous maps. For instance, there
exists a continuous map h : [0, 1] R
2
such that h([0, 1]) is the square 0 x
1
, x
1
1. (This is
a famous Peano curve passing through every point of the square.)
Corollary 9.20 is a simplest special case of Sards theorem which asserts that the set of critical
values of a suciently smooth map has volume 0. More precisely,
Proposition 9.22. (A. Sard, 1942) Given a C
k
-smooth map f : A W (where A is a compact
subset of V , dimV = n, dimW = m) let us denote by
(f) : x C; rank d
x
f < m.
Then if k max(n m+ 1, 1) then Vol
m
(f((f)) = 0.
If m > n then (f) = A, and hence the statement is equivalent to Corollary 9.20.
Proof. We prove the proposition only for the case m = n. The proof in this case is similar to the
proof of Corollary 9.19. C
1
-smoothness of f and compactness of (f) imply that there exists a
constant K such that
[[d
x
f(h)[[ K[[h[[ (9.5.4)
for any x A and h V
x
. C
1
-smoothness of f also implies that for any > 0 there exists such
> 0 that if x A and [[h[[ then
[[f(x +h) f(x) d
x
f(h)[[ [[h[[. (9.5.5)
Take a partition T of a rectangle P A by N
n
smaller rectangles of equal size. Let B be the
union of the rectangles intersecting (f). For any such rectangle P
j
B choose a point c
j
(f).
Viewing P
j
as a subset of V
c
j
we can take its image

P
j
= d
c
j
(P
bj
) W
f(c
j
)
. Then the parallelepiped

P
j
is contained in a subspace L W
f(c
j
)
of dimension r = rank(d
c
j
f) < m. In view of (9.5) it also
contained in a ball of radius
K

r
N
centered at the point f(c
j
). On the other hand, the inequality
(9.5.5) implies that if N is large enough then for any point u f(P
j
) there is a point in u

P
j
such [[u u[[

N
. This means that f(P
j
) is contained in a parallelepiped centered at c
j
and
generated by r orthogonal vectors of length
C
1
N
parallel to the subspace L and n r > 0 vectors
120
of length
C
2

N
which are orthogonal to L, where C
1
, C
2
are some positive constants. The volume of
this parallelepiped is equal to
C
r
1
C
nr
2

nr
N
n
=
C
3

nr
N
n
,
and hence Vol(f(P
j
) <
C
3

nr
N
n
. The set B contains no more that N
n
cubes P
j
, and hence
Vol(f((f)) Volf(C) N
n
C
3

nr
N
n
= C
3

nr

0
0.

Properties which hold almost everywhere

We say that some property holds almost everywhere (we will abbreviate a.e.) if it holds in the
complement of a set of volume 0. For instance, we say that a bounded function f : P R is almost
everywhere continuous (or a.e. continuous) if it is continuous in the complement of a set A P of
volume 0. For instance, a characteristic function of any measurable set is a.e. continuous. Indeed,
it is constant away from the set A which according to Proposition 9.16.4 has volume 0.
Proposition 9.23. Suppose that the bounded functions f, g : P R coincide a.e. Then if f is
integrable, then so is g and we have
_
P
f =
_
P
g.
Proof. Denote A = x P : f(x) ,= g(x). By our assumption, VolA = 0. Hence, for any there
exists a > 0 such that for every partition T with (T) the union B

of all rectangles of
the partition which have non-empty intersection with A has volume < . The functions f, g are
bounded, i.e. there exists C > 0 such C [f(x)[, [g(x)[ C for all x P. Due to integrability of
f we can choose small enough so that [U(f, T) L(f, T)[ when (T) . Then we have
[U(g, T) U(f, T)[ =

J: P
J
B

sup
P
J
g sup
P
J
f)

2CVolB

2C.
Similarly, [L(g, T) L(f, T)[ 2C, and hence
[U(g, T) L(g, T)[ [U(g, T) U(f, T)[ +[U(f, T) L(f, T)[ +[L(f, T) L(g, T)[
+ 4C
0
0,
121
and hence g is integrable and
_
P
g = lim
(P)0
U(g, T) = lim
(P)0
U(f, T) =
_
P
f.

Proposition 9.24. 1. Suppose that a function f : P R is a.e. continuous. Then f is inte-

grable.
2. Let A V be compact and measurable, f : U W a C
1
-smooth map dened on a neighbor-
hood U A. Suppose that dimW = dimV . Then f(A) is measurable.
Proof. 1. Let us begin with a
Warning. One could think that in view of Proposition 9.23 it is sucient to consider only the case
when the function f is continuous. However, this is not the case, because for a given a.e. continuos
function one cannot, in general, nd a continuos function g which coincides with f a.e.
Let us proceed with the proof. Given a partition T we denote by J
A
the set of multi-indices j such
that Int P
j
A ,= , and by J
A
the complementary set of multi-indices, i.e. for each j J
A
we have
P
j
A = . Let us denote C :=

jJ
A
P
j
. According to Proposition 9.16.3 for any > 0 there exists
a partition T such that Vol(C) =

jJ
A
Vol(P
j
) < . By assumption the function f is continuous
over a compact set B =

jJ
A
P
j
, and hence it is uniformly continuous over it. Thus there exists
> 0 such that [f(x) f(x

)[ < provided that x, x

B and [[x x

[[ < . Thus we can further

subdivide our partition, so that for the new ner partition T

we have (T

) < . By assumption
the function f is bounded, i.e. there exists a constant K > 0 such that M
j
(f) m
j
(f) < K for all
indices j. Then we have
U(f; T

) L(f, T

) =

j
(M
j
(f) m
j
(f))Vol(P
j
) =

j;P
j
B
(M
j
(f) m
j
(f))Vol(P
j
) +

j;P
j
C
(M
j
(f) m
j
(f))Vol(P
j
) <
VolB +KVolC < (VolP +K).
Hence inf
P
U(f; T) = sup
P
L(f; T), i.e. the function f is integrable.
122
2. If x is an interior point of A and detDf(x) ,= 0 then the inverse function theorem implies
that f(x) Int f(A). Denote C = x A; det Df(x) = 0. Hence, f(A) f(A) f(C).
But Vol(A) = 0 because A is measurable and Vol f(C) = 0 by Sards theorem 9.22. Therefore,
Vol f(A) = 0 and thus f(A) is measurable.
Orthogonal invariance of the volume and volume of a parallelepiped
The following lemma provides a way of computing the volume via packing by balls rather then
cubes. An admissible set of balls in A is any nite set of disjoint balls B
1
, . . . , B
K
A
Lemma 9.25. Let A be a measurable set. Then VolA is the supremum of the total volume of
admissible sets of balls in A. Here the supremum is taken over all admissible sets of balls in A.
Proof. Let us denote this supremum by . The monotonicity of volume implies that VolA.
Suppose that < V olA. Let us denote by
n
the volume of an n-dimensional ball of radius 1 (we
will compute this number later on). This ball is contained in a cube of volume 2
n
. It follows then
that the ratio of the volume of any ball to the volume of the cube to which it is inscribed is equal
to
n
2
n
. Choose an <
n
2
n
(VolA). Then there exists a nite set of disjoint balls B
1
, . . . , B
K
A
such that Vol(
K

1
B
j
) > . The volume of the complement C = A
K

1
B
j
satises
VolC = VolAVol
_
K
_
1
B
j
_
> VolA.
Hence there exists a partition T of P by cubes such that the total volume of cubes Q
1
, . . . , Q
L
con-
tained in C is > VolA. Let us inscribe in each of the cubes Q
j
a ball

B
j
. Then B
1
, . . . , B
K
,

B
1
, . . . ,

B
L
is an admissible set of balls in A. Indeed, all these balls are disjoint and contained in A. The total
volume of this admissible set is equal to
K

1
VolB
j
+
L

1
Vol

B
i
+

n
2
n
(VolA) > ,
in view of our choice of , but this contradicts to our assumption < VolA. Hence, we have
= VolA.
123
Lemma 9.26. Let A V be any measurable set in a Euclidean space V . Then for any linear
orthogonal transformation F : V V the set F(A) is also measurable and we have Vol(F(A)) =
Vol(A).
Proof. First note that if VolA = 0 then the claim follows from Lemma 9.19. Indeed, an orthogonal
transformation is, of course a smooth map.
Let now A be an arbitrary measurable set. Note that F(A) = F(A). Measurability of A
implies Vol(A) = 0. Hence, as we just have explained, Vol(F(A)) = Vol(F(A)) = 0, and hence
F(A) is measurable. According to Lemma 9.25 the volume of a measurable set can be computed
as a supremum of the total volume of disjoint inscribed balls. But the orthogonal transformation
F moves disjoint balls to disjoint balls of the same size, and hence VolA = VolF(A).
Next proposition shows that the volume of a parallelepiped can be computed by formula (3.3.1)
from Section 3.3.
Proposition 9.27. Let v
1
, . . . , v
n
V be linearly independent vectors. Then
Vol P(v
1
, . . . , v
n
) = [x
1
x
n
(v
1
. . . , v
n
)[. (9.5.6)
Proof. The formula (9.5.6) holds for rectangles, i.e. when v
j
= c
j
e
j
for some non-zero numbers c
j
,
j = 1, . . . n. Using Lemma 9.26 we conclude that it also holds for any orthogonal basis. Indeed, any
such basis can be moved by an orthogonal transformation to a basis of the above form c
j
e
j
, j =
1, . . . n. Lemma 9.26 ensures that the volume does not change under the orthogonal transformation,
while Proposition 2.17 implies the same about [x
1
x
n
(v
1
. . . , v
n
)[.
The Gram-Schmidt orthogonalization process shows that one can pass from any basis to an
orthogonal basis by a sequence of following elementary operations:
- reordering of basic vectors, and
shears, i.e. an addition to the last vector a linear combination of the other ones:
v
1
, . . . , v
n1
, v
n
v
1
, . . . , v
n1
, v
n
+
n1

j
v
j
.
Note that the reordering of vectors v
1
, . . . , v
n
changes neither VolP(v
1
, . . . , v
n
), nor the absolute
value
124
[x
1
x
n
(v
1
. . . , v
n
)[. On the other hand, a shear does not change
x
1
x
n
(v
1
. . . , v
n
).
It remains to be shown that a shear does not change the volume of a parallelepiped. We will consider
here only the case n = 2 and will leave to the reader the extension of the argument to the general
case.
Let v
1
, v
2
be two orthogonal vectors in R
2
. We can assume that v
1
= (a, 0), v
2
= (0, b) for
a, b > 0, because we already proved the invariance of volume under orthogonal transformations.
Let v

2
= v
2
+v
1
= (a

, b), where a

= a +b. Let us partition the rectangle P = P(v

1
, v
2
) into N
2
smaller rectangles P
i,j
, i, j = 0, . . . , N 1, of equal size. We number the rectangles in such a way
that the rst index corresponds to the rst coordinate, so that the rectangles P
00
, . . . , P
N1,0
form
the lower layer, P
01
, . . . , P
N1,1
the second layer, etc. Let us now shift the rectangles in k-th layer
horizontally by the vector (
kb
N
, 0). Then the total volume of the rectangles, denoted

P
ij
remains
the same, while when N the volume of part of the parallelogram P(v
1
, v

2
) that is not covered
by rectangles

P
i,j
, i, j = 0, . . . , N 1 converges to 0.

9.6 Fubinis Theorem

Let us consider R
n
as a direct product of R
k
and R
nk
for some k = 1, . . . , n 1. We will denote
coordinates in R
k
by x = (x
1
, . . . , x
k
) and coordinates in R
nk
by y = (y
1
, . . . , y
nk
), so the
coordinates in R
n
are denoted by (x
1
, . . . , x
k
, y
1
, . . . , y
nk
). Given rectangles P
1
R
k
and P
2

R
nk
their product P = P
1
P
2
is a rectangle in R
n
.
The following theorem provides us with a basic tool for computing multiple integrals.
Theorem 9.28 (Guido Fubini). Suppose that a function f : P R is integrable over P. Given a
point x P
1
let us dene a function f
x
: P
2
R by the formula f
x
(y) = f(x, y), y P
2
. Then
_
P
fdV
n
=
_
P
1
_
_
_
_
P
2
f
x
dV
nk
_
_
_dV
k
=
_
P
1
_
_
_
P
2
f
x
dV
nk
_
_
dV
k
.
125
Figure 9.2: Guido Fubini (1879-1943)
In particular, if the function f
x
is integrable for all (or almost all) x P
1
then one has
_
P
fdV
n
=
_
P
1
_
_
_
P
2
f
x
dV
nk
_
_
dV
k
.
Here by writing dV
k
, dV
nk
and dV
n
we emphasize the integration with respect to the k-, (nk)-
and n-dimensional volumes, respectively.
Proof. Choose any partition T
1
of P
1
and T
2
of P
2
. We will denote elements of the partition T
1
by P
j
1
and elements of the partition T
2
by P
i
2
. Then products of P
j,i
= P
j
1
P
i
2
form a partition T
of P = P
1
P
2
. Let us denote
I(x) :=
_
P
2
f
x
, I(x) :=
_
P
2
f
x
, x P
1
.
Let us show that
L(f, T) L(I, T
1
) U(I, T
1
) U(f, T). (9.6.1)
Indeed, we have
L(f, T) =

i
m
j,i
(f)Vol
n
P
j,i
.
Here the rst sum is taken over all multi-indices j of the partition T
1
, and the second sum is taken
126
over all multi-indices i of the partition T
2
. On the other hand,
L(I, T
1
) =

j
inf
xP
j
1
_
_
_
_
P
2
f
x
dV
nk
_
_
_Vol
k
P
j
1
.
Note that for every x P
j
1
we have
_
P
2
f
x
dV
nk
L(f
x
; T
2
) =

i
m
i
(f
x
)Vol
nk
(P
i
2
)

i
m
i,j
(f)Vol
nk
(P
i
2
),
and hence
inf
xP
j
1
_
P
2
f
x
dV
nk

i
m
i,j
(f)Vol
nk
(P
i
2
).
Therefore,
L(I, T
1
)

i
m
i,j
(f)Vol
nk
(P
i
2
)Vol
k
(P
j
1
) =

i
m
j,i
(f)Vol
n
(P
j,i
) = L(f, T).
Similarly, one can check that U(I, T
1
) U(f, T). Together with an obvious inequality L(I, T
1
)
U(I, T
1
) this completes the proof of (9.6.1). Thus we have
max(U(I, T
1
) L(I, T
1
), U(I, T
1
) L(I, T
1
)) U(I, T
1
) L(I, T
1
) U(f, T) L(f, T).
By assumption for appropriate choices of partitions, the right-hand side can be made < for any
a priori given > 0. This implies the integrability of the function I(x) and I(x) over P
1
. But then
we can write
_
P
1
I(x)dV
nk
= lim
(P
1
)0
L(I; T
1
)
and
_
P
1
I(x)dV
nk
= lim
(P
1
)0
U(I; T
1
).
We also have
lim
(P)0
L(f; T) = lim
(P)0
U(f; T) =
_
P
fdV
n
.
Hence, the inequality (9.6.1) implies that
_
P
fdV
n
=
_
P
1
_
_
_
_
P
2
f
x
dV
nk
_
_
_dV
k
=
_
P
1
_
_
_
P
2
f
x
dV
nk
_
_
dV
k
.

127
Corollary 9.29. Suppose f : P R is a continuous function. Then
_
P
f =
_
P
1
_
P
2
f
x
=
_
P
2
_
P
1
f
y
.
Thus if we switch back to the notation x
1
, . . . , x
n
for coordinates in R
n
, and if P = a
1
x
1

b
1
, . . . , a
n
x
n
b
n
then we can write
_
P
f =
bn
_
an
_
_
. . .
_
_
b
1
_
a
1
f(x
1
, . . . , x
n
)dx
1
_
_
. . .
_
_
dx
n
. (9.6.2)
The integral in the right-hand side of (9.6.2) is called an iterated integral. Note that the order
of integration is irrelevant there. In particular, for continuous functions one can change the order
of integration in the iterated integrals.

9.7 Integration of n-forms over domains in n-dimensional space

Dierential forms are much better suited to be integrated than functions. For integrating a function,
one needs a measure. To integrate a dierential form, one needs nothing except an orientation of
the domain of integration.
Let us start with the integration of a n-form over a domain in a n-dimensional space. Let be
a n-form on a domain U V, dimV = n.
Let us x now an orientation of the space V . Pick any coordinate system (x
1
. . . x
n
) that agrees
with the chosen orientation.
We proceed similar to the way we dened an integral of a function. Let us x a rectangle
P = P(a
1
, b
1
; a
2
, b
2
; . . . ; a
n
, b
n
) = a
i
x
i
b
i
; i = 1, . . . , n. Choose its partition T by N
n
smaller rectangles P
j
= t
1
jn
x
1
t
1
jn+1
, . . . , t
n
j
1
x
1
t
n
j
1
+1
, where j = (j
1
, . . . , j
n
) and each
index j
k
takes values between 0 and N 1. Let us x a point c
j
P
j
and denote by C the set of all
such c
j
. We also denote by t
j
the point with coordinates t
1
j
1
, . . . , t
n
jn
and by T
j,m
V
c
j
, m = 1, . . . , n
the vector t
j+1m
t
j
, parallel-transported to the point c
j
. Here we use the notation j + 1
m
for the
multi-index j
1
, . . . , j
m1
, j
m
+ 1, j
m+1
, . . . , j
n
. Thus the vector T
j,m
is parallel to the m-th basic
vector and has the length [t
jm+1
t
jm
[.
128
Given a dierential n-form on P we form an integral sum
I(; T, C) =

j
(T
j
1
, T
j
2
, . . . , T
j,n
), (9.7.1)
where the sum is taken over all elements of the partition. We call an n-form integrable if there
exists a limit lim
(P)0
I(; T, C) which we denote by
_
P
and call the integral of over P. Note
that if = f(x)dx
1
dx
n
then the integral sum I(, T, C) from (9.7.1) coincides with the
integral sum I(f; T, C) from (9.5.2) for the function f. Thus the integrability of is the same as
integrability of f and we have
_
P
f(x)dx
1
dx
n
=
_
P
fdV. (9.7.2)
Note, however, that the equality (9.7.2) holds only if the coordinate system (x
1
, . . . , x
n
) denes the
given orientation of the space V . The integral
_
P
f(x)dx
1
dx
n
changes its sign with a change
of the orientation while the integral
_
P
fdV is not sensitive to the orientation of the space V .
It is not clear from the above denition whether the integral of a dierential form depends on
our choice of the coordinate system. It turns out that it does not, as the following theorem, which
is the main result of this section, shows. Moreover, we will see that one even can use arbitrarty
curvilinear coordinates.
In what follows we use the convention introduced at the end of Section 6.1. Namely by a
dieomorphism between two closed subsets of vector spaces we mean a dieomorphism between
their neighborhoods.
Theorem 9.30. Let A, B R
n
be two measurable compact subsets. Let f : A B be an orientation
preserving dieomorphism. Let be a dierential n-form dened on B. Then if is integrable over
B then f

is integrable over A and we have

_
A
f

=
_
B
. (9.7.3)
For an orientation reversing dieomorphism f we have
_
A
f

=
_
B
.
Let = g(x)dx
1
dx
n
. Then f

= g f det Dfdx
1
dx
n
, and hence the formula
(9.7.3) can be rewritten as
_
P
g(x
1
, . . . , x
n
)dx
1
dx
n
=
_
P
g f det Df dx
1
dx
n
.
129
Here
det Df =

f
1
x
1
. . .
f
1
xn
. . . . . . . . .
fn
x
1
. . .
fn
xn

is the determinant of the Jacobian matrix of f = (f

1
, . . . , f
n
).
Hence, in view of formula (9.7.2) we get the following change of variables formula for multiple
integrals of functions.
Corollary 9.31. [Change of variables in a multiple integral] Let g : B R be an integrable
function and f : A B a dieomorphism. Then the function g f is also integrable and
_
B
gdV =
_
A
g f[ det Df[dV . (9.7.4)
We begin the proof with the following special case of Theorem 9.30.
Proposition 9.32. The statement of 9.30 holds when = dx
1
dx
n
and the set A is the unit
cube I = I
n
.
We will use below the following notation.
For any set A V and any positive number > 0 we denote by A the set x, x A. For
any linear operator F : V W between two Euclidean spaces V and W we dene its norm [[F[[
by the formula
[[F[[ = max
||v||=1
[[F(v)[[ = max
vV,v=0
[[F(v)[[
[[v[[
.
Equivalently, we can dene [[F[[ as follows. The linear map F maps the unit sphere in the space V
onto an ellipsoid in the space W. Then [[F[[ is the biggest semi-axis of this ellipsoid.
Let us begin by observing the following geometric fact:
Lemma 9.33. Let I = [x
i
[
1
2
, i = 1, . . . , n R
n
be the unit cube centered at 0 and F : R
n
R
n
a non-degenerate linear map. Take any (0, 1) and set =

||F
1
||
. Then for any boundary point
z I we have
B

(F(z)) (1 +)F(I) (1 )F(I), (9.7.5)

see Fig. 9.7
130
Figure 9.3: Image of a cube under a dieomorphism and its linearization
131
Proof. Inclusion (9.7.5) can be rewritten as
F
1
(B

(F(z)) (1 +)I (1 )I.

But the set F
1
(B

(F(z)) is an ellipsoid centered at z whose greatest semi-axis is equal to [[F

1
[[.
Hence, if [[F
1
[[ then F
1
(B

(F(z)) (1 +)I (1 )I.

Recall that we denote by I the cube I scaled with the coecient , i.e. I = [x
i
[

2
, i =
1, . . . , n R
n
. We will also need
Lemma 9.34. Let U R
n
is an open set, f : U R
n
such that f(0) = 0. Suppose that f is
dierentiable at 0 and its dierential F = d
0
f : R
n
R
n
at 0 is non-degenerate. Then for any
(0, 1) there exists > 0 such that
(1 )F(I) f(I) (1 +)F(I), (9.7.6)
see Fig. 9.7.
Proof. First, we note that inclusion (9.7.5) implies, using linearity of F, that for any > 0 we
have
B

(F(z)) (1 +)F(I) (1 )F(I), (9.7.7)

where z (I) and, as in Lemma 9.33, we assume that = [[F
1
[[.
According to the denition of dierentiability we have
f(h) = F(h) +o([[h[[).
Denote :=

n
=

n||F
1
||
. There exists > 0 such that if [[h[[ then
[[f(h) F(h)[[ [[h[[ .
Denote :=

n
. Then I B

(0), and hence [[f(z) F(z)[[ for any z I. In particular,

for any point z (I) we have
f(z) B
e
(F(z)) = B

ne
(F(z)) = B

(F(z)),
and therefore in view of (9.7.7)
f((I)) (1 +)F(I) (1 )F(I).
But this is equivalent to inclusion (9.7.6).
132
Lemma 9.35. Let F : R
n
R
n
be a non-degenerate orientation preserving linear map, P =
P(v
1
, . . . , v
n
) a parallelepiped, and = dx
1
dx
n
. Then
_
F(P)
=
_
P
F

. Here we assume that

the orientation of P and F(P) are given by the orientation of R
n
.
Proof. We have
_
F(P)
=
_
F(P)
dx
1
. . . dx
n
= VolF(P) = (det F)VolP. On the other hand, F

=
det F, and hence
_
P
F

= det F
_
P
= (det F)VolP.
Proof of Proposition 9.32. We have f

= (det Df)dx
1
dx
n
, and hence the form f

is
integrable because f is C
1
-smooth, and hence det Df is continuous.
Choose a partition T of the cube I by N
n
small cubes I
K
, K = 1, . . . , N
n
, of the same size
1
N
.
Let c
K
I
K
be the center of the cube I
K
. Then
_
I
f

=
N
n

K=1
_
I
K
f

.
Note that in view of the uniform continuity of the function det Df, for any > 0 the number N
can be chosen so large that [ det Df(x) det Df(x

)[ < for any two points x, x

I
K
and any
K = 1, . . . , N
n
. Let
K
be the form det Df(c
k
)dx
1
dx
n
on I
K
. Then

_
I
K
f

_
I
K

_
I
K
[det Df(x) det Df(c
K
)[ dx
1
dx
n
Vol(I
K
) =

N
n
.
Thus

_
I
f

N
n

K=1
_
I
K

. (9.7.8)
Next, let us analyze the integral
_
I
K

K
. Denote F
K
:= d
c
K
(f). We can assume without loss of
generality that c
K
= 0, and f(c
K
) = 0, and hence F
K
can be viewed just as a linear map R
n
R
n
.
Using Lemma 9.34 we have for a suciently large N
(1 )F
K
(I
K
) f(I
K
) (1 +)F
K
(I
K
).
Again in view of compactness of I the number can be chosen the same for all cubes I
K
. Hence
(1 )
n
Vol(F
K
(I
K
)) Volf(I
K
) (1 +)
n
Vol(F
K
(I
K
)). (9.7.9)
133
Note that Volf(I
K
) =
_
f(I
K
)
, and hence summing up inequality (9.7.9) over K we get
(1 )
n
N
n

K=1
Vol(F
K
(I
K
))
N
n

K=1
Volf(I
K
) =
N
n

K=1
_
f(I
K
)
=
_
f(I)
(1 +)
n
N
n

K=1
Vol(F
K
(I
K
)).
(9.7.10)
Note that
K
= F

K
and by Lemma 9.35 we have
_
I
K

K
=
_
I
K
F

K
=
_
F
K
(I
K
)
= Vol(F
K
(I
K
)).
Hence, it follows from (9.7.10) that
(1 )
n
N
n

K=1
_
I
K

K

_
f(I)
(1 +)
n
N
n

K=1
_
I
K

K
. (9.7.11)
Recall that from (9.7.8) we have
_
I
f

N
n

K=1
_
I
K

_
I
f

+.
Combining with (9.7.11) we get
(1 )
n
_
_
_
I
f

_
_

_
f(I)
(1 +)
n
_
_
_
I
f

+
_
_
. (9.7.12)
Passing to the limit when 0 we get
_
I
f

_
f(I)

_
I
f

, (9.7.13)
i.e.
_
I
f

=
_
f(I)
.
Proof of Theorem 9.30. Let us recall that the dieomorphism f is dened as a dieomorphism
between open neighborhoods U A and U

B. We also assume that the form is extended to U

as equal to 0 outside B. The form can be written as hdx

1
dx
n
. Let us take a partition T of
a rectangular containg U

by cubes I
j
of the same size . Consider forms
+
j
:= M
j
(h)dx
1
dx
n
and

j
:= m
j
(h)dx
1
dx
n
on I
j
, where m
j
(h) = inf
I
j
h, M
j
(h) = sup
I
j
(h). Let

be the
form on U

equal to

j
on each cube I
j
. The assumption of integrability of over B guarantees
that for any > 0 if is chosen small enough we have
_
B

+

_
B
. The forms f

are a.e.
continuous, and hence integrable over A and we have
_
A
f

_
A
f

_
A
f

+
. Hence, if we prove
that
_
A
f

=
_
B

then this will imply that is integrable and

_
A
=
_
B
.
134
On the other hand,
_
B

j
_
I
j

j
and
_
A
f

j
_
B
j
f

j
, where B
j
= f
1
(I
j
). But according
to Proposition 9.32 we have
_
B
j
f

j
=
_
I
j

j
, and hence
_
A
f

=
_
B

.
9.8 Manifolds and submanifolds
9.8.1 Manifolds
Manifolds of dimension n are spaces which are locally look like open subsets of R
n
but globally
could be much more complicated. We give a precise denition below.
Let U, U

R
n
be open sets. A map f : U U

is called a homeomorpism if it is continuous

one-to-one map which has a continuous inverse f
1
: U

U.
A map f : U U

is called a C
k
-dieomorpism, k = 1, . . . , , if it is C
k
-smooth, one-to-one
map which has a C
k
-smooth inverse f
1
: U

U. Usually we will omit the reference to the class

of smoothness, and just call f a dieomorphism, unless it will be important to emphasize the class
of smoothness.
A set M is called an n-dimensional C
k
-smooth (resp. topological) manifold if there exist subsets
U

X, , where is a nite or countable set of indices, and for every a map

: U

R
n
such that
M1. M =

.
M2. The image G

(U

) is an open set in R
n
.
M3. The map

viewed as a map U

is one-to-one.
M4. For any two sets U

, U

, , the images

(U

),

(U

) R
n
are open and
the map
h

:=

(U

(U

) R
n
is a C
k
-dieomorphism (resp. homeomorphism).
Sets U

are called coordinate neighborhoods and maps

: U

R
n
are called coordinate maps.
The pairs (U

) are also called local coordinate charts. The maps h

are called transiton maps

135
between dierent coordinate charts. The inverse maps

=
1

: G

are called (local)

parameterization maps. An atlas is a collection A = U

of all coordinate charts.

One says that two atlases A = U

and A

= U

on the same manifold

X are equivalent, or that they dene the same smooth structure on X if their union A A

=
(U

), (U

)
,
is again an atlas on X. In other words, two atlases dene the same
smooth structure if transition maps from local coordinates in one of the atlases to the local coor-
dinates in the other one are given by smooth functions.
A subset G M is called open if for every the image

(G U

) R
n
is open. In
particular, coordinate charts U

themselves are open, and we can equivalently say that a set G is

open if its intersection with every coordinate chart is open. By a neighborhood of a point a M we
will mean any open subset U M such that a U.
Given two smooth manifolds M and

M of dimension m and n then a map f : M

M is
called continuous if if for every point a M there exist local coordinate charts (U

) in M and
(

) in

M

, such that a U

, f(U

)

U

and the composition map

G

(U

R
n
is continuous.
Similarly, for k = 1, . . . , a map f : M

M is called C
k
-smooth if for every point a M
there exist local coordinate charts (U

) in M and (

) in

M

, such that a U

, f(U

)

U

and the composition map

G

(U

R
n
is C
k
-smooth. In other words, a map is continuous or smooth, if it is continuous or smooth when
expressed in local coordinates.
A map f : M N is called a dieomorphism if it is smooth, one-to-one, and the inverse map
is also smooth. One-to-one continuous maps with continuous inverses are called homeomorphisms.
Note that in view of the chain rule the C
k
-smoothness is independent of the choice of local
coordinate charts (U

) and (

). Note that for C

k
-smooth manifolds one can talk only
about C
l
-smooth maps for l k. For topological manifolds one can talk only about continuous
maps.
136
If one replaces condition M2 in the denition of a manifold by
M2
b
. The image G

(U

) is either an open set in R

n
or an intersection of an open set in R
n
with R
n
+
= x
1
0
then one gets a denition of a manifold with boundary.
A slightly awkward nuance in the above denition is that a manifold with boundary is not a
manifold! It would be, probably, less confusing to write this as a 1 word manifold-with-boundary,
but of course nobody does that.
The points of a manifold M with boundary which are mapped by coordinate maps

to points
in R
n1
= R
n
+
are called the boundary points of M. The set of boundary points is called the
boundary of M and denoted by M. It is itself a manifold of dimension n 1.
Note that any (interior) point a of an n-dimensional manifold M has a neighborhood B dieo-
morphic to an open ball B
1
(0) R
n
, while any boundary point has a neighborhood dieomorphic
to a semi-ball B
1
(0) x
1
0 R
n
.
Exercise 9.36. Prove that a boundary point does not have a neighborhood dieomorphic to an
open ball. In other words, the notion of boundary and interior point of a manifold with boundary
are well dened.
Next we want to introduce a notion of compactness for subsets in a manifold. Let us recall that
for subsets in a Euclidean vector space we introduced three equivalent denition of compactness,
see Section 6.1. The rst denition, COMP1 is unapplicable because we cannot talk about bounded
sets in a manifold. However, denitions COMP2 and COMP3 make perfect sense in an arbitrary
manifold. For instance, we can say that a subset A M is compact if from any innite sequence
of points in A one can choose a subsequence converging to a point in A.
A compact manifold (without boundary) is called closed. Note that the word closed is used
here in a dierent sense than a closed set. For instance, a closed interval is not a closed manifold
because it has a boundary. An open interval or a real line R is not a closed manifold because it is
not compact. On the other hand, a circle, or a sphere S
n
of any dimension n is a closed manifold.
The notions of connected and path connected subsets of a manifold are dened in the same way
as in an Euclidean space.
137
9.8.2 Gluing construction
The construction which is described in this section is called gluing or quotient construction. It
provides a rich source of examples of manifolds. We discuss here only very special cases of this
construction.
a) Let M be a manifold and U, U

its two open disjoint subsets. Let us moreover assume that

each point x M has a neighborhood which does not intersect at least one of the sets U and U

.
4
Consider a dieomorphism f : U U

.
Let us denote by M/f(x) x the set obtained from M by identifying each point x U with
its image f(x) U

. In other words, a point of M/f(x) x is either a point from x M(UU

),
or a pair of points (x, f(x)), where x U. Note that there exists a canonical projection : M
M/f(x) x. Namely (x) = x if x / U U

, (x) = (x, f(x)) if x U and (x) = (f

1
(x), x)
if x U

. By our assumption each point x M has a coordinate neighborhood G

x
x such that
f(G
x
U) G
x
= . In particular, the projection [
Gx
: G
x


G
x
= (G
x
) is one-to-one. We will
declare by denition that

G
x
is a coordinate neighborhood of (x) M/f(x) x and dene a
coordinate map

:

G
x
R
n
by the formula

=
1
. It is not dicult to check that this
construction dene a structure of an n-dimensional manifold on the set M/f(x) x. We will
call the resulted manifold the quotient manifold of M, or say that M/f(x) x is obtained from
M by gluing U with U

with the dieomorphism f.

Though the described above gluing construction always produce a manifold, the result could be
quite pathological, if no additional care is taken. Here is an example of such pathology.
Example 9.37. Let M = I I

be the union of two disjoint open intervals I = (0, 2) and I

= (3, 5).
Then M is a 1-dimensional manifold. Denote U := (0, 1) I, U

:= (3, 4) I

. Consider a
dieomorphism f : U U

given by the formula f(t) = t + 3, t U. Let

M = M/f(x) x be
the corresponding quotient manifold. In other words,

M is the result of gluing the intervals I and
I

along their open sub-intervals U and U

. Note that the points 1 I and 4 I

are not identied,

but 1 , 4 are identied for an arbitrary small > 0. This means that any neighborhood of 1
and any neighborhood of 4 have non-empty intersection.
4
Here is an example when this condition is not satised: M = (0, 2), U = (0, 1), U

= (1, 2). In this case any

neighborhood of the point 1 intersect both sets, U and U

.
138
Figure 9.4: Felix Hausdor (1868-1942)
In order to avoid such pathological examples one usually (but not always) requires that manifolds
satisfy an additional axiom, called Hausdor property:
M5. Any two distinct points x, y M have non-intersecting neighborhoods U x, G y.
In what follows we always assume that the manifolds satisfy the Hausdor property M5.
Let us make the following general remark about dieomorphisms f : (a, b) (c, d) between
two open intervals. Such dieomorphism is simply a dierentiable function whose derivative never
vanishes and whose range is equal to the interval (c, d). If derivative is positive then the dieomor-
phism is orientation preserving, and it is orientation reversing otherwise. The function f always
extends to a continuous (but necessarily dierentiable function f : [a, b] [c, d] such that f(a) = c
and f(b) = d in the orientation preserving case, and f(a) = d f(b) = c in the orientation reversing
case.
Lemma 9.38. Given a, b, a

(0, 1) such that a < b and a

< b

consider an orientation preserving

dieomorphisms f : (0, a) (0, a

) and (b, 1) (b

1). Then for any a (0, a) and

b (b, 1) there
exists a dieomorphism F : (0, 1) (0, 1) which coincides with f on (0, a) and coincides with g on
(

b, 1).
Proof. Choose real numbers c, c,

d, d such that a < c < c <

d < d < b. Consider a cut-o C

-
function : (0, 1) (0, 1) which is equal to 1 on (0, a] [c,

d] [

b, 1) and equal to 0 on [a, c] [d, b].

139
For positive numbers > 0 and C > 0 (which we will choose later) consider a function h
,C
on
(0, 1) dened by the formula
h
,C
(x) =
_

_
(1 )(x)f

(x) +, x (0, a);

, x [a, c] [d, b];
(1 )C(x) +, x (c, d);
(1 )(x)g

(x) +, x (b, 1).

Note that h
,C
(x) = f

(x) on (0, a], h

,C
(x) = g

(x) on [

b, 1) and equal to C on [c,

d]. Dene the

function F
,C
: (0, 1) (0, 1) by the formula
F
,C
(x) =
x
_
0
h(u)du.
Note that the derivative F

,C
is positive, and hence the function F
,C
is strictly increasing. It
coincides with f on (0, a] and coincides up to a constant with g on (

b, 1). Note that when and

C are small we have F
,C
(

b) < b

< g(

b), and lim

C
F
,C
(

b) = . Hence, by continuity one can

choose , C > 0 in such a way that F
,C
(

b) = g(

b). Then the function F = F

,C
is a dieomorphism
(0, 1) (0, 1) with the required properties.
Lemma 9.39. Suppose that a 1-dimensional manifold M (which satises the Hausdor axiom M5)
is covered by two coordinate charts, M = U U

, with coordinate maps : U (0, 1),

: U

(0, 1) such that (U U

) = (0, a) (b, 1),

(U U

) = (0, a

) (b

1) for some a, a

, b, b

(0, 1)
with a < b, a

< b

. Then M is dieomorphic to the circle S

1
.
Proof. Denote by and

the parameterization maps

1
and (

)
1
, and set G := (U U

)
and G

:=

(U U

). Let h =

: G G

be the transition dieomorphism. There could

be two cases: h((0, a)) = (0, a

), h((b, 1)) = (b

, 1) and h((0, a)) = (b

, 1), h((b, 1)) = (0, a). We will

analyze the rst case. The second one is similar.
Let h be the continuous extension of h to [0, a] [b, 1]. We claim that h(0) = a

, h(a) = 0,
h(b) = 1 and h(1) = b

. Indeed, assuming otherwise we come to a contradiction with the Hausdor

property M5. Indeed, suppose h(a) = a

. Note the points A := (a), A

:=

(a

) M are disjoint.
On the other hand, for any neighborhood A its image () I contains an interval (a , a),
140
and similarly for any neighborhood

its image

() I contains an interval (a

, a

). for
a suciently small . But h((a , a)) = (a

, a

) for some

> 0 and hence

((a

, a

)) =

h((a, a )) =

((a, a )) = ((a, a )) ,
i.e.

,= . In other words, any neighborhoods of the distinct points A, A

M intersect, which
violates axiom M5. Similarly we can check that h(b) = b

.
Now take the unit circle S
1
R
2
and consider the polar coordinate on S
1
. Let us dene a
map g

: U

S
1
by the formula =

(x). Thus g

is a dieomorphism of U

onto an arc
of S
1
given in polar coordinates by < < 0. The points A

(a

) and B

(b

) are
mapped to points with polar coordinates = a

and = b

. On the intersection U U

we can describe the map g

in terms of the coordinate in U. Thus we get a map f := g

:
(0, a) (b, 1) S
1
. We have f(0) = g

(A

), f(a) = g

(0), f(1) = g

(B

), f(b) = g

(1). Here we
denoted by f the continuous extension of f to [0, a] [b, 1]. Thus f((0, a)) = a

< < 0 and

f((b, 1) = < < 3 b

. Note that the dieomorphism f is orientation preserving assuming

that the circle is oriented counter-clockwise. Using Lemma 9.38 we can nd a dieomorphism F
from (0, 1) to the arc a

< < 3 b

S
1
which coincides with f on (0, a) (

b, 1) for
any a (0, a) and

b (b, 1). Denote a

:= h(a),

= h(

b). Notice that the neighborhoods U and

((a,

b)) cover M. Hence, the required dieomorphism

F : M S
1
we can dene by the
formula

F(x) =
_

_
g(x), x

U

;
F (x), x U.

Similarly (and even simpler), one can prove

Lemma 9.40. Suppose that a 1-dimensional manifold M (which satises the Hausdor axiom M5)
is covered by two coordinate charts, M = U U

, with coordinate maps : U (0, 1),

: U

(0, 1) such that U U

is connected. Then M is dieomorphic to the open interval (0, 1).

Theorem 9.41. Any (Hausdor) connected closed 1-dimensional manifold is dieomorphic to the
circle S
1
.
141
Exercise 9.42. Show that the statement of the above theorem is not true without the axiom M5,
i.e. the assumption that the manifold has the Hausdor property.
Proof. Let M be a connected closed 1-dimensional manifold. Each point x M has a coordinate
neighborhood U
x
dieomorphic to an open interval. All open intervals are dieomorphic, and hence
we can assume that each neighborhood G
x
is parameterized by the interval I = (0, 1). Let
x
: I
G
x
be the corresponding parameterization map. We have

xM
U
x
= M, and due to compactness of
M we can choose nitely many U
x
1
, . . . , U
x
k
such that
k

i=1
U
x
i
= M. We can further assume that
none of these neighborhoods is completrely contained inside another one. Denote U
1
:= U
x
1
,
1
:=

x
1
. Note that U
1

k

2
U
x
k
,= . Indeed, if this were the case then due to connectedness of M we
would have
k

2
U
x
i
= and hence M = U
1
, but this is impossible because M is compact. Thus,
there exists i = 2, . . . , k such that U
x
i
U
1
,= . We set U
2
:= U
x
i
,
2
=
x
i
. Consider open sets
G
1,2
:=
1
1
(U
1
U
2
), G
2,1
=
1
2
(U
1
U
2
) I. The transition map h
1,2
:=
1
2

1
[
G
1,2
: G
1,2

G
2,1
is a dieomorphism.
Let us show that the set G
1,2
(and hence G
2,1
) cannot contain more that two connected com-
ponents. Indeed, in that case one of the components of G
1,2
has to be a subinterval I

= (a, b)
I = (0, 1) where 0 < a < b < 1. Denote I

:= h
1,2
(I

). Then at least of of the boundary values

of the transition dieomorphism h
1,2
[
I
, say h
1,2
(a), which is one of the end points of I

, has to
be an interior point c I = (0, 1). We will assume for determinacy that I

= (c, d) I. But this

contradicts the Hausdor property M5. The argument repeats a similar argument in the proof of
Lemma 9.39.
Indeed, note that
1
(a) ,=
2
(c). Indeed,
1
(a) belongs to U
1
U
2
and
2
(c) is in U
2
U
1
.
Take any neighborhood
1
(a) in M. Then
1
() is an open subset of I which contains the
point a. Hence
1
((a, a + )) , and similarly, for any neighborhood

2
(c) in M we have

2
((c, c+))

for a suciently small > 0. But

1
((a, a+)) =
2
(h
1,2
((a, a

))) =
2
(c, c+

),
where c +

= h
1,2
(a+). Hence

,= , i.e. any two neighborhoods of two distict points

1
(a)
and
2
(c) have a non-empty intersection, which violates the Hausdor axiom M5.
If G
1,2
(0, 1) consists of two components then the above argument shows that each of these
components must be adjacent to one of the ends of the interval I, and the same is true about the
142
components of the set G
2,1
I. Hence, we can apply Lemma 9.39 to conclude that the union U
1
U
2
is dieomorphic to S
1
. We also notice that in this case all the remaining neighborhoods U
x
j
must
contain in U
1
U
2
. Indeed, each U
x
i
which intersects the circle U
1
U
2
must be completely contained
in it, because otherwise we would again get a contradiction with the Hausdor property. Hence,
we can eleiminate all neighborhoods which intersect U
1
U
2
. But then no other neighborhoods
could be left because otherwise we would have M = (U
1
U
2
)

Ux
j
(U
1
U
2
)=
U
x
j
, i.e. the manifold
M could be presented as a union of two disjoint non-empty open sets which is impossible due to
connectedness of M. Thus we conclude that in this case M = U
1
U
2
is dieomorphic to S
1
.
Finally in the case when G
1,2
consists of 1 component, i.e. when it is connected, one can Use
Lemma 9.40 to show that U
1
U
2
is dieomorphic to an open interval. Hence, we get a covering of
M by k 1 neighborhood dieomorphic to S
1
. Continuing inductively this process we will either
nd at some step two neighborhoods which intersect each other along two components, or continue
to reduce the number of neighborhoods. However, at some moment the rst situation should occur
because otherwise we would get that M is dieomorphic to an interval which is impossible because
by assumption M is compact.
b) Let M be a manifold, f : M M be a dieomorphism. Suppose that f satises the fol-
lowing property: There exists a positive integer p such that for any point x M we have f
p
(x) =
f f f
. .
p
(x) = x, but the points x, f(x) . . . , f
p1
(x) are all disjoint. The set x, f(x) . . . , f
p1
(x)
M is called the trajectory of the point x under the action of f. It is clear that trajectory of two
dierent points either coincide or disjoint. Then one can consider the quotient space X/f, whose
points are trajectories of points of M under the action of f. Similarly to how it was done in a) one
can dene on M/f a structure of an n-dimensional manifold.
c) Here is a version of the construction in a) for the case when trajectory of points are innite.
Let f : M M be a dieomorphism which satises the following property: for each point x M
there exists a neighborhood U
x
x such that all sets
. . . , f
2
(U
x
), f
1
(U
x
), U
x
, f(U
x
), f
2
(U
x
), . . .
are mutually disjoint. In this case the trajectory . . . , f
2
(x), f
1
(x), x, f(x), f
2
(x), . . . of each
point is innite. As in the case b), the trajectories of two dierent points either coincide or disjoint.
143
The set M/f of all trajectories can again be endowed with a structure of a manifold of the same
dimension as M.
9.8.3 Examples of manifolds
1. n-dimensional sphere S
n
. Consider the unit sphere S
n
= [[x[[ =

n+1

1
x
2
j
= 1 R
n+1
. Let
introduce on S
n
the structure of an n-dimensional manifold. Let N = (0, . . . , 1) and S = (0, . . . , 1)
be the North and South poles of S
n
, respectively.
Denote U

= S
n
S, U
+
= S
n
N and consider the maps p

: U

R
n
given by the formula
p

(x
1
, . . . , x
n+1
) =
1
1 x
n+1
(x
1
, . . . , x
n
). (9.8.1)
The maps p
+
: U
+
R
n
and p

: U

R
n
are called stereographic projections from the North
and South poles, respectively. It is easy to see that stereographic projections are one-to one maps.
Note that U
+
U

= S
n
S, N and both images, p
+
(U
+
U

) and p

(U
+
U

) coincide with
R
n
0. The map p

p
1
+
: R
n
0 R
n
0 is given by the formula
p

p
1
+
(x) =
x
[[x[[
2
, (9.8.2)
and therefore it is a dieomorphism R
n
0 R
n
0.
Thus, the atlas which consists of two coordinate charts (U
+
, p
+
) and (U

, p

) denes on S
n
a
structure of an n-dimensional manifold. One can equivalently denes the manifold S
n
as follows.
Take two disjoint copies of R
n
, let call them R
n
1
and R
n
2
. Denote M = R
n
1
R
n
2
, U = R
n
1
0 and
U

= R
n
2
0. Let f : U U

be a difeomorphism dened by the formula f(x) =

x
||x||
2
, as in (9.8.2).
Then S
n
can be equivalently described as the quotient manifold M/f.
Note that the 1-dimensional sphere is the circle S
1
. It can be d as follows. Consider the map
T : R R given by the formula T(x) = x + 1, x R. It satises the condition from 9.8.2 and
hence, one can dene the manifold R/T. This manifold is dieomorphic to S
1
.
2. Real projective space. The real projective space RP
n
is the set of all lines in R
n+1
passing
through the origin. One introduces on RP
n
a structure of an n-dimensional manifold as follows.
For each j = 1, . . . , n + 1 let us denote by U
j
the set of lines which are not parallel to the ane
subspace
j
= x
j
= 1. Clearly,
n+1

1
U
j
= RP
n
. There is a natural one-to one map
j
: U
j

j
144
which associates with each line U
j
the unique intersection point of with
j
. Furthermore,
each
j
can be identied with R
n
, and hence pairs (U
j
,
j
), j = 1, . . . , n + 1 can be chosen as an
atlas of coordinate charts. We leave it to the reader to check that this atlas indeed dene on RP
n
a structure of a smooth manifold, i.e. that the transition maps between dierent coordinate charts
are smooth.
Exercise 9.43. Let us view S
n
as the unit sphere in R
n+1
. Consider a map p : S
n
RP
n
which
associates to a point of S
n
the line passing through this point and the origin. Prove that this
two-to-one map is smooth, and moreover a local dieomorphism, i.e. that the restriction of p to
a suciently small neighborhood of each point is a dieomorphism. Use it to show that RP
n
is
dieomorphic to the quotient space S
n
/f, where f : S
n
S
n
is the antipodal map f(x) = x.
3. Products of manifolds and n-dimensional tori. Given two manifolds, M and N of
dimension m and n, respectively, one can naturally endow the direct product
M N = (x, y); x M, y N
with a structure of a manifold of dimension m+n. Let (U

and (V

be atlases
for M and N, so that

: U

R
m
,

: V

R
n
are dieomorphisms on open
subsets of R
m
and R
n
. Then the smooth structure on M N can be dened by an atlas
(U

)
,
,
where we denote by

: U

R
m
R
n
= R
m+n
are dieomorphisms dened
by the formula (x, y) (

(x)

(y)) for x U

and y V

.
One can similarly dene the direct product of any nite number of smooth manifolds. In par-
ticular the n-dimensional torus T
n
is dened as the product of n circles: T
n
= S
1
S
1
. .
n
. Let
us recall, that the circle S
1
is dieomorphic to R/T, i.e. a point of S
1
is a real number up to adding
any integer. Hence, the points of the torus T
n
can viewed as the points of R
n
up to adding any
vector with all integer coordinates.
145
9.8.4 Submanifolds of an n-dimensional vector space
Let V be an n-dimensional vector space. A subset A V is called a k-dimensional submanifold
of V , or simply a k-submanifold of V , 0 k n, if for any points a A there exists a local
coordinate chart (U
a
, u = (u
1
, . . . , u
n
) R
n
) such that u(a) = 0 (i.e. the point a is the origin in
this coordinate system) and
A U
a
= u = (u
1
, . . . , u
n
) U
a
; u
k+1
= = u
n
= 0. (9.8.3)
We will always assume the local coordinates at least as smooth as necessary for our purposes (but
at least C
1
-smooth), but more precisely, one can talk of C
m
- submanifolds if the implied coordinate
systems are at least C
m
-smooth.
Note that in the above we can replace the vector space V by any n-dimensional manifold, and
thus will get a notion of a k-dimesional submanifold of an n-dimensional manifold V .
Example 9.44. Suppose a subset A U V is given by equations F
1
= = F
nk
= 0 for
some C
m
-smooth functions F
1
, . . . , F
nk
on U. Suppose that for any point a A the dierentials
d
a
F
1
, . . . , d
a
F
nk
V

a
are linearly independent. Then A U is a C
m
-smooth submanifold.
Indeed, for each a A one can choose a linear functions l
1
, . . . , l
k
V

a
such that together
with d
a
F
1
, . . . , d
a
F
nk
V

a
they form a basis of V

. Consider functions L
1
, . . . , L
k
: V R,
dened by L
j
(x) = l
j
(x a) so that d
a
(L
j
) = l
j
, j = 1, . . . , k. Then the Jacobian det D
a
F of the
map F : (L
1
, . . . , L
k
, F
1
, . . . , F
nk
) : U R
n
does not vanish at a, and hence the inverse function
theorem implies that this map is invertible in a smaller neighborhood U
a
U of the point a A.
Hence, the functions u
1
= L
1
, . . . , u
k
= L
k
, u
k+1
= F
1
, . . . , u
n
= F
nk
can be chosen as a local
coordinate system in U
a
, and thus A U
a
= u
k+1
= = u
n
= 0.
Note that the map u

= (u
1
, . . . , u
k
) maps U
A
a
= U
a
A onto an open neighborhood

U =
u(U
a
) R
k
of the origin in R
k
R
n
, and therefore u

= (u
1
, . . . u
k
) denes a local coordinates,
so that the pair (U
A
a
, u

) is a coordinate chart. The restriction

= [
e
U
of the parameterization
map = u
1
maps

U onto U
A
a
. Thus

a parameterization map for the neighborhood U
A
a
. The
atlas (U
A
a
, u

)a A denes on a a structure of a k-dimensional manifold. The complementary

dimension n k is called the codimension of the submanifold A. We will denote dimension and
codimension of A by dimA and codimA, respectively.
146
As we already mentioned above in Section 9.3 1-dimensional submanifolds are usually called
curves. We will also call 2-dimensional submanifolds surfaces and codimension 1 submanifolds hyper-
surfaces. Sometimes k-dimensional submanifolds are called k-surfaces. Submanifolds of codimension
0 are open domains in V .
An important class form graphical k-submanifolds. Let us recall that given a map f : B R
nk
,
where B is a subset B R
k
, then graph is the set

f
= (x, y) R
k
R
nk
= R
n
; x B, y = f(x).
A (C
m
)-submanifold A V is called graphical with respect to a splitting : R
k
R
nk
V ,
if there exist an open set U R
k
and a (C
m
)-smooth map f : U R
nk
such that
A = (
f
).
In other words, A is graphical if there exists a coordinate system in V such that
A = x = (x
1
, . . . , x
n
); (x
1
, . . . x
k
) U, x
j
= f
j
(x
1
, . . . , x
k
), j = k + 1, . . . , n.
for some open set U R
k
and smooth functions, f
k+1
, . . . , f
n
: U R.
For a graphical submanifold there is a global coordinate system given by the projection of the
submanifold to R
k
.
It turns out that that any submanifold locally is graphical.
Proposition 9.45. Let A V be a submanifold. Then for any point a A there is a neighborhood
U
a
a such that U
a
A is graphical with respect to a splitting of V . (The splitting may depend on
the point a A).
We leave it to the reader to prove this proposition using the implicit function theorem.
One can generalize the discussion in this section and dene submanifolds of any manifold M,
and not just the vector space V . In fact, the denition (9.8.3) can be used without any changes to
dene submanifolds of an arbitrary smooth manifold.
A map f : M Q is called an embedding of a manifold M into another manifold Q if it is a
dieomorphism of M onto a submanifold A Q. In other words, f is an embedding if the image
A = f(M) is a submanifold of Q and the map f viewed as a map M A is a dieomorphism.
147
One can prove that any n-dimensional manifold can be embedded into R
N
with a suciently large
N (in fact N = 2n + 1 is always sucient).
Hence, one can think of manifold as submanifold of some R
n
given up to a dieomorphism, i.e.
ignoring how this submanifold is embedded in the ambient space.
In the exposition below we mostly restrict our discussion to submanifolds of R
n
rather than
general abstract manifolds.
9.8.5 Submanifolds with boundary
A slightly dierent notion is of a submanifold with boundary. A subset A V is called a k-
dimensional submanifold with boundary, or simply a k-submanifold of V with boundary, 0 k < n,
if for any points a A there is a neighborhood U
a
a in V and local (curvi-linear) coordinates
(u
1
, . . . , u
n
) in U
a
with the origin at a if one of two conditions is satised: condition (9.8.3), or the
following condition
A U
a
= u = (u
1
, . . . , u
n
) U
a
; u
1
0, u
k+1
= = u
n
= 0. (9.8.4)
In the latter case the point a is called a boundary point of A, and the set of all boundary points is
called the boundary of A and is denoted by A.
As in the case of submanifolds without boundary, any submanifold with boundary has a struc-
ture of a manifold with boundary.
Exercise 9.46. Prove that if A is k-submanifold with boundary then A is a (k 1)-dimensional
submanifold (without boundary).
Remark 9.47. 1. As we already pointed out when we discussed manifolds with boundary, a
submanifold with boundary is not a submanifold!
2. As it was already pointed out when we discussed 1-dimensional submanifolds with boundary,
the boundary of a k-submanifold with boundary is not the same as its set-theoretic boundary,
though traditionally the same notation A is used. Usually this should be clear from the
context, what the notation A stands for in each concrete case. We will explicitly point this
dierence out when it could be confusing.
148
Figure 9.5: The parameterization introducing local coordinates near an interior point a and a
boundary point b.
A compact manifold (without boundary) is called closed. The boundary of any compact manifold
with boundary is closed, i.e. (A) = .
Example 9.48. An open ball B
n
r
= B
n
R
(0) =
n

1
x
2
j
< 1 R
n
is a codimension 0 submanifold,
A closed ball D
n
r
= D
n
R
(0) =
n

1
x
2
j
1 R
n
ia codimension 0 submanifold with boundary.
Its boundary D
n
R
is an (n 1)-dimensional sphere S
n1
R
=
n

1
x
2
j
= 1 R
n
. It is a closed
hypersurface. For k = 0, 1 . . . n1 let us denote by L
k
the subspace L
k
= x
k+1
= = x
n
= 0
R
n
. Then the intersections
B
k
R
= B
n
R
L
k
, D
k
R
= D
n
R
L
k
, and S
k1
R
= S
n1
R
L
k
R
n
are, respectively a k-dimensional submanifold, a k-dimensional submanifold with boundary and a
closed (k 1)-dimensional submanifold of R
n
. Among all above examples there is only one (which
one?) for which the manifold boundary is the same as the set-theoretic boundary.
A neighborhood of a boundary point a A can be always locally parameterized by the semi-
149
open upper-half ball
B
+
(0) = x = (x
1
, . . . , x
k
) R
k
; x
1
0,
n

1
x
2
j
< 1.
We will nish this section by dening submanifolds with piece-wise smooth boundary. A subset
A V is called a k-dimensional submanifold of V with piecewise smooth boundary or with boundary
with corners, 0 k < n, if for any points a A there is a neighborhood U
a
a in V and local
(curvi-linear) coordinates (u
1
, . . . , u
n
) in U
a
with the origin at a if one of three condittions satised:
conditions (9.8.3), (9.8.4) or the following condition
A U
a
= u = (u
1
, . . . , u
n
) U
a
; l
1
(u) 0, . . . , l
m
(u) 0, u
k+1
= = u
n
= 0, (9.8.5)
where m > 1 and l
1
, . . . , l
m
(R
k
)

are linear functions In the latter case the point a is called a

corner point of A.
Note that the system of linear inequalities l
1
(u) 0, . . . , l
m
(u) 0 denes a convex cone in
R
k
. Hence, near a corner point of its boundary the manifold is dieomorphic to a convex cone.
Thus convex polyhedra and their dieomorphic images are important examples of submanifolds
with boundary with corners.
9.9 Tangent spaces and dierential
Suppose we are given two local parameterizations : G A and

:

G A. Suppose that
0 G

G and (0) =

(0) = a A. Then there exists a neighborhood U a in A such that
U (G)

G).
Denote G
1
:=
1
(U),

G
1
:=

1
(

U). Then one has two coordinate charts on U: u = (u

1
, . . . , u
k
) =
([
G
1
)
1
: U G
1
, and u = ( u
1
, . . . , u
k
) =
_

[
e
G
1
_
1
: U

G
1
.
Denote h := u

[
e
G
1
=

G
1
G
1
. We have

= u

= h,
and hence the dierentials d
0
and d

0
of parameterizations and

at the origin map R
k
0
isomor-
phically onto the same k-dimensional linear subspace T V
a
. Indeed, d
0

= d
0
d
0
h. Thus the
space T = d
0
(R
k
0
) V
a
is independent of the choice of parameterization. It is called the tangent
150
space to the submanifold A at the point a A and will be denoted by T
a
A. If A is a submanifold
with boundary and a A then there are dened both the k-dimensional tangent space T
a
A and
its (k 1)-dimensional subspace T
a
(A) T
a
A tangent to the boundary.
Example 9.49. 1. Suppose a submanifold A V is globally parameterized by an embedding
: G A V , G R
k
. Suppose the coordinates in R
k
are denoted by (u
1
, . . . , u
k
). Then
the tangent space T
a
A at a point a = (b), b G is equal to the span
Span
_

u
1
(a), . . . ,

u
k
(a)
_
.
2. In particular, suppose a submanifold A is graphical and given by equations
x
k+1
= g
1
(x
1
, . . . , x
k
), . . . , x
n
= g
nk
(x
1
, . . . , x
k
), (x
1
, . . . , x
k
) G R
k
.
Take points b = (b
1
, . . . b
k
) G and a = (b
1
, . . . , b
k
, g
1
(b), . . . , g
nk
(b)) A. Then T
a
A =
Span(T
1
, . . . T
k
) , where
T
1
=
_
_
1, 0, . . . , 0
. .
k
,
g
1
x
1
(b), . . . ,
g
nk
x
1
(b)
_
_
,
T
1
=
_
_
0, 1, . . . , 0
. .
k
,
g
1
x
2
(b), . . . ,
g
nk
x
2
(b)
_
_
,
. . .
T
1
=
_
_
0, 0, . . . , 1
. .
k
,
g
1
x
k
(b), . . . ,
g
nk
x
k
(b)
_
_
.
3. Suppose a hypersurface R
n
is given by an equation = F = 0 for a smooth function
F dened on an neighborhood of and such that d
a
F ,= 0 for any a . In other words,
the function F has no critical points on . Take a point a . Then T
a
R
n
a
is given by a
linear equation
n

1
F
x
j
(a)h
j
= 0, h = (h
1
, . . . , h
n
) R
n
a
.
Note that sometimes one is interested to dene T
a
as an ane subspace of R
n
= R
n
0
and
not as a linear subspace of R
n
a
. We get the required equation by shifting the origin:
T
a
= x = (x
1
, . . . , x
n
) R
n
;
n

1
F
x
j
(a)(x
j
a
j
) = 0.
151
If for some parameterization : G A with (0) = a the composition f is dierentiable at
0, and the linear map
d
0
(f ) (d
0
)
1
: T
a
A W
f(a)
is called the dierential of f at the point a and denoted, as usual, by d
a
f. Similarly one can dene
C
m
-smooth maps A W.
Exercise 9.50. Show that a map f : A W is dierentiable at a point a A i for some
neighborhood U of a in V there exists a map F : U W that is dierentiable at a and such
that F[
UA
= f[
UA
, and we have dF[
TaA
= d
a
f. As it follows from the above discussion the map
dF[
TaA
is independent of this extension. Similarly, any C
m
-smooth map of A locally extends to a
C
m
-smooth map of a neighborhood of A in V .
Suppose that the image f(A) of a smooth map f : A W is contained in a submanifold B W.
In this case the image d
a
f(T
a
A) is contained in T
f(a)
B. Hence, given a smooth map f : A B
between two submanifolds A V and B W its dierential at a point a can be viewed as a linear
map d
a
f : T
a
A T
f(a)
B.
Let us recall, that given two submanifolds A V and B W (with or without boundary), a
smooth map f : A B is called a dieomorphism if there exists a smooth inverse map : B A,
i.e. f g : B B and g f : A A are both identity map. The submanifolds A and B are called
in this case dieomorphic.
Exercise 9.51. 1. Let A, B be two dieomorphic submanifolds. Prove that
(a) if A is path-connected then so is B;
(b) if A is compact then so is B;
(c) if A = then B = ;
(d) dimA = dimB;
5
2. Give an example of two dieomorphic submanifolds, such that one is bounded and the other
is not.
5
In fact, we will prove later that even homeomorphic manifolds should have the same dimension.
152
3. Prove that any closed connected 1-dimensional submanifold is dieomorphic to the unit circle
S
1
= x
2
1
+x
2
2
= 1 R
2
.
9.10 Vector bundles and their homomorphisms
Let us put the above discussion in a bit more global and general setup.
A collection of all tangent spaces T
a
A
aA
to a submanifold A is called its tangent bundle
and denoted by TA or T(A). This is an example of a more general notion of a vector bundle of
rank r over a set A V . One understands by this a family of r-dimensional vector subspaces
L
a
V
a
, parameterized by points of A and continuously (or C
m
-smoothly) depending on a. More
precisely one requires that each point a A has a neighborhood U A such that there exist linear
independent vector elds v
1
(a), . . . , v
r
(a) L
a
which continuously (smoothly, etc.) depend on a.
Besides the tangent bundle T(A) over a k-submanifold A an important example of a vector
bundle over a submanifold A is its normal bundle NA = N(A), which is a vector bundle of rank
n k formed by orthogonal complements N
a
A = T

a
A V
a
of the tangent spaces T
a
A of A. We
assume here that V is Euclidean space.
A vector bundle L of rank k over A is called trivial if one can nd k continuous linearly
independent vector elds v
1
(a), . . . , v
k
(a) L
a
dened for all a A. The set A is called the base of
the bundle L.
An important example of a trivial bundle is the bundle TV = V
a

aV
.
Exercise 9.52. Prove that the tangent bundle to the unit circle S
1
R
2
is trivial. Prove that the
tangent bundle to S
2
R
3
is not trivial, but the tangent bundle to the unit sphere S
3
R
4
is
trivial. (The case of S
1
is easy, of S
3
is a bit more dicult, and of S
2
even more dicult. It turns
out that the tangent bundle TS
n1
to the unit sphere S
n1
R
n
is trivial if and only if n = 2, 4
and 8. The only if part is a very deep topological fact which was proved by F. Adams in 1960.
Suppose we are given two vector bundles, L over A and

L over

A and a continuous (resp. smooth)
map : A

A. By a continuous (resp. smooth) homomorphism : L

L which covers the map
: A

A we understand a continuous (resp. smooth) family of linear maps
a
: L
a


L
(a)
. For
instance, a C
m
-smooth map f : A B denes a C
m1
-smooth homomorphism df : TA TB
153
which covers f : A B. Here df = d
a
f
aA
is the family of linear maps d
a
f : T
a
A T
f(a)
B,
a A.
9.11 Orientation
By an orientation of a vector bundle L = L
a

aA
over A we understand continuously depending
on a orientation of all vector spaces L
a
. An orientation of a submanifold k is the same as an
orientation of its tangent bundle T(A). A co-orientation of a k- submanifold A is an orientation of
its normal bundle N(A) = T

A in V . Note that not all bundles are orientable, i.e. some bundles
admit no orientation. But if L is orientable and the base A is connected, then L admits exactly two
orientations. Here is a simplest example of a non-orientable rank 1 bundle of the circle S
1
R
2
.
Let us identify a point a S
1
with a complex number a = e
i
, and consider a line l
a
R
2
a
directed
by the vector e
i
2
. Hence, when the point completes a turn around S
1
the line l
a
rotates by the
angle . We leave it to the reader to make a precise argument why this bundle is not orientable .
In fact, rank 1 bundles are orientable if and only if they are trivial.
If the ambient space V is oriented then co-orientation and orientation of a submanifold A deter-
mine each other according to the following rule. For each point a, let us choose any basis v
1
, . . . , v
k
of T
a
(A) and any basis w
1
, . . . , w
nk
of N
a
(A). Then w
1
, . . . , w
nk
, v
1
, . . . , v
k
is a basis of V
a
= V .
Suppose one of the bundles, say N(A), is oriented. Let us assume that the basis w
1
, . . . , w
nk
de-
nes this orientation. Then we orient T
a
A by the basis v
1
, . . . , v
k
if the basis w
1
, . . . , w
nk
, v
1
, . . . , v
k
denes the given orientation of V , and we pick the opposite orientation of T
a
A otherwise.
Example 9.53. (Induced orientation of the boundary of a submanifold.) Suppose A is an
oriented manifold with boundary. Let us co-orient the boundary A by orienting the rank 1 normal
bundle to T(A) in T(A) by the unit ourtward normal to T(A) in T(A) vector eld. Then the
above rule determine an orientation of T(A), and hence of A.
9.12 Integration of dierential k-forms over k-dimensional sub-
manifolds
Let be a dierential k-form dened on an open set U V .
154
Figure 9.6: The orientation of the surface is induced by its co-orientation by the normal vector n.
The orientation of the boundary us induced by the orientation of the surface.
155
Consider rst a k-dimensional compact submanifold with boundary A U dened parametri-
cally by an embedding : G A U, where G R
k
is possibly with boundary. Suppose that A
is oriented by this embedding. Then we dene
_
A
:=
_
G

.
Note that if we dene A by a dierent embedding

:

G A, then we have

= h, where
h =
1

:

G G is a dieomorphism. Hence, using Theorem 9.30 we get
_
e
G

=
_
e
G
h

) =
_
G

,
and hence
_
A
is independent of a choice of parameterization, provided that the orientation is
preserved.
Let now A be any compact oriented submanifold with boundary. Let us choose a partition of
unity 1 =
K

j
in a neighborhood of A such that each function is supported in some coordinate
neighborhood of A. Denote
j
=
j
. Then =
K

j
, where each form
j
is supported in one
of coordinate neighborhoods. Hence there exist orientation preserving embeddings
j
: G
j
A
of domains with boundary G
j
R
k
, such that
j
(G
j
) Supp(
j
), j = 1, . . . , K. Hence, we can
dene
_
A

j
:=
_
G
j

j
and
_
A
:=
K

1
_
A

j
.
Lemma 9.54. The above denition of
_
A
is independent of a choice of a partition of unity.
Proof. Consider two dierent partitions of unity 1 =
K

j
and 1 =
e
K

j
subordinated to coverings
U
1
, . . . , U
K
and

U
1
, . . . ,

U
e
K
, respectively. Taking the product of two partitions we get another
partition 1 =
K

i=1
e
K

j=1

ij
, where
ij
:=
i

j
, which is subordinated to the covering by intersections
U
i
U
j
, i = 1, . . . , K, j = 1, . . . ,

K. Denote
i
:=
i
,
j
:=

j
and
ij
=
ij
. Then
K

i=1

ij
=
j
,
156
e
K

j=1

ij
=
i
and =
K

i
=
e
K

1

j
. Then, using the linearity of the integral we get
K

1
_
A

i
=
K

i=1
e
K

j=1
_
A

ij
=
e
K

1
_
A

j
.

When k = 1 the above denition of the integral coincides with the denition of the integral of
a 1-form over an oriented curve which was given above in Section 9.2.
Let us extend the denition of integration of dierential forms to an important case of integration
of 0-form over oriented 0-dimensional submanifolds. Let us recall a compact oriented 0-dimensional
submanifold of V is just a nite set of points a
1
, . . . , a
m
V with assigned signs to every point.
So in view of the additivity of the integral it is sucient to dene integration over 1 point with a
sign. On the other hand, a 0-form is just a function f : V R. So we dene
_
a
f := f(a).
A partition of unity is a convenient tool for studying integrals, but not so convenient for practical
computations. The following proposition provides a more practical method for computations.
Proposition 9.55. Let A be a compact oriented submanifold of V and a dierential k-form given
on a neighborhood of A. Suppose that A presented as a union A =
N

1
A
j
, where A
j
are codimension
0 submanifolds of A with boundary with corners. Suppose that A
i
and A
j
for any i ,= j intersect
only along pieces of their boundaries. Then
_
A
=
N

1
_
A
j
.
In particular, if each A
j
is parameterized by an orientation preserving embedding
j
: G
j
A,
where G
j
R
k
is a domain with boundary with corners. Then
_
A
=
N

1
_
G
j

j
.
157
We leave the proof to the reader as an exercise.
Exercise 9.56. Compute the integral
_
S
1/3(x
1
dx
2
dx
3
+x
2
dx
3
dx
1
+x
3
dx
1
dx
2
),
where S is the sphere
x
2
1
+x
2
2
+x
2
3
= 1,
cooriented by its exterior normal vector.
Solution. Let us present the sphere as the union of northern and southern hemispheres:
S = S

S
+
, where S

= S x
3
0, S
+
= S x
3
0.
Then
_
S
=
_
S
+
+
_
S

. Let us rst compute

_
S
+
.
We can parametrize S
+
by the map (u, v) (u, v,

R
2
u
2
v
2
), (u, v) u
2
+ v
2
R
2
=
T
R
. One can check that this parametrization agrees with the prescribed coorientation of S. Thus,
we have
_
S
+
= 1/3
_
D
R
_
udv d
_
R
2
u
2
v
2
+vd
_
R
2
u
2
v
2
du +
_
R
2
u
2
v
2
du dv
_
.
Passing to polar coordinates (r, ) in the plane (u, v) we get
_
S
+
= 1/3
_
P
r cos d(r sin) d
_
R
2
r
2
+r sind
_
R
2
r
2
d(r cos )
+
_
R
2
r
2
d(r cos ) d(r sin),
where P = 0 r R, 0 2. Computing this integral we get
158
_
S
+
=
1
3
_
P

r
3
cos
2
d dr

R
2
r
2
+
r
3
sin
2
dr d

R
2
r
2
+
_
R
2
r
2
dr d
=
1
3
_
P
_
r
3

R
2
r
2
+r
_
R
2
r
2
_
dr d
=
2
3
_
R
0
rR
2

R
2
r
2
dr =
2R
2
3
_
R
2
r
2
_
R
0
=
2R
3
3
Similarly, one can compute that
_
S

=
2R
3
3
.
Computing this last integral, one should notice the fact that the parametrization
(u, v) (u, v,
_
R
2
u
2
v
2
)
denes the wrong orientation of S

. Thus one should use instead the parametrization

(u, v) (v, u,
_
R
2
u
2
v
2
),
and we get the answer
_
S
=
4R
3
3
.
This is just the volume of the ball bounded by the sphere. The reason for such an answer will be
clear below from Stokes theorem.
159
160
Part III
Stokes theorem and its applications
161
Chapter 10
Stokes theorem
10.1 Statement of Stokes theorem
Theorem 10.1. Let A V be a compact oriented submanifold with boundary (and possibly with
corners). Let be a C
2
-smooth dierential form dened on a neighborhood U A. Then
_
A
=
_
A
d .
Here d is the exterior dierential of the form and A is the oriented boundary of A.
We will discuss below what exactly Stokes theorem means for the case k 3 and n = dimV 3.
Let us begin with the case k = 1, n = 2. Thus V = R
2
. Let x
1
, x
2
be coordinates in R
2
and U
a domain in R
2
bounded by a smooth curve = U. Let us co-orient with the outward normal
to the boundary of U. This denes a counter-clockwise orientation of .
Let = P
1
(x
1
, x
2
)dx
1
+P
2
(x
1
, x
2
)dx
2
be a dierential 1-form. Then the above Stokes formula
asserts
_
U
d =
_

,
or
_
U
_
P
2
x
1

P
1
x
2
_
dx
1
dx
2
=
_

P
1
dx
1
+P
2
dx
2
.
163
Figure 10.1: George Stokes (1819-1903)
This is called Greens formula. In particular, when d = dx
1
dx
2
, e.g. = xdy or =
1
2
(xdy ydx), the integral
_

computes the area of the domain U.

Consider now the case n = 3, k = 2. Thus
V = R
3
, = P
1
dx
2
dx
3
+P
2
dx
3
dx
1
+P
3
dx
1
dx
2
.
Let U R
3
be a domain bounded by a smooth surface S. We co-orient S with the exterior normal
. Then
d =
_
P
1
x
1
+
P
2
x
2
+
P
3
x
3
_
dx
1
dx
2
dx
3
.
Thus, Stokes formula
_
S
=
_
U
d
gives in this case
_
S
P
1
dx
2
dx
3
+P
2
dx
3
dx
1
+P
3
dx
1
dx
2
=
_
U
_
P
1
x
1
+
P
2
x
2
+
P
3
x
3
_
dx
1
dx
2
dx
3
This is called the divergence theorem or Gauss-Ostrogradskis formula.
164
Figure 10.2: George Green (1793-1841)
Consider the case k = 0, n = 1. Thus is just a function f on an interval I = [a, b]. The
boundary I consists of 2 points: I = a, b. One should orient the point a with the sign and
the point b with the sign +.
Thus, Stokes formula in this case gives
_
[a,b]
df =
_
{a,+b}
f,
or
_
b
a
f

(x)dx = f(b) f(a).

This is Newton-Leibnitz formula. More generally, for a 1-dimensional oriented connected curve
R
3
with boundary = B (A) and any smooth function f we get the formula
_

df =
_
B(A)
f = f(B) f(A),
which we already proved earlier, see Theorem 9.11 .
Consider now the case n = 3, k = 1.
Thus V = R
3
and = P
1
dx
1
+P
2
dx
2
+P
3
dx
3
. Let S R
3
be an oriented surface with boundary
165
Figure 10.3: Carl Friedrich Gauss (1777-1855) Mikhail Ostrogradski (1801-1862)
. We orient in the same way, as in Greens theorem. Then Stokes formula
_
S
d =
_

gives in this case

_
S
_
P
3
x
2

P
2
x
3
_
dx
2
dx
3
+
_
P
1
x
3

P
3
x
1
_
dx
3
dx
1
+
_
P
2
x
1

P
1
x
2
_
dx
1
dx
2
=
_

P
1
dx
1
+P
2
dx
2
+P
3
dx
3
.
This is the original Stokes theorem.
Stokes theorem allows one to clarify the geometric meaning of the exterior dierential.
Lemma 10.2. Let be a dierential k-form in a domain U V . Take any point a U and
vectors X
1
, . . . , X
k+1
V
a
. Given > 0 let us consider the parallelepiped P(X
1
, . . . , X
k+1
) as a
subset of V with vertices at points a
i
1
...i
k+1
= a +
k+1

1
i
j
X
j
, where each index i
j
takes values 0, 1.
Then
d
a
(X
1
, . . . X
k+1
) = lim
0
1

k+1
_
P(X
1
,...,X
k+1
)
.
166
Proof. First, it follows from the denition of integral of a dierential form that
d
a
(X
1
, . . . , X
k+1
) = lim
0
1

k+1
_
P(X
1
,...,X
k+1
)
d. (10.1.1)
Then we can continue using Stokes formula
d
a
(X
1
, . . . , X
k+1
) = lim
0
1

k+1
_
P(X
1
,...,X
k+1
)
d = lim
0
1

k+1
_
P(X
1
,...,X
k+1
)
. (10.1.2)

10.2 Proof of Stokes theorem

We prove in this section Theorem 10.1. We will consider only the case when A is a manifold with
boundary without corners and leave the corner case as an exercise to the reader.
Let us cover A by coordinate neighborhoods such that in each neighborhood A is given either
by (9.8.3) or (9.8.4). First we observe that it is sucient to prove the theorem for the case of a
form supported in one of these coordinate neighborhoods. Indeed, let us choose nitely many such
neighborhoods covering A. Let 1 =
N

j
be a partition of unity subordinated to this covering. We
set
j
=
j
, so that =
N

j
, and each of
j
is supported in one of coordinate neighborhoods.
Hence, if formula 10.1 holds for each
j
it also holds for .
Let us now assume that is supported in one of coordinate neighborhoods. Consider the
corresponding parameterization : G U V , G R
n
, introducing coordinates u
1
, . . . , u
n
. Then
AU = (GL), where L is equal to the subspace R
k
= u
k+1
= . . . u
n
= 0 in the case (9.8.3) and
the upper-half space R
k
u
1
0. By denition, we have
_
A
d =
_
U
d =
_
GL

d =
_
GL
d

.
1
Though the form =

[
GL
is dened only on G L, it is supported in this neighborhood,
and hence we can extend it to a smooth form on the whole L by setting it equal to 0 outside the
neighborhood. With this extension we have
_
GL
d =
_
L
d . The (k 1)-form can be written in
coordinates u
1
, . . . , u
k
as
=
j

1
f
j
(u)du
1
. . .
j
du
k
.
1
We assume here that the coordinates u1, . . . , u
k
dene the given orientation of A.
167
Then
_
GL
d =
_
L
_
k

1
f
j
u
j
_
du
1
du
k
.
Let us choose a suciently R > 0 so that the cube I = [u
i
[ R, i = 1, . . . , k contains Supp( ).
Thus in the case (9.8.3) we have
_
GL
d =
k

1
_
R
k
f
j
u
j
dV =
k

1
R
_
R
. . .
R
_
R
f
j
u
j
du
1
. . . du
n
=
k

1
R
_
R
. . .
_
_
R
_
R
f
j
u
j
du
j
_
_
du
1
. . . du
j1
du
j+1
. . . du
n
= 0 (10.2.1)
because
R
_
R
f
j
u
j
du
j
= f
j
(u
1
, . . . , u
i1
, R, u
i
, . . . , u
n
) f
j
(u
1
, . . . , u
i1
, R, u
i
, . . . , u
n
)) = 0.
On the other hand, in this case
_
A
= 0, because the support of does not intersect the boundary
of A. Hence, Stokes formula holds in this case. In case (9.8.4) we similarly get
_
GL
d =
k

1
_
{u
1
0}
f
j
u
j
dV =
k

1
R
_
0
_
_
R
_
R
. . .
R
_
R
f
j
u
j
du
n
. . . du
2
_
_
du
1
=
R
_
R
_
_
R
_
R
. . .
R
_
0
f
1
u
1
du
1
. . . du
n1
_
_
du
n
=

R
_
R
. . .
R
_
R
f
1
(0, u
2
, . . . , u
n
)du
2
. . . du
n
. (10.2.2)
because all terms in the sum with j > 1 are equal to 0 by the same argument as in (10.2.1). On
the other hand, in this case
_
A
=
_
{u
1
=0}

=
_ _
{u
1
=0}
f
1
(0, u
2
, . . . , u
n
)du
2
du
n
=

R
_
R
. . .
R
_
R
f
1
(0, u
2
, . . . , u
n
)du
2
. . . du
n
. (10.2.3)
168
The sign minus appears in the last equality in front of the integral because the induced orien-
tation on the space u
1
= 0 as the boundary of the upper-half space u
1
0 is opposite to the
orientation dened by the volume form du
2
du
n
. Comparing the expressions (10.2.2) and
(10.2.3) we conclude that
_
A
d =
_
A
, as required.
10.3 Integration of functions over submanifolds
In order to integrate functions over a submanifold we need a notion of volume for subsets of the
submanifold.
Let A V be an oriented k-dimensional submanifold, 0 k n. By denition, the volume
form =
A
of A (or the area form if k = 2, or the length form if k = 1) is a dierential k-form
on A whose value on any k tangent vectors v
1
, . . . , v
k
T
x
A equals the oriented volume of the
parallelepiped generated by these vectors.
Given a function f : A R we dene its integral over A by the formula
_
A
fdV =
_
A
f
A
, (10.3.1)
and, in particular,
Vol A =
_
A

A
.
Notice that the integral
_
A
fdV is independent of the orientation of A. Indeed, changing the orienta-
tion we also change the sign of the form
A
, and hence the integral remains unchanged. This allows
us to dene the integral
_
A
fdV even for a non-orientable A. Indeed, we can cover A by coordinate
charts, nd a subordinated partition of unity and split correspondingly the function f =
N

1
f
j
in such a way that each function f
j
is supported in a coordinate neighborhood. By orienting in
arbitrary ways each of the coordinate neighborhoods we can compute each of the integrals
_
A
f
j
dV ,
j = 1, . . . , N. It is straightforward to see that the integral
_
A
fdV =

j
_
A
f
j
dV is independent of the
choice of the partition of unity.
Let us study in some examples how the form
A
can be eectively computed.
169
Example 10.3. Volume form of a hypersurface. Let us x a Cartesian coordiantes in V . Let A V
is given by the equation
A = F = 0
for some function F : V R which has no critical points on A. The vector eld F is orthogonal
to A, and
n =
F
[[F[[
is the unit normal vector eld to A. Assuming A to be co-oriented by n we can write down the
volume form of A as the contraction of n with the volume form = dx
1
dx
n
of R
n
, i.e.

A
= n =
1
[[F[[
n

1
(1)
i1
F
x
i
dx
1

i

. . . dx
n
.
In particular, if n = 3 we get the following formula for the area form of an implicitely given
2-dimensional surface A = F = 0 R
3
:

A
=
1
_
_
F
x
1
_
2
+
_
F
x
2
_
2
+
_
F
x
3
_
2
_
F
x
1
dx
2
dx
3
+
F
x
2
dx
3
dx
1
+
F
x
3
dx
1
dx
2
_
. (10.3.2)
Example 10.4. Length form of a curve.
Let R
n
be an oriented curve given parametrically by a map : [a, b] R
n
. Let =

be
the length form. Let us compute the form

. Denoting the coordinate in [a, b] by t and the unit

vector eld on [a, b] by e we have

= f(t)dt,
where
f(t) =

(e) =

(t)
_
= [[

(t)[[ .
In particular the length of is equal to
_

=
b
_
a
[[

(t)[[dt =
b
_
a

_
n

i=1
(x

i
(t))
2
dt ,
where
(t) = (x
1
(t), . . . , x
n
(t)) .
170
Similarly, given any function f : R we have
_

fds =
b
_
a
f ((t)) [[

(t)[[dt .
Example 10.5. Area form of a surface given parametrically.
Suppose a surface S R
n
is given parametrically by a map : U R
n
where U in the plane R
2
with coordinates (u, v).
Let us compute the pull-back form

S
. In other words, we want to express
S
in coordinates
u, v. We have

S
= f(u, v)du dv .
To determine f(u, v) take a point z = (u, v) R
2
and the standard basis e
1
, e
2
R
2
z
. Then
(

S
)
z
(e
1
, e
2
) = f(u, v)du dv(e
1
, e
2
) . (10.3.3)
On the other hand, by the denition of the pull-back form we have
(

S
)
z
(e
1
, e
2
) = (
S
)
(z)
(d
z
(e
1
), d
z
(e
2
)) . (10.3.4)
But d
z
(e
1
) =

u
(z) =
u
(z) and d
z
(e
2
) =

v
(z) =
v
(z). Hence from (10.3.3) and (10.3.4) we
get
f(u, v) =
S
(
u
,
v
) . (10.3.5)
The value of the form
S
on the vectors
u
,
v
is equal to the area of the parallelogram gener-
ated by these vectors, because the surface is assumed to be oriented by these vectors, and hence

S
(
u
,
v
) > 0. Denoting the angle between
u
and
v
by we get
2

S
(
u
,
v
) = [[
u
[[ [[
v
[[ sin.
Hence

S
(
u
,
v
)
2
= [[
u
[[
2
[[
v
[[
2
sin
2
= [[
u
[[
2
[[
v
[[
2
(1 cos
2
) = [[
u
[[
2
[[
v
[[
2
(
u

v
)
2
,
and therefore,
f(u, v) =
S
(
u
,
v
) =
_
[[
u
[[
2
[[
v
[[
2
(
u

v
)
2
.
2
See a computation in a more general case below in Example 10.6.
171
It is traditional to introduce the notation
E = [[
u
[[
2
, F =
u

v
, G = [[
v
[[
2
,
so that we get

S
=
_
EGF
2
du dv ,
and hence we get for any function f : S R
_
S
fdS =
_
S
f
S
=
_
U
f((u, v))
_
EGF
2
du dv =
_ _
U
f((u, v))
_
EGF
2
dudv . (10.3.6)
Consider a special case when the surface S dened as a graph of a function over a domain
D R
2
. Namely, suppose
S = z = (x, y), (x, y) D R
2
.
The surface S as parametrized by the map
(x, y)

(x, y, (x, y)).
Then
E = [[
x
[[
2
= 1 +
2
x
, G = [[
y
[[
2
= 1 +
2
y
, F =
x

y
=
x

y
,
and hence
EGF
2
= (1 +
2
x
)(1 +
2
y
)
2
x

2
y
= 1 +
2
x
+
2
y
.
Therefore, the formula (10.3.6) takes the form
_
S
fdS =
_ _
D
f((x, y))
_
EGF
2
dx dy =
_ _
D
f (x, y, (x, y))
_
1 +
2
x
+
2
y
dxdy . (10.3.7)
Note that the formula (10.3.7) can be also deduced from (10.3.2). Indeed, the surface
S = z = (x, y), (x, y) D R
2
,
can also be dened implicitly by the equation
F(x, y, z) = z (x, y) = 0, (x, y) D.
172
We have
F = (

x
,

y
, 1),
and, therefore,
_
S
fdS =
_
S
f(x, y, z)
[[F[[
_
F
x
dy dz +
F
y
dz dx +
F
z
dx dy
_
=
_ _
D
f(x, y, (x, y))
_
1 +
_

x
_
2
+
_

y
_
2
_

x
dy d

y
d dx +dx dy
_
=
_ _
D
f(x, y, (x, y))
_
1 +
_

x
_
2
+
_

y
_
2
_
1 +
_

x
_
2
+
_

y
_
2
_
dxdy
=
_ _
D
f(x, y, (x, y))

1 +
_

x
_
2
+
_

y
_
2
dxdy .
Example 10.6. Integration over a parametrically given k-dimensional submanfold.
Consider now a more general case of a parametrically given k-submanifold A in an n-dimensional
Euclidean space V . We x a Cartersian coordinate system in V and thus identify V with R
n
with
the standard dot-product.
Let U R
k
be a compact domain with boundary and : U R
n
be an embedding. We assume
that the submanifold with boundary A = (U) is oriented by this parameterization. Let
A
be the
volume form of A. We will nd an explicit expression for

A
. Namely. denoting coordinates in
R
k
by (u
1
, . . . , u
k
) we have

A
= f(u)du
1
du
k
, and our goal is to compute the function f.
By denition, we have
f(u) =

(
A
)
u
(e
1
, . . . , e
k
) = (
A
)
u
(d
u
(e
1
), . . . , d
u
(e
k
)) =
(
A
)
u
_

u
1
(u), . . . ,

u
k
(u)
_
= Vol
k
_
P
_

u
1
(u), . . . ,

u
k
(u)
__
(10.3.8)
In Section 4.2 we proved two formulas for the volume of a parallelepiped. Using formula (4.2.1)
we get
Vol
k
_
P
_

u
1
(u), . . . ,

u
k
(u)
__
=

1i
1
<<i
k
n
Z
2
i
1
...i
k
. (10.3.9)
173
where
Z
i
1
...i
k
=

i
1
u
1
(u) . . .

i
1
u
k
(u)
. . . . . . . . .

i
k
u
1
(u) . . .

i
k
u
k
(u)

. (10.3.10)
Thus
_
A
fdV =
_
A
f
A
=
_
U
f((u))

1i
1
<<i
k
n
Z
2
i
1
...i
k
du
1
du
k
.
Rewriting this formula for k = 2 we get
_
A
fdV =
_
U
f((u))

1i<jn

i
u
1

i
u
2

j
u
1

j
u
2

2
du
1
du
2
. (10.3.11)
Alternatively we can use formula (??). Then we get
Vol
k
_
P
_

u
1
(u), . . . ,

u
k
(u)
__
=
_
det
_
(D)
T
D
_
, (10.3.12)
where
D =
_
_
_
_
_

1
u
1
. . .

1
u
k
. . . . . . . . .
n
u
1
. . .
n
u
k
_
_
_
_
_
is the Jacobi matrix of .
3
Thus using this expression we get
_
A
fdV =
_
A
f
A
=
_
U
f((u))
_
det
_
(D)
T
D
_
dV. (10.3.13)
Exercise 10.7. In case n = 3, k = 2 show explicitely equivalence of formulas (10.3.6), (10.3.11)
and (10.3.13).
Exercise 10.8. Integration over an implicitly dened k-dimensional submanfold. Suppose that
A = F
1
= = F
nk
= 0 and the dierentials of dening functions are linearly independent at
points of A. Show that

A
=
(dF
1
dF
nk
)
[[ dF
1
dF
nk
[[
.
3
The symmetric matrix (D)
T
D is the Gram matrix of vectors

u
1
(u), . . . ,

u
k
(u), and its entrees are pairwise
scalar products of these vectors, see Remark ??.
174
Example 10.9. Let us compute the volume of the unit 3-sphere S
3
= x
2
1
+x
2
2
+x
2
3
+x
2
4
= 1.
By denition, Vol(S
3
) =
_
S
3
n , where = dx
1
dx
2
dx
3
dx
4
, and n is the outward unit
normal vector to the unit ball B
4
. Here S
3
should be co-oriented by the vector eld n. Then using
Stokes theorem we have
_
S
3
n =
_
S
3
(x
1
dx
2
dx
3
dx
4
x
2
dx
1
dx
3
dx
4
+x
3
dx
1
dx
2
dx
4
x
4
dx
1
dx
2
dx
3
) =
4
_
B
4
dx
1
dx
2
dx
3
dx
4
= 4
_
B
4
dV.
(10.3.14)
Introducing polar coordinates (r, ) and (, ) in the coordinate planes (x
1
, x
2
) and (x
3
, x
4
) and
using Fubinis theorem we get
_
B
4
=
2
_
0
2
_
0
1
_
0

1r
2
_
0
rddrdd =
2
2
1
_
0
(r r
3
)dr =

2
2
. (10.3.15)
Hence, Vol(S
3
) = 2
2
.
Exercise 10.10. Find the ratio
Voln(B
n
R
)
Vol
n1
(S
n1
R
)
.
10.4 Work and Flux
We introduce in this section two fundamental notions of vector analysis: a work of a vector eld
along a curve, and a ux of a vector eld through a surface. Let be an oriented smooth curve in
a Euclidean space V and T the unit tangent vector eld to . Let v be another vector eld, dened
along . The function v, T) equals the projection of the vector eld v to the tangent directions
to the curve. If the vector eld v is viewed as a force eld, then the integral
_

v, T)ds has the

meaning of a work Work

( v) performed by the eld v to transport a particle of mass 1 along the

curve in the direction determined by the orientation. If the curve is closed then this integral is
sometimes called the circulation of the vector eld v along and denoted by
_

v, T)ds. As we
175
already indicated earlier, the sign
_
in this case has precisely the same meaning as
_
, and it is used
only to stress the point that we are integrating along a closed curve.
Consider now a co-oriented hypersurface V and denote by n the unit normal vector eld
to which determines the given co-orientation of . Given a vector eld v along we will view
it as the velocity vector eld of a ow of a uid in the space. Then we can interpret the integral
_

v, n)dV
as the ux Flux

( v) of v through , i.e. the volume of uid passing through in the direction of

n in time 1.
Lemma 10.11. 1. For any co-oriented hypersurface and a vector eld v given in its neigh-
borhood we have
v, n)

= ( v )

,
where is the volume form in V .
2. For any oriented curve and a vector eld v near we have
v, T)

= T( v)[

.
Proof. 1. For any n 1 vectors T
1
, . . . , T
n1
T
x
we have
v (T
1
, . . . , T
n1
) = ( v, T
1
, . . . , T
n1
) = VolP( v, T
1
, . . . , T
n1
).
Using (3.3.1) we get
VolP( v, T
1
, . . . , T
n1
) = v, n)Vol
n1
P(T
1
, . . . , T
n1
) =
v, n)VolP( n, T
1
, . . . , T
n1
) = v, n)

(T
1
, . . . , T
n1
). (10.4.1)
2. The tangent space T
x
is generated by the vector T, and hence we just need to check that
v, T)

(T) = T( v)(T). But

(T) = Vol( n, T) = 1, and hence

T( v)(T) = v, T) = v, T)

(T).
176

Note that if we are given a Cartesian coordinate system in V and v =

n

1
a
j

x
j
, then
v =
n1

1
(1)
i1
a
i
dx
1
i

. . . dx
n
, T( v) =
n

1
a
i
dx
i
.
Thus, we have
Corollary 10.12.
Flux

( v) =
_

v, n)dV =
_

( v ) =
_

1
(1)
i1
a
i
dx
1
i

. . . dx
n
;
Work

( v) =
_

v, T) =
_

T( v) =
_

1
a
i
dx
i
.
In particular if n = 3 we have
Flux

( v) =
_

a
1
dx
2
dx
3
+a
2
dx
3
dx
1
+a
3
dx
1
dx
2
.
Let us also recall that in a Euclidean space V we have v = T(v). Hence, the equation
= v is equivalent to the equation
v = T
1
_

_
= (1)
n1
T
1
().
In particular, when n = 3 we get v = T
1
(). Thus we get
Corollary 10.13. For any dierential (n1)-form and an oriented compact hypersurface we
have
_

= Flux

v,
where v = (1)
n1
T
1
().
Integration of functions along curves and surfaces can be interpreted as the work and the ux
of appropriate vector elds. Indeed, suppose we need to compute an integral
_

fds. Consider the

tangent vector eld v(x) = f(x)T(x), x , along . Then v, T) = f and hence the integral
_

fds can be interpreted as the work Work

( v). Therefore, we have

_

fds = Work

( v) =
_

T( v)
177
.
Note that we can also express v through by the formula v = T
1
, see Section 8.7.
Similarly, to compute an integral
_

fdS let us co-orient the surface with a unit normal to

vector eld n(x), x and set v(x) = f(x) n(x). Then v, n) = f, and hence
_

fdS =
_

v, n)dS = Flux

( v) =
_

,
where = v .
10.5 Integral formulas of vector analysis
We interpret in this section Stokes formula in terms of integrals of functions and operations on
vector elds. Let us consider again dierential forms, which one can associate with a vector eld
v in an Euclidean 3-space. Namely, this is a dierential 1-form = T( v) and a dierential 2-form
= v , where = dx dy dz is the volume form.
Using Corollary 10.12 we can reformulate Stokes theorem for domains in a R
3
as follows.
Theorem 10.14. Let v be a smooth vector eld in a domain U R
3
with a smooth (or piece-wise)
smooth boundary . Suppose that is co-oriented by an outward normal vector eld. Then we have
Flux

v =
_ _ _
U
div vdxdydz .
Indeed, div v = d. Hence we have
_
U
div vdV =
_
U
(d)dx dy dz =
_
U
d =
_

=
_

v = Flux

v .
This theorem claries the meaning of div v:
Let B
r
(x) be the ball of radius r centered at a point x R
3
, and S
r
(x) = B
r
(x) be its
boundary sphere co-oriented by the outward normal vector eld. Then
div v(x) = lim
r0
Flux
Sr(x)
v
Vol(B
r
(x))
.
178
Theorem 10.15. Let be a piece-wise smooth compact oriented surface in R
3
with a piece-wise
smooth boundary = oriented respectively. Let v be a smooth vector eld dened near . Then
Flux

(curl v) =
_

curl v, n)dV =
_

v Tds = Work

v .
To prove the theorem we again use Stokes theorem and the connection between integrals of
functions and dierential forms. Set = T( v). Then curl v = T
1
d. We have
_

v Tds =
_

=
_

d = Flux

(T
1
(d)) = Flux

(curl v) .
Again, similar to the previous case, this theorem claries the meaning of curl. Indeed, let us
denote by D
r
(x, w) the 2-dimensional disc of radius r in R
3
centered at a point x R
3
and
orthogonal to a unit vector w R
3
x
. Set
c(x, w) = lim
r0
Work
Dr(x, w)
v
r
2
.
Then
c(x, w) = lim
r0
Flux
Dr(x, w)
(curl v)
r
2
= lim
r0
_
Dr(x, w)
curl v, w)
r
2
= curl v, w).
Hence, [[curl v(x)[[ = max
w
c(x, w) and direction of curl v(x) coincides with the direction of the
vector w for which the maximum value of c(x, w) is achieved.
10.6 Expressing div and curl in curvilinear coordinates
Let us show how to compute div v and curl v of a vector eld v in R
3
given in a curvilinear
coordinates u
1
, u
2
, u
3
, i.e. expressed through the coordinate vector elds

u
1
,

u
2
and

u
3
. Let
= f(u
1
, u
2
, u
3
)du
1
du
2
du
3
be the volume form dx
1
dx
2
dx
3
expressed in coordinates u
1
, u
2
, u
3
.
Let us rst compute (du
1
du
2
du
3
). We have
(du
1
du
2
du
3
) =
_
1
f
dx
1
dx
2
dx
3
_
=
1
f
.
179
Let
v = a
1

u
1
+a
2

u
2
+a
3

u
3
.
Then we have
div v = d( v )
= d
__
3

1
a
i

u
i
_
fdu
1
du
2
du
3
_
= d (fa
1
du
2
du
3
+fa
2
du
3
du
1
+fa
3
du
1
du
2
)
=
__
(fa
1
)
u
1
+
(fa
2
)
u
2
+
(fa
3
)
u
3
_
du
1
du
2
du
3
_
=
1
f
_
(fa
1
)
u
1
+
(fa
2
)
u
2
+
(fa
3
)
u
3
_
=
a
1
u
1
+
a
2
u
2
+
a
3
u
3
+
1
f
_
f
u
1
a
1
+
f
u
2
a
2
+
f
u
3
a
3
_
.
In particular, we see that the divergence of a vector eld is expressed by the same formulas as in
the cartesian case if and only if the volume form is proportional to the form du
1
du
2
du
3
with
a constant coecient.
For instance, in the spherical coordinates the volume form can be written as
= r
2
sindr d d ,
4
and hence the divergence of a vector eld
v = a

r
+b

+c

can be computed by the formula

div v =
a
r
+
b

+
c

+
2a
r
+c cot .
The general formula for curl v in curvilinear coordinates looks pretty complicated. So instead
of deriving the formula we will explain here how it can be obtained in the general case, and then
illustrate this procedure for the spherical coordinates.
By the denition we have
4
Note that the spherical coordinates ordered as (r, , ) determine the same orientation of R
3
as the cartesian
coordinates (x, y, z).
180
curl v = D
1
( (d (D( v)))) .
Hence we rst need to compute D( v).
To do this we need to introduce a symmetric matrix
G =
_
_
_
_
_
g
11
g
12
g
13
g
21
g
22
g
23
g
31
g
32
g
33
_
_
_
_
_
,
where
g
ij
=

u
i
,

u
j
), i, j = 1, 2, 3 .
The matrix G is called the Gram matrix.
Notice, that if D( v) = A
1
du
1
+Bdu
2
+Cdu
3
then for any vector h = h
1

u
1
+h
2

u
2
+h
3

u
3
.
we have
D( v)(h) = A
1
h
1
+A
2
h
2
+A
3
h
3
=
_
A
1
A
2
A
3
_
_
_
_
_
_
h
1
h
2
h
3
_
_
_
_
_
= v, h) .
But
v, h) =
_
a
1
a
2
a
3
_
G
_
_
_
_
_
h
1
h
2
h
3
_
_
_
_
_
.
Hence
_
A
1
A
2
A
3
_
=
_
a
1
a
2
a
3
_
G,
or, equivalently, because the Gram matrix G is symmetric we can write
_
_
_
_
_
A
1
A
2
A
3
_
_
_
_
_
= G
_
_
_
_
_
a
1
a
2
a
3
_
_
_
_
_
,
and therefore,
A
i
=
n

i=1
g
ij
a
j
, i = 1, 2, 3 .
181
After computing
= d(D( v)) = B
1
du
2
du
3
+B
2
du
3
du
1
+B
3
du
1
du
2
we compute curl v by the formula = curl v = D
1
(). Let us recall (see Proposition 4.13 above)
that for any vector eld w the equality Dw = is equivalent to the equality w = , where
= fdu
1
du
2
du
3
is the volume form. Hence, if
curl v = c
1

u
1
+c
2

u
2
+c
3

u
3
then we have
w = fc
1
du
2
du
3
+fc
2
du
3
du
1
+fc
3
du
1
du
2
,
and therefore,
curl v =
B
1
f

u
1
+
B
2
f

u
2
+
B
3
f

u
3
.
Let us use the above procedure to compute curl v of the vector eld
v = a

r
+b

+c

given in the spherical coordinates. The Gram matrix in this case is the diagonal matrix
G =
_
_
_
_
_
1 0 0
0 r
2
0
0 0 r
2
sin
2

_
_
_
_
_
.
Hence,
Dv = adr +br
2
d +cr
2
sin
2
d ,
and
= d(Dv) = da dr +d(br
2
) d +d(cr
2
sin
2
) d
=
_
r
b

+r
2
c sin2 +r
2
sin
2

_
d d
+
_

+r
2
b
r
+ 2br
_
dr d +
_
2rc sin
2
r
2
sin
2

c
r
+
a

_
d dr .
Finally, we get the following expression for curl v:
curl v =
r
b

+r
2
c sin2 +r
2
sin
2

r
2
cos

r
+

+r
2 b
r
+ 2br
r
2
cos

+
2rc sin
2
r
2
sin
2

c
r
+
a

r
2
cos

.
182
Chapter 11
Applications of Stokes formula
11.1 Integration of closed and exact forms
Let us recall that a dierential k-form is called closed if d = 0, and that it is called exact if
there exists a (k 1)-form , called primitive of , such that = d.
Any exact form is closed, because d(d) = 0. Any n-form in a n-dimensional space is closed.
Proposition 11.1. a) For a closed k-form dened near a (k + 1)-dimensional submanifold
with boundary we have
_

= 0 .
b) If is exact k-form dened near a closed k-dimensional submanifold S then
_
S
= 0 .
The proof immediately follows from Stokes formula. Indeed, in case a) we have
_

=
_

d = 0 .
In case b) we have = d and S = . Thus
_
S
d =
_

= 0 .
Proposition 11.1b) gives a necessary condition for a closed form to be exact.
183
Example 11.2. The dierential 1-form =
1
x
2
+y
2
(xdy ydx) dened on the punctured plane
R
2
0 is closed but not exact.
Indeed, it is straightforward to check that is exact (one can simplify computations by passing
to polar coordinates and computing that = d). To check that it is not exact we compute the
integral
_
S
, where S in the unit circle x
2
+y
2
= 1. We have
_
S
=
2
_
0
d = 2 ,= 0 .
More generally, an (n 1)-form

n
=
n

i=1
(1)
i1
x
i
r
n
dx
1

i
. . . dx
n
(dx
i
is missing)
is closed in R
n
0. However, it is not exact. Indeed, let us show that
_
S
n1

n
,= 0, where S
n1
is the
unit sphere oriented as the boundary of the unit ball. Let us recall that the volume form
S
n1 on
the unit sphere is dened as

S
n1 = n =
n

i=1
(1)
i1
x
i
r
dx
1

i
. . . dx
n
.
Notice that
n
[
S
n1 =
S
n1, and hence
_
S
n1

n
=
_
S
n1

S
n1 =
_
S
n1
dV = Vol(S
n1
) > 0.
11.2 Approximation of continuous functions by smooth ones
Theorem 11.3. Let C V be a compact domain with smooth boundary. Then any continuous
function f : C R can be C
0
-approximated by C

- smooth functions, i.e. for any > 0 there

exists a C

-smooth function g : C R such that [f(x) g(x)[ < for any x C. Moreover, if
the function f is already C

-smooth in a neighborhood of a closed subset B Int C, then one can

arrange that the function g coincides with f over B.
Lemma 11.4. There is a continuous extension of f to V .
184
Sketch of the proof. Let n be the outward normal vector eld to the boundary C. If the
boundary is C

-smooth then so is the vector eld n. Consider a map : C [0, 1] V given

by the formula (x, t) = x +t n, x C, t [0, 1]. The dierential of at the points of C 0 has
rank n. (Exercise: prove this.) Hence by the inverse function theorem for a suciently small > 0
the map is a dieomorphism of C [0, ) onto U Int C for some open neighborhood U C.
Consider a function F : C [0, ) R, dened by the formula
F(x, t) =
_
1
2t

_
f(x)
if t [0,

2
] and f(x, t) = 0 if t (

2
, ). Now we can extend f to U by the formula f(y) = F
1
(y)
if y U C, and setting it to 0 outside U.
Consider the function

=
1
_
D(0)

0,
dV

0,
,
where
0,
is a bump function dened above in (9.1.2). It is supported in the disc D

(0), non-
negative, and satises
_
D(0)

dV = 1.
. Given a continuous function f : V R we dene a function f

: V R by the formula
f

(x) =
_
f(x y)

(y)d
n
y. (11.2.1)
Then
Lemma 11.5. 1. The function f

is C

-smooth.
2. For any > 0 there exists > 0 such that for all x C we have [f(x) f

(x)[ < provided

that < .
Proof. 1. By the change of variable formula we have, replacing the variable y by u = y x:
f

(x) =
_
D(0)
f(x y)

(y)d
n
y =
_
D(x)
f(u)

(x +u)d
n
u =
_
V
f(u)

(x +u)d
n
u.
185
But the expression under the latter integral depends on x C

-smoothly as a parameter. Hence, by

the theorem about dierentiating integral over a parameter, we conclude that the function f

in
C

-smooth.
2. Fix some
0
> 0. The function f is uniformly continuous in U

0
(C). Hence there exists > 0
such that x, x

0
(C) and [[x x

[[ < we have [f(x) f(x

)[ < . Hence, for < min(

0
, )
and for x C we have
[f

(x) f(x)[ = [
_
D(0)
f(x y)

(y)d
n
y
_
D(0)
f(x)

(y)d
n
y[
_
D(0)
[f(x y) f(x)[

(y)d
n
y
_
D(0)

(y)d
n
y = . (11.2.2)
Proof of Theorem 11.3. Lemma 11.5 implies that for a suciently small the function g = f

is the required C

-smooth -approximation of the continuous function f. To prove the second part

of the theorem let us assume that f is already C

-smooth on a neighborhood U, B U C. Let

us choose a cut-o function
B,U
constructed in Lemma 9.2 and dene the required approximation
g by the formula f

+ (f f

)
B,U
.
Theorem 11.3 implies a similar theorem form continuous maps C R
n
by applying it to all
coordinate functions.
11.3 Homotopy
Let A, B be any 2 subsets of vector spaces V and W, respectively. Two continuous maps f
0
, f
1
: A
B are called homotopic if there exists a continuous map F : A[0, 1] B such that F(x, 0) = f
0
(x)
and F(x, 1) = f
1
(x) for all t [0, 1]. Notice that the family f
t
: A B, t [0, 1], dened by the
formula f
t
(x) = F(x, t) is a continuous deformation connecting f
0
and f
1
. Conversely, any such
continuous deformation f
t

t[0,1]
provides a homotopy between f
0
and f
1
.
Given a subset C A, we say that a homotopy f
t

t[0,1]
is xed over C if f
t
(x) = f
0
(x) for
all x C and all t [0, 1].
A set A is called contractible if there exists a point a A and a homotopy f
t
: A A, t [0, 1],
such that f
1
= Id and f
0
is a constant map, i.e. f
1
(x) = x for all x A and f
0
(x) = a A for all
x A.
186
Example 11.6. Any star-shaped domain A in V is contractible. Indeed, assuming that it is star-
shaped with respect to the origin, the required homotopy f
t
: A A, t [0, 1], can be dened by the
formula f
t
(x) = tx, x A.
Remark 11.7. In what follows we will always assume all homotopies to be smooth. According to
Theorem 11.3 this is not a serious constraint. Indeed, any continuous map can be C
0
-approximated
by smooth ones, and any homotopy between smooth maps can be C
0
-approximated by a smooth
homotopy between the same maps.
Lemma 11.8. Let U V be an open set, A a compact oriented manifold (possibly with boundary)
and a smooth closed dierential k-form on U. Let f
0
, f
1
: A U be two maps which are homotopic
relative to the boundary A. Then
_
A
f

0
=
_
A
f

1
.
Proof. Let F : A [0, 1] U be the homotopy map between f
0
and f
1
. By assumption d = 0,
and hence
_
A[0,1]
F

d = 0. Then, using Stokes theorem we have

0 =
_
A[0,1]
F

d =
_
A[0,1]
dF

=
_
(A[0,1])
F

=
_
A[0,1]
F

+
_
A1
F

+
_
A0
F

where the boundary (A [0, 1]) = (A 1) (A 0) (A [0, 1]) is oriented by an outward

normal vector eld n. Note that n =

t
on A 1 and n =

t
on A 0, where we denote by t
the coordinate corresponding to the factor [0, 1]. First, we notice that F

[
A[0,1]
= 0 because the
map F is independent of the coordinate t, when restricted to A [0, 1]. Hence
_
A[0,1]
F

= 0.
Consider the inclusion maps A A[0, 1] dened by the formulas j
0
(x) = (x, 0) and j
1
(x) = (x, 1).
Note that j
0
, j
1
are dieomorphisms A A 0 and A A 1, respectively. Note that the map
j
1
preserves the orientation while j
0
reverses it. We also have F j
0
= f
0
, F j
1
= f
1
. Hence,
_
A1
F

=
_
A
f

1
and
_
A0
F

=
_
A
f

0
. Thus,
0 =
_
(A[0,1])
F

=
_
A[0,1]
F

+
_
A1
F

+
_
A0
F

=
_
A
f

1

_
A
f

0
.

187
Lemma 11.9. Let A be an oriented m-dimensional manifold, possibly with boundary. Let (A)
denote the space of dierential forms on A and (A[0, 1]) denote the space of dierential forms
on the product A[0, 1]. Let j
0
, j
1
: A A[0, 1] be the inclusion maps j
0
(x) = (x, 0) A[0, 1]
and j
1
(x) = (x, 1) A[0, 1]. Then there exists a linear map K : (A[0, 1]) (A) such that
If is a k-form, k = 1, . . . , m then K() is a (k 1)-form;
d K +K d = j

1
j

0
, i.e. for each dierential k-form
k
(A [0, 1] one has dK() +
K(d) = j

1
j

0
.
Remark 11.10. Note that the rst d in the above formula denotes the exterior dierential
k
(A)

k
(A), while the second one is the exterior dierential
k
(A[0, 1])
k
(A[0, 1]).
Proof. Let us write a point in A [0, 1] as (x, t), x A, t [0, 1]. To construct K() for a given

k
(A[0, 1] we rst contract with the vector eld

t
and then integrate the resultant form
with respect to the t-coordinate. More precisely, note that any k-form on A[0, 1] can be written
as = (t) +dt (t), t [0, 1], where for each t [0, 1]
(t)
k
(A), (t)
k1
(A).
Then

t
= (t) and we dene K() =
1
_
0
(t)dt.
If we choose a local coordinate system (u
1
, . . . , u
m
) on A then (t) can be written as (t) =

1i
1
<<i

m
h
i
1
...i
k
(t)du
i
1
du
i
k
, and hence
K() =
1
_
0
(t)dt =

1i
1
<<i

m
_
_
1
_
0
h
i
1
...i
k
(t)dt
_
_
du
i
1
du
i
k
.
Clearly, K is a linear operator
k
(AI)
k1
(A).
Note that if = (t) +dt (t)
k
(A[0, 1]) then
j

0
= (0), j

1
= (1).
We further have
K() =
1
_
0
(t)dt;
d = d
UI
= d
U
(t) +dt

(t) dt d
U
(t) = d
U
(t) +dt (

(t) d
U
(t)),
188
where we denoted

(t) :=
(t)
t
and I = [0, 1]. Here the notation d
UI
stands for exterior dierential
on (U I) and d
U
denotes the exterior dierential on (U). In other words, when we write d
U
(t)
we view (t) as a form on U depending on t as a parameter. We do not write any subscript for d
when there could not be any misunderstanding.
Hence,
K(d) =
1
_
0
(

(t) d
U
(t))dt = (1) (0)
1
_
0
d
U
(t)dt;
d(K()) =
1
_
0
d
U
(t)dt.
Therefore,
d(K()) +K(d()) = (1) (0)
1
_
0
d
U
(t)dt +
1
_
0
d
U
(t)dt
= (1) (0) = j

1
() j

0
().

Theorem 11.11. (Poincar es lemma) Let U be a contractible domain in V . Then any closed form
in U is exact. More precisely, let F : U [0, 1] U be the contraction homotopy to a point a U,
i.e. F(x, 1) = x, F(x, 0) = a for all x U. Then if a closed k-form in U then
= dK(F

),
where K :
k+1
(U [0, 1])
k
(U) is an operator constructed in Lemma 11.9.
Proof. Consider a contraction homotopy F : U [0, 1] U. Then F j
0
(x) = a U and
F j
1
(x) = x for all x U. Consider an operator K : U) (U) constructed above. Thus
K d +d K = j

1
j

0
.
Let be a closed k-form on U. Denote := F

. Thus is a k-form on U [0, 1]. Note that

d = dF

= F

d = 0, j

1
= (F j
1
)

= and j

0
= (F j
0
)

= 0. Then, using Lemma 11.9

we have
K(d) +dK() = dK() = j

1
j

0
= , (11.3.1)
189
i.e. = dK(F

).
In particular, any closed form is locally exact.
Example 11.12. Let us work out explicitly the formula for a primitive of a closed 1-form in a
star-shaped domain U R
n
. We can assume that U is star-shaped with respect to the origin.
Let =
n

1
f
i
dx
i
be a closed 1-form. Then according to Theorem 11.11 we have = dF, where
F = K(

), where : U [0, 1] U is a contraction homotopy, i.e. (x, 1) = x, (x, 1) = 0 for

x U. can be dened by the formula (x, t) = tx, x U, t [0, 1]. Then

=
n

1
f
i
(tx)d(tx
i
) =
n

1
tf
i
(tx)dx
i
+
n

1
x
i
f
i
(tx)dt.
Hence,
K() =
1
_
0

=
1
_
0
_
n

1
x
i
f
i
(tx)
_
dt.
Note that this expression coincides with the expression in formula (9.4.1) in Section 9.4.
Exercise 11.13. Work out an explicit expression for a primitive of a closed 2-form = Pdy dz +
Qdz dx +Rdx dy on a star-shaped domain U R
3
.
Example 11.14. 1. R
n
+
= x
1
0 R
n
is not dieomorphic to R
n
. Indeed, suppose there exists
such a dieomorphism f : R
n
+
R
n
. Denote a := f(0). Without loss of generality we can assume
that a = 0. Then

f = f[
R
n
+
\0
is a dieomorphism R
n
+
0 R
n
0. But R
n
+
0 is star-shaped with
respect to any point with positive coordinate x
1
, and hence it is contractible. In particular any
closed form on R
n
+
0 is exact. On the other hand, we exhibited above in 11.2 a closed (n1)-form
on R
n
0 which is not exact.
2. Borsuks theorem: There is no continuous map D
n
D
n
which is the identity on D
n
. We
denote here by D
n
the unit disc in R
n
and by D
n
its boundary (n 1)-sphere.
Proof. Suppose that there is such a map f : D
n
D
n
. One can assume that f is smooth.
Indeed, according to Theorem 11.3 one can approximate f by a smooth map, keeping it xed on
the boundary where it is the identity map, and hence smooth. Take the closed non-exact form
n
from Example 11.2 on D
n
. Then
n
= f

n
is a closed (n 1)-form on D
n
which coincides with
190

n
on D
n
. D
n
is star-shaped, and therefore
n
is exact,
n
= d. But then
n
= d([
D
n) which
is a contradiction.
3. Brouwers fixed point theorem: Any continuous map f : D
n
D
n
has at least 1 xed point.
Proof. Suppose f : D
n
D
n
has no xed points. Let us dene a map F : D
n
D
n
as follows.
For each x D
n
take a ray r
x
from the point f(x) which goes through x till it intersects D
n
at a
point which we will denote F(x). The map is well dened because for any x the points x and f(x)
are distinct. Note also that if x D
n
then the ray r
x
intersects D
n
at the point x, and hence
F(x) = x in this case. But existence of such F is ruled out by Borsuk theorem.
k-connected manifolds
A subset A V is called k-connected, k = 0, 1, . . . , if for any m k any two continuous maps
of discs f
0
, f
1
: D
m
A which coincide along D
m
are homotopic relative to D
m
. Thus, 0-
connectedness is equivalent to path-connectedness. 1-connected submanifolds are also called simply
connected.
Exercise 11.15. Prove that k-connectedness can be equivalently dened as follows: A is k-connected
if any map f : S
m
A, m k is homotopic to a constant map.
Example 11.16. 1. If A is contractible then it is k-connected for any k. For some classes of
subsets, e.g. submanifolds, the converse is also true (J.H.C Whiteheads theorem) but this is
a quite deep and non-trivial fact.
2. The n-sphere S
n
is (n1)-connected but not n-connected. Indeed, to prove that S
n1
simply
connected we will use the second denition. Consider a map f : S
k
S
n
. We rst notice
that according to Theorem 11.3 we can assume that the map f is smooth. Hence, according
to Corollary 9.20 Vol
n
f(S
k
) = 0 provided that k < n. In particular, there exists a point
p S
n
f(S
k
). But the complement of a point p in S
n
is dieomorphic to S
n
vis the
stereographic projection from the point p. But R
n
is contractible, and hence f is homotopic
to a constant map. On the other hand, the identity map Id : S
n
S
n
is not homotopic to
a constant map. Indeed, we know that there exists a closed n-form on S
n
, say the form
n
191
from Example 11.2, such that
_
S
n

n
,= 0. Hence,
_
S
n
Id

n
,= 0. On the other hand if Id were
homotopic to a constant map this integral would vanish.
Exercise 11.17. Prove that R
n+1
0 is (n 1) connected but not n-connected.
Proposition 11.18. Let U V be a m-connected domain. Then for any k m any closed
dierential k-form in U is exact.
Proof. We will prove here only the case m = 1. Though the general case is not dicult, it requires
developing certain additional tools. Let be a closed dierential 1-form. Choose a reference point
b U. By assumption, U is path-connected. Hence, any other point x can be connected to b by
a path
x
: [0, 1] U, i.e.
x
(0) = b,
x
(1) = x. Let us dene the function F : U R by the
formula F(x) =
_
x
. Note that due to the simply-connectedness of the domain U, any : [0, 1] U
connecting b and x is homotopic to
x
relative its ends, and hence according to Lemma 11.8 we
have
_
x
=
_

. Thus the above denition of the function F is independent of the choice of paths

x
. We claim that the function F is dierentiable and dF = . Note that if the primitive of exists
than it has to be equal to F up to an additive constant. But we know that in a suciently small
ball B

(a) centered at any point a U there exists a primitive G of . Hence, G(x) = F(x) +const,
and the the dierentiability of G implies dierentiablity of F and we have dF = dG = .
11.4 Winding and linking numbers
Given a loop : S
1
R
2
0 we dene its winding number around 0 as the integral
w() =
1
2
1
_
S

2
=
1
2
_
S
1
xdy ydx
x
2
+y
2
,
where we orient S
1
as the boundary of the unit disc in R
2
. For instance, if j : S
1
R
2
is the
inclusion map then w(j) = 1. For the loop
n
parameterized by the map t (cos nt, sinnt), t
[0, 2] we have w(
n
) = n.
Proposition 11.19. 1. For any loop the number w() is an integer.
2. If loops
0
,
1
: S
1
R
2
0 are homotopic in R
2
0 then w(
0
) = w(
1
).
192
Figure 11.1: w() = 2.
3. w() = n then the loop is homotopic (as a loop in R
2
0) to the loop
n
: [0, 1] R
2
0
given by the formula
n
(t) = (cos 2nt, sin2nt).
Proof. 1. Let us dene the loop parametrically in polar coordinates:
r = r(s), = (s), s [0, 1],
where r(0) = r(1) and (1) = (0) + 2n. The form
2
in polar coordinates is equal to d, and
hence
_

=
1
2
1
_
0

(s)ds =
(1) (0)
2
= n.
2. This is an immediate corollary of Proposition 11.8.
3. Let us write both loops and
n
in polar coordinates. Respectively,we have r = r(t), = (s)
for and r = 1, = 2ns for
n
, s [0, 1]. The condition w() = n implies, in view of part 1, that
(1) = (0) + 2n. Then the required homotopy
t
, t [0, 1], connecting the loops
0
= and

1
=
n
can be dened by the parametric equations r = (1t)r(s)+t, =
t
(s) = (1t)(s)+2nst.
Note that for all t [0, 1] we have
t
(1) =
t
(0) + 2n. Therefore,
t
is a loop for all t [0, 1].
193
Figure 11.2: l(
1
,
2
) = 1.
Given two disjoint loops , : S
1
R
3
(i.e. (s) ,= (t) for any s, t S
1
) consider a map
F
,
: T
2
R
3
0, where T
2
= S
1
S
1
is the 2-torus, dened by the formula
F
,
(s, t) = (s) (t).
Then the number
l(, ) :=
1
4
2
_
T
F

3
=
1
4
2
_
T
F

,
_
xdy dz +ydz dx +zdx dy
(x
2
+y
2
+z
2
)
3
2
_
is called the linking number of loops , .
1
Exercise 11.20. Prove that
1. The number l(, ) remains unchanged if one continuously deforms the loops , keeping
them disjoint;
2. The number l(, ) is an integer for any disjoint loops , ;
1
This denition of the linking number is due to Carl Friedrich Gauss.
194
3. l(, ) = l(, );
4. Let (s) = (cos s, sins, 0), s [0, 2] and (t) = (1 +
1
2
cos t, 0,
1
2
sint), t [0, 2]. Then
l(, ) = 1.
11.5 Properties of k-forms on k-dimensional manifolds
A k-form on k-dimensional submanifold is always closed. Indeed, d is a (k +1)-form and hence
it is identically 0 on a k-dimensional manifold.
Remark 11.21. Given a k-dimensional submanifold A V , and a k-form on V , the dierential
d
x
does not need to vanish at a point x A. However, d
x
[
Tx(A)
does vanish.
The following theorem is the main result of this section.
Theorem 11.22. Let A V be an orientable compact connected k-dimensional submanifold,
possibly with boundary, and a dierential k-form on A.
1. Suppose that A ,= . Then is exact, i.e. there exists a (k 1)-form on A such that
d = .
2. Suppose that A is closed, i.e. A = . Then is exact if and only if
_
A
= 0.
To prove Theorem 11.22 we will need a few lemmas.
Lemma 11.23. Let I
k
be the k-dimensional cube 1 x
j
1, j = 1, . . . , k.
1. Let be a dierential k-form on I
k
such that
Supp()
_
0 I
k1
[0, 1] I
k1
_
= .
Then there exists a (k1)-form such that d = and such that Supp()
_
1 I
k1
[1, 1] I
k1
_
=
.
2. Let be a dierential k-form on I
k
such that Supp() Int I
k
and
_
I
k
= 0. Then there exists
a (k 1)-form such that d = and Supp() Int I
k
.
195
Proof. We have
= f(x
1
, . . . , x
k
)dx
1
dx
k
,
In the rst case of the lemma the function f vanishes on 0 I
k1
[1, 1] I
k1
. We will look
for in the form
= g(x
1
, . . . , x
k
)dx
2
dx
k
.
Then
d =
g
x
1
(x
1
, . . . , x
k
)dx
1
dx
2
dx
k
.
and hence the equation d = is equivalent to
g
x
1
(x
1
, . . . , x
k
) = f(x
1
, . . . , x
k
).
Hence, if we dene
g(x
1
, . . . , x
k
) :=
x
1
_
1
f(u, x
2
, . . . , x
k
)du,
then the form = g(x
1
, . . . , x
k
)dx
2
dx
k
has the required properties.
The second part of the lemma we will prove here only for the case k = 2. The general case can
be handled similarly by induction over k. We have in this case Supp(f) Int I
2
and
_
I
2
fdS = 0.
Let us denote h(x
2
) :=
1
_
1
f(x
1
, x
2
)dx
1
. Note that h(u) = 0 if u is suciently close to 1 or 1.
According to Fubinis theorem,
1
_
1
h(x
2
)dx
2
= 0. We can assume that f(x
1
, x
2
) = 0 for x
1
1 ,
and hence
u
_
1
f(x
1
, . . . , x
k1
, t)dt = h(x
1
, . . . , x
k1
) for u [1 , 1]. Consider any non-negative
C

-function : [1, 1] R such that (u) = 1 for u [1, 1

2
3
] and (u) = 0 for u [1

3
, 1].
Dene a function g
1
: I
2
R by the formula
g
1
(x
1
x
2
) =
_

_
x
1
_
1
f(u, x
2
)du, x
1
[1, 1 ],
h(x
2
)(x
1
), x
1
(1 , 1].
Denote
1
= g
1
(x
1
, x
2
)dx
2
. Then d = on [1, 1 ] [0, 1] and d
1
= h(x
2
)

(x
1
)dx
1
dx
2
on
[1 , 1] [0, 1]. Note that Supp(
1
) Int I
2
.
196
Let us dene
g
2
(x
1
, x
2
) :=
_

_
0, x
1
[1, 1 ],

(x
1
)
x
2
_
1
h(u)du, x
1
(1 , 1]
and denote
2
= g
2
(x
1
, x
2
)dx
1
. Then d
2
= 0 on [1, 1 ] [1, 1] and
d
2
= h(x
2
)

(x
1
)dx
1
dx
2
on [1, 1][1, 1]. Note that g
2
(x
1
, 1) =

(x
1
)
1
_
1
h(u)du = 0. Taking into account that h(u) = 0
when u is suciently close to 1 or 1 we conclude that h(x
1
, x
2
) = 0 near I
2
, i.e. Supp(
2
) Int I
2
.
Finally, if we dene =
1
+
2
then we have d = and Supp() Int I
2
.
The following lemma is a special case of the, so-called, tubular neighborhood theorem.
Lemma 11.24. Let A V be a compact k-dimensional submanifold with boundary. Let :
[1, 1] A be an embedding such that (1) A,

(1) T
(1)
(A) and ([0, 1)) Int A.
Then the embedding can be extended to an embedding : [1, 1] I
k1
A such that
(t, 0) = (t), for t [1, 1], 0 I
k1
;
(1 I
k1
) A, ([1, 1) I
k1
) Int A;

t
(1, x) / T(A) for all x I
k1
.
There are many ways to prove this lemma. We will explain below one of the arguments.
Proof. Step 1. We rst construct k 1 ortonormal vector elds
1
, . . . ,
k
along = ([1, 1])
which are tangent to A and normal to . To do that let us denote by N
u
the normal (k 1)-
dimensional space N
u
to T
u
in T
u
A. Let us observe that in view of compactness of there is an
> 0 with the following property: for any two points u = (t), u

= (t

) , t, t

[1, 1], such

that [t t

[ the orthogonal projection N

u
N
u
is non-degenerate (i.e. is an isomorphism).
Choose N <
1
2
and consider points u
j
= (t
j
), where t
j
= 1 +
2j
N
, j = 1, . . . N. Choose any
orthonormal basis
1
(0), . . . ,
k
(0) N
u
0
, parallel transport these vectors to all points of the arc

1
= ([1, t
1
]), project them orthogonally to the normal spaces N
u
in these points, and then
orthonormalize the resulted bases via the Gram-Schmidt process. Thus we constructed orthonormal
vector elds
1
(t), . . .
k
(t) N
(t)
, t [1, t
1
]. Now we repeat this procedure beginning with the
197
basis
1
(t
1
), . . .
k
(t
1
) N
(t
1
)
= N
u
1
and extend the vector elds
1
, . . . ,
k
to
2
= ([t
1
, t
2
]).
Continuing this process we will construct the orhonormal vector elds
1
, . . . ,
k
along the whole
curve .
2
Step 2. Consider a map : [1, 1] I
k1
V given by the formula
(t, x
1
, . . . , x
k1
) = (t) +
k1

1
x
j

j
(t), t, x
1
, . . . , x
k1
[1, 1],
where a small positive number will be chosen later. The map is an embedding if is chosen
small enough.
3
Unfortunately the image ([1, 1] I
k1
) is not contained in A. We will correct
this in the next step.
Step 3. Take any point a A and denote by
a
the orthogonal projection V T
a
A. Let us make
the following additional assumption (in the next step we will show how to get rid of it): there exists
a neighborhood U a = (1) in A such that
a
(U) N
a
T
a
A. Given > 0 let us denote by
B

(a) the (k 1)-dimensional ball of radius in the space N

T
a
A. In view of compactness of A
one can choose an > 0 such that for all points a there exists an embedding e
a
: B

(a) A
such that
a
e
a
= Id. Then for a suciently small <

k1
the map

: [1, 1] I
k1
A
dened by the formula

(t, x) = e
(t)
(t, x), t [1, 1], x I
k1
is an embedding with the required properties.
Step 4 It remains to show how to satisfy the additional condition at the boundary point (1)
A which were imposed above in Step 3. Take the point a = (1) A. Without loss of
generality we can assume that a = 0 V . Choose an orthonormal basis v
1
, . . . , v
n
of V such that
v
1
, . . . , v
k
N
a
and v
k
is tangent to and pointing inward . Let (y
1
, . . . , y
n
) be the corresponding
cartesian coordinates in V . Then there exists a neighborhood U a in A which is graphical in
these coordinates and can be given by
y
j
=
j
(y
1
, . . . , y
k
), j = k + 1, . . . , n, y
k

k
(y
1
, . . . , y
k1
), [y
i
[ , i = 1, . . . , k 1,
2
Strictly speaking, the constructed vector elds only piece-wise smooth, because we did not make any special
precautions to ensure smoothness at the points uj, j = 1, . . . , N1. This could be corrected via a standard smoothing
procedure.
3
Exercise: prove it!
198
where all the rst partial derivatives of the functions
k
, . . . ,
n
vanish at the origin. Take a C

cut-o function : [0, )] R which is equal to 1 on [0,

1
2
] and which is supported in [0, 1] (see
Lemma 9.2). Consider a map F given by the formula
F(y
1
, . . . , y
n
) = (y
1
, . . . , y
k1
, y
k

k
(y
1
, . . . , y
k1
)
_
[[y[[

_
, y
k+1
, . . . , y
n
).
For a suciently small > 0 this is a dieomorphism supported in an -ball in V centered in the
origin. On the other hand, the manifold

A = F(A) satises the extra condition of Step 3.
Lemma 11.25. Let A V be a (path)-connected submanifold with a non-empty boundary. Then
for any point a A there exists an embedding
a
: [1, 1] A such that
a
(0) = a,
a
(1) A
and

a
(1) T
a(1)
(A).
Sketch of the proof. Because A is path-connected with non-empty boundary, any interior point
can be connected by a path with a boundary point. However, this path need not be an embedding.
First, we perturb this path to make it an immersion : [1, 1] A, i.e. a map with non-vanising
derivative. This can be done as follows. As in the proof of the previous lemma we consider a su-
ciently small partition of the path, so that two neighboring subdivision points lie in a coordinate
neighborhood. Then we can connect these points by a straight segment in these coordinate neigh-
borhoods. Finally we can smooth the corners via the standard smoothing procedure. Unfortunately
the constructed immersed path may have self-intersection points. First, one can arrange that
there are only nitely many intersections, and then cut-out the loops, i.e. if (t
1
) = (t
2
) for
t
1
< t
2
we can consider a new piece-wise smooth path which consists of [
[1,t
1
]
and [
[t
2
,1]
The
new path has less self-intersection points, and thus continuing by induction we will end with a
piece-wise smooth embedding. It remains to smooth again the corners.
Proof of Theorem 11.22. 1. For every point a A choose an embedding
a
: [1, 1] A, as in
Lemma 11.25, and using Lemma 11.24 extend
a
to an embedding
a
: [1, 1] I
k1
A such
that
-
a
(t, 0) = (t), for t [1, 1], 0 I
k1
;
-
a
(1 I
k1
) A,
a
([1, 1) I
k1
) Int A;
-
a
t
(1, x) / T(A) for all x I
k1
.
199
Due to compactness of A we can choose nitely many such embeddings
j
=
a
j
, j = 1, . . . , N,
such that
N

j
((1, 1] Int(I
k1
) = A. Choose a partition of unity subordinated to this covering
and split the k-form as a sum =
K

j
, where each
i
is supported in
j
((1, 1] Int(I
k1
))
for some j = 1, . . . , N. To simplify the notation we will assume that N = K and each
j
is
supported in
j
((1, 1] Int(I
k1
)), j = 1, . . . , N. Consider the pull-back form
j
=

j
on
I
k
= [1, 1]I
k1
. According to Lemma 11.23.1 there exists a (k1)-form

j
such that Supp(

j
)
(1, 1] Int(I
k1
) and d

j
=
j
. Let us transport the form

j
back to A. Namely, set
j
equal to
(
1
j
)

j
on
j
((1, 1] Int(I
k1
)) A and extend it as 0 elsewhere on A. Then d
j
=
j
, and
hence d(
N

j
) =
N

j
= .
2. Choose a point a A and parameterize a coordinate neighborhood U A by an embedding
: I
k
A such that (0) = a. Take a small closed ball D

(0) I
k
R
k
and denote

D = (D

(0)).
Then

A = A Int

D is a submanifold with non-empty boundary, and

A =

D. Let us use part

1 of the theorem to construct a form

on

A such that d

= [
e
A
. Let us extent the form

in
any way to a form, still denoted by

on the whole submanifold A. Then d

= + where
Supp()

D Int (I
k
). Note that
_
(I
k
)
=
_
A
=
_
A

_
A
d

= 0
because
_
A
= 0 by our assumption, and
_
A
d

= 0 by Stokes theorem. Thus,

k
_
I

= 0, and
hence, we can apply Lemma 11.23.2 to the form

on I
k
and construct a (k 1)-form on
I
k1
such that d =

and Supp() Int I

k
. Now we push-forward the form to A, i.e. take
the form

on A which is equal to (
1
)

on (I
k
) and equal to 0 elsewhere. Finally, we have
d(

) = d

+ = , and hence =

+

is the required primitive of on A.

Corollary 11.26. Let A be an oriented compact connected k-dimensional submanifold with non-
empty boundary and a dierential k-form on A from Theorem 11.22. Then for any smooth map
f : A A such that f[
A
= Id we have
_
A
f

=
_
A
.
200
Proof. According to Theorem 11.22.1 there exists a form such that = d. Then
_
A
f

=
_
A
f

d =
_
A
df

=
_
A
f

=
_
A
=
_
A
.

Degree of a map
Consider two closed connected oriented submanifolds A V , B W of the same dimension k. Let
be an n-form on B such that
_
B
= 1. Given a smooth map f : A B the integer deg(f) :=
_
A
f

is called the degree of the map f.

Proposition 11.27. 1. Given any two k-forms on B such
_
B
=
_
B
we have
_
A
f

=
_
A
f

,
for any smooth map f : A B, and thus deg(f) is independent of the choice of the form
on B with the property
_
A
= 1.
2. If the maps f, g : A B are homotopic then deg(f) = deg(g).
3. Let b B be a regular value of the map f. Let f
1
(b) = a
1
, . . . , a
d
. Then
deg(f) =
d

1
sign(det Df(a
j
)).
In particular, deg(f) is an integer number.
Proof. The second part follows from Lemma 11.8. To prove the rst part, let us write = +,
where
_
B
= 0. Using Theorem 11.22.2 we conclude that = d for some (k 1)-form on B.
Then
_
A
f

=
_
A
f

+
_
A
f

=
_
A
f

+
_
A
df

=
_
A
f

.
Let us prove the last statement of the theorem. By the inverse function theorem there exists a
neighborhood U b in B and neighborhoods U
1
a
1
, . . . , U
d
a
d
in A such that the restrictions
of the map f to the neighborhoods U
1
, . . . , U
d
are dieomorphisms f[
U
j
: U
j
U, j = 1, . . . , d. Let
us consider a form on B such that Supp U and
_
B
=
_
U
= 1. Then
deg(f) =
_
A
f

=
d

1
_
U
j
f

=
d

1
sign(det Df(a
j
)),
201
because according to Theorem ?? we have
_
U
j
f

= sign(det Df(a
j
))
_
U
= sign(det Df(a
j
)).
for each j = 1, . . . , d.
Remark 11.28. Any continuous map f : A B can be approximated by a homotopic to f smooth
map A B, and any two such smooth approximations of f are homotopic. Hence this allows us
to dene the degree of any continuous map f : A B.
Exercise 11.29. 1. Let us view R
2
as C. In particular, we view the unit sphere S
1
= S
1
1
(0) as the
set of complex numbers of modulus 1:
S
1
= z C; [z[ = 1.
Consider a map h
n
: S
1
S
1
given by the formula h
n
(z) = z
n
, z S
1
. Then deg(h
n
) = n.
2. Let f : S
n1
S
n1
be a map of degree d. Let p

be the north and south poles of S

n+1
, i.e.
p

= (0, . . . , 0, 1). Given any point x = (x

1
, . . . , x
n+1
) S
n
p
+
, p

we denote by (x) the

point
1

1
x
2
j
(x
1
, . . . , x
n
) S
n1
and dene a map f : S
n
S
n
by the formula
f(x) =
_

_
p

, if x = p

,
_
n

1
x
2
j
f((x)), x
n+1
_
, if x ,= p

.
Prove that deg((f)) = d.
4
3. Prove that two maps f, g : S
n
S
n
are homotopic if and only if they have the same degree.
In particular, any map of degree n is homotopic to the map h
n
. (Hint: For n=1 this follows from
Proposition 11.19. For n > 1 rst prove that any map is homotopic to a suspension.)
4. Give an example of two non-homotopic orientation preserving dieomorphisms T
2
T
2
. Note
that the degree of both these maps is 1. Hence, for manifolds, other than spheres, having the same
degree is not sucient for their homotopy.
4
The map f is called the suspension of the map f.
202
5. Let , : S
1
R
3
be two disjoint loops in R
3
. Consider a map

F
,
: T
2
S
2
dened by the
formula

F
,
(s, t) =
(s) (t)
[[(s) (t)[[
, s, t S
1
.
Prove that l(, ) = deg(

F
,
). Use this to solve Exercise 11.20.4 above.
203