Differential Geometry

A course for Master students

Hacen Zelaci

Mathematical department, University of El Oued.

Current address: El Oued - Algeria
E-mail address: [email protected]
 ABSTRACT 3

Abstract
A branch of geometry dealing with geometrical forms, mainly with curves and
surfaces, by methods of mathematical analysis. In differential geometry the prop-
erties of curves and surfaces are usually studied on a small scale, i.e. the study
concerns properties of sufficiently small pieces of them. Properties of families of
curves and surfaces are also studied.
Differential geometry arose and developed in close connection with mathemat-
ical analysis, the latter having grown, to a considerable extent, out of problems
in geometry. Many geometrical concepts were defined prior to their analogues in
analysis. For instance, the concept of a tangent is older than that of a derivative,
and the concepts of area and volume are older than that of the integral.
Differential geometry first appeared in the 18th century and is linked with the
names of L. Euler and G. Monge. The first synoptic treatise on the theory of
surfaces was written by Monge (Une application d’analyse à la géométrie, 1795).
In 1827 a study under the (English) title A general study on curved surfaces was
published by C.F. Gauss; this study laid the foundations of the theory of surfaces in
its modern form. From that time onwards differential geometry ceased to be a mere
application of analysis, and has become an independent branch of mathematics.
The discovery of non-Euclidean geometry by N.I. Lobachevskii in 1826 played
a major role in the development of geometry as a whole, including differential ge-
ometry. Lobachevskii rejected in fact the a priori concept of space, which was
predominating in mathematics and in philosophy. He found that spaces different
from Euclidean spaces exist. This idea of Lobachevskii was reflected in numerous
mathematical studies. Thus, in 1854 B. Riemann published his course ber die Hy-
pothesen, welche der Geometrie zuGrunde liegen and thus laid the foundations of
Riemannian geometry, the application of which to higher-dimensional manifolds is
related to the geometry of n-dimensional space similarly as the relation between
the interior geometry of a surface and Euclidean geometry on a plane.

Historically, differential geometry arose and developed as a result of and in

connection to the mathematical analysis of curves and surfaces. Mathematical
analysis of curves and surfaces had been developed to answer some of the nagging
and unanswered questions that appeared in calculus, like the reasons for relation-
ships between complex shapes and curves, series and analytic functions. These
unanswered questions indicated greater, hidden relationships. The general idea of
natural equations for obtaining curves from local curvature appears to have been
first considered by Leonhard Euler in 1736, and many examples with fairly simple
behavior were studied in the 1800s. When curves, surfaces enclosed by curves, and
points on curves were found to be quantitatively, and generally, related by mathe-
matical forms, the formal study of the nature of curves and surfaces became a field
of study in its own right, with Monge’s paper in 1795, and especially, with Gauss’s
publication of his article, titled ”Disquisitiones Generales Circa Superficies Curvas”,
in Commentationes Societatis Regiae Scientiarum Gottingesis Recentiores in 1827.
Initially applied to the Euclidean space, further explorations led to non-Euclidean
space, and metric and topological spaces.
4

This course gives an introduction to this domain, it doesn’t requires any back-
ground in this mathematical discipline.

Our main references are [Spi99], [RS18],[Oan18], [Pet], [Sch08], [Gua18],

[Csi14].
 Contents

Abstract 3
Chapter 1. Introduction 7
Chapter 2. Differential manifolds 9
1. Abstract manifolds 9
2. Tangent space 11
3. Vector fields and flows 13
4. Parallel transport 16
5. Orientability 17
6. Tensors 19
7. Differential forms 21
8. Integration of differential forms 23
Chapter 3. Vector bundles 25
Chapter 4. Riemannian manifolds 27
Chapter 5. Exercises 29
1. Solutions 31
Bibliography 35

5
6 CONTENTS

notations: Through out this work, we use the following general notations:
• N, Z, Q, R, C for the set of nonnegative integers, integers, the fields of
rational numbers, real numbers, complex numbers respectively.
• Maps are generally denoted by f, g, h....
• Integers are denoted by k, n, m....
• We say that a function f is C k if it is k−times differentiable and its
k th −differential is continuous.
• The differential of a function f at p is denoted dp f .
• The Jacobian of a function f at p is denoted Jp (f ).
 CHAPTER 1

Introduction

Recall the definition of a submanifold of Rn .

Definition 0.1. Let M ⊂ Rn be a subset. We say that M is a submanifold of
dimension d (6 n) and of class C k if for all x in M , there is a neighborhood U of
x in Rn , a neighborhood V of 0 in Rn and a C k -diffeomorphism φ : U → V such
that φ(U ∩ M ) = V ∩ (Rd × {0}).
Let U ⊂ Rn be an open set and f : U → Rm be a smooth function. An element
c ∈ Rm is called a regular value of f if, for all p ∈ U , we have
f (p) = c =⇒ dp f : Rn → Rm is surjective.
We have
Proposition 0.2. A subset M ⊂ Rn is a submanifold iff for any p ∈ M ,
there exists an open set U ⊂ Rn and a smooth map f : U → Rn−m such that
p ∈ U ,U ∩ M = f −1 (0) and 0 is a regular value for f .
Recall that a vector v ∈ Rn is called a tangent vector of M at p if there exists
a smooth curve γ :] − 1, 1[→ M such that
γ(0) = p, γ 0 (0) = v.
The set of tangent vectors to M at p is called the linear tangent space of M at p
and it is denoted Tp M . The affine tangent space of M at p is the translate of Tp M
−
→
with the vector 0p.
In the notation of the above proposition, we have Tp M = Ker(dp f ).
A vector field on a smooth manifold M is a smooth map
X : M → T M,
such that for any p ∈ M , X(p) ∈ Tp M . In other words, X is a section for the
canonical projection T M → M . We denote by Γ(M ) the space of vector fields.
Let X be a vector field on M and I ⊂ R an interval. A smooth map γ : I → M
is called an integral curve for X if for any t ∈ I we have
γ 0 (t) = X(γ(t)).

7
 CHAPTER 2

Differential manifolds

1. Abstract manifolds
Let T be a topological space. A neighborhood of a point x ∈ T is a subset
U ⊂ T with the property that it contains an open set containing x. A map f :
T → W which is continuous, bijective and has a continuous inverse is called a
homeomorphism.
Definition 1.1. A topological space X is said to be Hausdorff if for every pair
of distinct points x, y ∈ X there exist disjoint neighborhoods of x and y.
Let M be a Hausdorff topological space, and let m > 0 be a fixed non negative
integer.
Definition 1.2. An m−dimensional smooth atlas of M is a collection (Oi )i∈I
of open sets Oi ⊂ M such that M = ∪i∈I Oi , together with a collection (Ui )i∈I of
open sets in Rm and a collection of homeomorphisms, called charts, φi : Oi → Ui =
φi (Ui ), with the following property of smooth transition on overlaps:
For each pair i, j ∈ I the map φj ◦ φ−1i : φi (Oi ∩ Oj ) → φj (Oi ∩ Oj ) is smooth.
Two smooth atlases are compatible if their union is again an atlas.
If A is an atlas, then so is the collection A¯ of all charts compatible with each
member of A . The atlas A¯ is obviously maximal. In other words, every atlas
extends uniquely to a maximal atlas.
Example 1.3. The map
ϕ(u, v) = (cos v, sin v, u), u, v ∈ R
is smooth and covers the cylinder S = {(x, y, z) ∈ R3 |x2 + y 2 = 1}, but it is not
injective. Let
U1 = {(u, v) ∈ R2 | − π < v < π} , U2 = {(u, v) ∈ R2 |0 < v < 2π},
and let ϕi denote the restriction of ϕ to Ui for i = 1, 2. Then ϕ1 and ϕ2 are
both injective, ϕ1 covers S with the exception of a vertical line on the back where
x = −1, and ϕ2 covers S with the exception of a vertical line on the front where
x = 1. Together they cover the entire set and thus they constitute an atlas.
Example 1.4. Assume that W ⊂ Rm and V ⊂ Rn are open non-empty sets,
and let f : Rm → Rn be a continuous map. Denote by M the graph of f , i.e.
M = {(x, y)| y = f (x)}.
Let U = (W × V ) ∩ M and let φ(x, y) = x be the projection of U onto W . Then
the pair (φ, U ) is a chart on M . The inverse map is given by φ−1 (x) = (x, f (x)).

9
10 2. DIFFERENTIAL MANIFOLDS

Definition 1.5. An abstract manifold (or just a manifold) of dimension m, is

a Hausdorff topological space M , equipped with a maximal m−dimensional smooth
atlas.
Remark 1.6. If in the definition, we don’t assume that M is a topological
space, and that Oi are just subset, then there is a unique topology on M making
Oi open and φi homeomorphisms.
Example 1.7. The stereographic projections endows the spheres S n with an at-
las given as follows; let U± be the subsets of S n obtained by removing (∓1, 0, · · · , 0).
Stereographic projection defines bijections φ± : U± → Rn given by
x1 xn
φ± (x0 , · · · , xn ) = ( ,··· , ).
1 ± x0 1 ± x0
One can check (exercise!) that
x
φ− ◦ φ−1 + (x) = ,
kxk2
hence it is smooth. The same is true for φ+ ◦ φ−1
+ . Hence {(φ+ , U+ ), (φ− , U− )} is
an atlas on S n .
Definition 1.8. Let f : M → N be a map between abstract manifolds. Then
f is called smooth if for each p ∈ M there exists a chart φ : O → U around p,
and a chart ψ : O0 → V around f (p), such that f (O) ⊂ O0 and such that the
coordinate expression ψ ◦ f ◦ φ−1 is smooth. A bijective map f : M → N , is called
a diffeomorphism if f and f −1 are both smooth.
Notice that a smooth map f : M → N is continuous. This follows immediately
from the definition above, by writing f = ψ −1 ◦ (ψ ◦ f ◦ φ−1 ) ◦ φ in a neighborhood
of each point. Also, every chart φ : O → U is a smooth diffeomorphism. And every
diffeomorphism φ of a non-empty open subset V ⊂ Rm onto an open subset in M
is a chart on M .

Now, an atlas of an abstract manifold is said to be countable if the set of charts

in the atlas is countable
Lemma 1.9. Let M be an abstract manifold. Then M has a countable atlas
if and only if there exists a countable base for the topology in M . In particular, a
submanifold of Rn has a countable atlas.
Proof. Assume M has a countable atlas. Let φ : O → U ⊂ Rm be a chart,
then there is a countable base for the topology of U . This induces a countable base
for the topology on O, because φ is a homeomorphism. The collection of these
bases for all the charts in the atlas gives a base for the topology of M .
Conversely, suppose that there exists a countable base (Vk )k of the topology of M .
For each k, choose a chart φ : O → U , such that Vk ⊂ O (if it exists!). This gives
a countable atlas A in M . Indeed, it is clearly countable. Let now x ∈ M . Then
there exists a chart ψ : O0 → U 0 such that x ∈ O0 , and there exists k such that
x ∈ Vk ⊂ O0 . Hence there exists a chart φ : O → U in A such that x ∈ O. So A is
an atlas. 
Theorem 1.10 (Whitney theorem). Let M be an abstract smooth manifold of
dimension d, and assume that there exists a countable atlas for M . Then there
exists a diffeomorphism of M onto a submanifold in R2d .
 2. TANGENT SPACE 11

Example 1.11 (Projective space). Let V be a finite dimensional vector space,

The projective space of V is the set of lines (through the origin) in V . In other
words,
P (V ) = {l ⊂ V |l is a 1-dimensional R-linear subspace}.
It can be seen a P (V ) = V ∗ / ∼, where the equivalence relation is given by
u ∼ v ⇔ u = λv for some λ ∈ R.
If V = R , then P (V ) is denoted simply Pn . The projective space Pn can be given
n

the structure of an abstract n−dimensional manifold as follows. Let π : x → [x]

denote the natural map Rn+1 r {0} → Pn , and let S n ⊂ Rn denote the unit sphere.
A set U ⊂ Pn is declared to be open if and only if its preimage π −1 (U ) is open in
Rn+1 (or equivalently, if π −1 (U ) ∩ S n is open in S n ). This makes Pn a Hausdorff
topological space (it is actually the quotient topology, that’s a map f : Pn → Y is
continuous if and only if f ◦ π is continuous).
For i = 1, . . . , n + 1, let Oi = {[x] ∈ Pn |xi 6= 0}. They are clearly open and cover
Pn . Define the maps φi : Oi → Rn given by
φi ([x1 : · · · : xn+1 ]) = (x1 /xi , · · · , xn+1 /xi ),
where the ith component is omitted. Clearly φi is a homeomorphism from Oi to
Rn . Moreover, the collection {(Oi , φi )} is a smooth atlas (exercise).
Note also that P1 is homeomorphic to the circle S 1 . It is clearly isomorphic to
S 1 / ∼, where ∼ identifies x and −x in S 1 . Then the map z → z 2 is a homeomor-
phism S 1 / ∼→ S 1 .
Definition 1.12. Let M be an abstract manifold. A subset N ⊂ M is called
an abstract submanifold of M if N is an abstract manifold on its own such that the
topology of N is induced from M and the inclusion map i : N → M is a smooth
immersion.

2. Tangent space
Let M be an m−dimensional manifold. A curve γ on M is a smooth map
γ : I → M , where I ⊂ R is open. This means (see Definition 1.8) that φ ◦ γ is
smooth for all chart φ on M . The expression φ ◦ γ is called coordinate expression
of γ with respect to φ. If p ∈ M is a point, a parametrized curve on M through p,
is a parametrized curve on M together with a point t0 ∈ I for which p = γ(t0 ).
Let γi : Ii → M , for i = 1, 2, be parametrized curves on M with p = γ1 (t1 ) =
γ2 (t2 ), and let φ be a chart around p. We say that γ1 and γ2 are tangential at p, if
the coordinate expressions satisfy
(φ ◦ γ1 )0 (t1 ) = (φ ◦ γ2 )0 (t2 ).
Lemma 2.1. Being tangential at p is an equivalence relation on curves through
p. It is independent of the chosen chart (φ, O).
Proof. The first part is easy.
If φ̃ is another chart then the coordinate expressions are related by
φ̃ ◦ γ = φ̃ ◦ φ−1 ◦ (φ ◦ γ)
on the overlap O ∩ Õ. The chain rule implies
(φ̃ ◦ γi )0 (ti ) = D(φ̃ ◦ φ−1 )(p)(φ ◦ γi )0 (ti ),
for each of the curves. This implies the result. 
12 2. DIFFERENTIAL MANIFOLDS

Denote the tangential equivalence relation at p by ∼p .

Definition 2.2. The tangent space Tp M is the set of ∼p −classes of parametrized
curves on M through p.
Theorem 2.3. Let (φ, U ) be a chart on M with φ(p) = q. For each element
v ∈ Rm let γv (t) = φ−1 (q + tv) for t close to 0. The map
Γ : v 7→ [γv ]
m
is a bijection of R onto Tp M . The inverse map is given by
[γ] 7→ (φ ◦ γ)0 (0),
for each curve γ on M with γ(0) = p.
The main difficulty of defining tangent vectors to a manifold is due to the
fact that an abstract manifold might not be naturally embedded into a fixed finite
dimensional linear space. Nevertheless, there is a universal embedding of each dif-
ferentiable manifold into an infinite dimensional linear space.

Let us denote by C ∞ (M ) the linear vector space of smooth functions on M ,

and by D(M ) the dual space of C ∞ (M ), that is the space of linear functions on
C ∞ (M ), and consider the embedding ι of M into D(M ) defined by the formula
[ι(p)](f ) = f (p); where p ∈ M ; f ∈ C ∞ (M ).
Having embedded the manifold M into D(M ), we can define tangent vectors to M
as elements of the linear space D(M ).
Definition 2.4. Let M be a differentiable manifold, p ∈ M . We say that a
linear function D ∈ D(M ) is a derivation of C ∞ (M ) at p if the equality
D(f g) = D(f )g(p) + f (p)D(g)
∞
holds for every f, g ∈ C (M ).
Derivations at a point p ∈ M form a linear subspace Derp (M ) of D(M ). Each
curve γ passes through p at 0 defines a derivation at the point p by the formula
Dγ (f ) = (f ◦ γ)0 (0).
Since two curves define the same derivation if and only if they are equivalent, there
is a one-to-one correspondence between the equivalence classes of curves and the
derivations obtained as Dγ for some γ. Hence the definition
Definition 2.5. A tangent vector to a manifold M at the point p ∈ M is a
derivation of the form Dγ , where γ is a smooth curve passes through p at 0.
The tangent space Tp M of M at the point p is the set of derivations Dγ along
curves in M passing through p at 0.
Definition 2.6. Let f : M → N be a smooth mapping between the differen-
tiable manifolds M and N , and let p ∈ M . The derivative of f at the point p is the
linear map Dp f : Tp M → Tf (p) N given in the following way: Let γ be a smooth
curve and let Dγ ∈ Tp M . Then Tp f (D) is the derivation D(f ◦γ) ∈ Tf (p) N .
In the first definition of the tangent space we have
Dp f ([γ]) = [f ◦ γ].
 3. VECTOR FIELDS AND FLOWS 13

Note that Dp f is well defined, that’s it doesn’t depend on the choice of γ. In-
deed, one sees that whenever γ1 and γ2 are tangential then so is f ◦ γ1 and f ◦ γ2 .

3. Vector fields and flows

A vector field on a smooth manifold M is simply a smooth map
X : M → T M,
such that for any p ∈ M , X(p) ∈ Tp M . In other words, X is a section for the
canonical projection T M → M . We denote by Γ(M ) the space of vector fields.
Definition 3.1. Let X be a vector field on M and I ⊂ R an interval. A
smooth map γ : I → M is called an integral curve for X if for any t ∈ I we have
(1) Dγ 0 (t) = X(γ(t)).
Associated to a chart φ = (x1 , · · · , xn ) on M near p, there is a basis of Tp M ,
formed by the tangent vectors (∂1 (p), · · · , ∂n (p)). The mapping
p 7→ ∂i (p)
gives a local smooth vector field in the domain of the chart for each i. Thus, every
smooth vector field X can be written in the form
Xn
X= ai ∂i ,
i
where ai are smooth functions near p.
Theorem 3.2. Let X be a smooth vector field on a differentiable manifold M .
Then for each point p ∈ M , there exists a unique maximal integral curve γp : I → M
of the vector field X such that 0 ∈ I.
The real vector space structure of the tangent space at any point makes it
possible to give Γ(M ) the structure of a real vector space where the addition is
defined point by point (X + Y )(p) = X(p) + Y (p) and the scaler multiplication
(λ · X)(p) = λX(p). In particular, we denote by 0 the zero section. Moreover, the
space Γ(M ) has also the structure of C ∞ (M )−module. The multiplication with a
function is also defined point by point (f · X)(p) = f (p)X(p).
Definition 3.3. A vector field X on M is called complete if, for each p ∈ M ,
there is an integral curve γ : R → M of X with γ(0) = p.
Lemma 3.4. Let M ⊂ Rn is a compact manifold. Then every vector field on
M is complete.
Proof. See Exercise ??. 
Since tangent vectors to a manifold at a point are identified with derivations at
this point, vector fields can be regarded as differential operators called derivations
of smooth functions assigning to a smooth function another smooth function by the
formula
[X(f )](p) = [X(p)](f ); where X ∈ Γ(M ); f ∈ C ∞ (M ); p ∈ M,
In this sense, a vector field X is seen as a linear mapping DX : C ∞ (M ) → C ∞ (M ),
satisfying the Leibniz rule
X(f g) = X(f )g + f X(g).
14 2. DIFFERENTIAL MANIFOLDS

Lie structure. Using this second point of view, we define a Lie brackets on
Γ(M ) by
[DX , DY ] := DX ◦ DY − DY ◦ DX .
None of the operators DX ◦ DY and DY ◦ DX are vector fields, since they are
second order operators, whereas vector fields are first order operators. However, it
turns out that their difference [DX , DY ] is again a vector field, this is given in the
following theorem
Theorem 3.5. The Lie bracket [DX , DY ] is a smooth vector field on M. That
is, there exists a vector field Z ∈ Γ(M ) such that [DX , DY ] = DZ .
Proof. It is sufficient to show that [DX , DY ] verifies the Leibniz rule. Let
f, g ∈ C ∞ (M ), then we have
[DX , DY ](f g) = (DX ◦ DY − DY ◦ DX )(f g) = DX (DY (f g)) − DY (DX (f g))
= DX (DY (f )g + f DY (g)) − DY (DX (f )g + f DX (g))
= DX (DY (f ))g + DY (f )DX (g) + DX (f )DY (g) + f DX (DY (g))
− DY (DX (f ))g − DX (f )DY (g) − DY (f )DX (g) − f DY (DX (g))
= [DX , DY ](f )g + f [DX , DY ](g).


Flows of a vector field. Let M as above and X ∈ Γ(M S ). Let p ∈ M , the

maximal existence interval of p is the open interval I(p) := I where the union
runs over all I ⊂ R open intervals containing 0 such that there exists an integral
curve γ : I → M for X with γ(0) = p.
By Theorem 3.2, there exists an integral curve γ : I(p) → M . The flow of X is the
map φ : D → M defined by
D := {(t, p)|p ∈ M, t ∈ I(p)},
and φ(t, p) := γ(t), where γ : I(p) → M is the unique integral curve.
Theorem 3.6. Let M ⊂ Rn be a smooth r−manifold and X ∈ Γ(M ) be a
smooth vector field on M . Let φ : D → M be the flow of X. Then the following
holds.
(i) Let p ∈ M and s ∈ I(p). Then
(2) I(φ(s, p)) = I(p) − s,
and, for every t ∈ R with s + t ∈ I(p), we have
(3) φ(s + t, p) = φ(t, φ(s, p)).
(ii) D is an open subset of R × M .
(iii) The map φ : D → M is smooth.
Proof. (i) The map γ : I(p) − s → M defined by γ(t) := φ(s + t, p) is a
solution of the initial value problem γ 0 (t) = X(γ(t)) with γ(0) = φ(s, p).
Hence I(p) − s ⊂ I(φ(s, p)) and equation (3) holds for every t ∈ R with
s + t ∈ I(p). In particular, with t = −s, we have p = φ(−s, φ(s, p)). Thus
we obtain equality in equation (2) by the same argument with the pair
(s, p) replaced by (−s, φ(s, p)).
 3. VECTOR FIELDS AND FLOWS 15

(ii) & (iii) Let (t0 , p0 ) ∈ D so that p0 ∈ M and t0 ∈ I(p0 ). Suppose t0 > 0. Then
K := {φ(t, p0 )|0 6 t 6 t0 } is a compact subset of M . (It is the image
of the compact interval [0, t0 ] under the unique solution γ : I(p0 ) → M
of equation (1).) Hence, by Lemma 3.7 bellow, there is an M −open set
U ⊂ M and an ε > 0 such that
K ⊂ U, (−ε, ε) × U ⊂ D,
and φ is smooth on (−ε, ε) × U . Choose N so large that t0 /N < ε. Define
U0 := U and, for k = 1, · · · , N , define the sets Uk ⊂ M inductively by
Uk := {p ∈ U |φ(t0 /N, p) ∈ Uk−1 }.
These sets are open in the relative topology of M. We prove by induction
on k that (−ε, kt0 /N +ε)×Uk ⊂ D and φ is smooth on (−ε, kt0 /N +ε)×Uk .
For k = 0 this holds by definition of ε and U . If k ∈ {1, · · · , N } and the
assertion holds for k − 1 then we have
p ∈ Uk ⇒ p ∈ U, φ(t0 /N, p) ∈ Uk−1
⇒ (−ε, ε) ⊂ I(p), (−ε, (k − 1)t0 /N + ε) ⊂ I(φ(t0 /N, p))
⇒ (−ε, kt0 /N + ε) ⊂ I(p).
Here the last implication follows from Equation (2). Moreover, for p ∈ Uk
and t0 /N − ε < t < kt0 /N + ε, we have, by Equation (3), that
φ(t, p) = φ(t − t0 /N, φ(t0 /N, p)).
Since φ(t0 /N, p) ∈ Uk−1 for p ∈ Uk the right hand side is a smooth map on
the open set (t0 /N − ε, kt0 /N + ε) × Uk . Since Uk ⊂ U , φ is also a smooth
map on (−ε, ε)×Uk and hence on (−ε, kt0 /N +ε)×Uk . This completes the
induction. With k = N we have found an open neighborhood of (t0 , p0 )
contained in D, namely the set (−ε, t0 + ε) × UN , on which φ is smooth.
The case t0 6 0 is treated similarly. This proves (ii) and (iii).


Lemma 3.7. Let M, X, D be as in Theorem 3.6 and let K ⊂ M be a compact

set. Then there exists an M −open set U ⊂ M and an ε > 0 such that K ⊂ U ,
(−ε, ε) × U ⊂ D, and φ is smooth on (−ε, ε) × U .
Proof. see [RS18, Lemma 2.4.10] 

Proposition 3.8. Let M ⊂ Rn be a smooth submanifold and X ∈ Γ(M ). Then

the following are equivalent.
(i) X is complete.
(ii) I(p) = R for all p ∈ M .
(iii) D = R × M.
Proof. Clear. 

Let us denote the space of diffeomorphisms of M by

Diff(M ) := {f : M → M |f is a diffeomorphism}.
This is a group. The group operation is composition and the neutral element is
the identity. Now equation (3) asserts that the flow of a complete vector field
16 2. DIFFERENTIAL MANIFOLDS

X ∈ Γ(M ) is a group homomorphism

R → Diff(M ) : t → φt .
This homomorphism is smooth and is characterized by the equation
d t
φ (p) = X(φt (p)), φ0 (p) = p,
dt
for all p ∈ M .
Theorem 3.9 (Normal form around a point). Let X ∈ Γ(M ) and p ∈ M
such that X(p) 6= 0. Then there exists a local chart (U, φ) around p such that if
φ = (x1 , · · · , xd ) then
∂
X|U = .
∂x1
In particular, φt (x1 , · · · , xd ) = (t + x1 , · · · , xd ).
Proof. Let (U, ψ) be any local coordinate chart around p with ψ(p) = 0. Since
we can compose this with a diffeomorphism of the target, we can assume that
∂
X(p) = (p).
∂x1
Define
χ(y1 , · · · , yd ) = φy1 (ψ −1 (0, y2 , · · · , yd )).
The differential d0 χ of this map at 0 is bijective. Indeed, one can compute
d0 χ = dp (φt )0 |0 ◦ d0 ψ −1 .

Using the equality

dp (φt )0 |0 = dφ0 (p)=p X ◦ dp φ0

and the fact that φ0 = id, we can see that d0 χ sends the coordinate ei to the basis
∂
. 
∂xi
4. Parallel transport
Let M ⊂ R be a submanifold. The Euclidean inner product h , i : Rn × Rn →
n

R induces an inner product

gp : Tp M × Tp M −→ R.
The set {gp }p is called the first fundamental form.

Note that this inner product induces an orthogonal projection Πp : Rn → Tp M

for each p ∈ M . The linear map Πp can be represented by n × n matrix over R,
and it is uniquely determined by the two conditions
Πp = Π2p ,
and for v ∈ Rn
Πp (v) = v ⇔ v ∈ Tp M.
Remark 4.1. The map Π : M → Rn×n is smooth.
The differential of Π is the linear map dp Π : Tp M → Rn×n which associate to
a vector v = γ 0 (0) ∈ Tp M the matrix
d
dp Π(v) := |t=0 Π(γ(t)) ∈ Rn×n .
dt
 5. ORIENTABILITY 17

Lemma 4.2. For any v, w ∈ Tp M we have

dp Π(v) · w = dp Π(w) · v ∈ Tp M ⊥ .
The collection of symmetric bilinear maps hp : (v, w) 7→ hp (v, w) := dp Π(v) · w
is called the second fundamental form.
Let M ⊂ Rn be a submanifold and γ : I → M a smooth curve. A vector
field along γ is a section of T M over I, that’s a smooth map Xγ : I → Rn such
that Xγ (t) ∈ Tγ(t) M . The derivative Xγ0 (t) is not in general in the tangent space
Tγ(t) M .
Definition 4.3. The covariant derivative of Xγ is the vector field ∇Xγ over
γ defined by
∇Xγ := Πp (Xγ0 (t)) ∈ Tγ(t) M.
Lemma 4.4. Let Xγ as above and λ : I → R be a smooth function. Then we
have
∇(λXγ ) = λ0 Xγ + λ∇(Xγ ).
Definition 4.5. Let I ⊂ R be an interval and let γ : I → M be a smooth
curve. A vector field Xγ along γ is called parallel if
∇Xγ (t) = 0.
for all t ∈ I.
In other words, Xγ is parallel if and only if Xγ0 (t) ⊥ Tγ(t) M.
Example 4.6. In particular, γ 0 is a vector field along γ and ∇γ 0 (t) = Πγ 0 (t) (γ 00 (t)).
Hence γ 0 is a parallel vector field along γ if and only if γ 00 (t) ⊥ Tγ(t) M for all t ∈ I.
Theorem 4.7. Let I ⊂ R be an interval and γ : I → M be a smooth curve.
Let t0 ∈ I and v0 ∈ Tγ(t0 ) M be given. Then there is a unique parallel vector field
Xγ along γ such that Xγ (t0 ) = v0 .
Definition 4.8 (Parallel transport). Let I ⊂ R be an interval and let γ : I →
M be a smooth curve. For t0 , t ∈ I we define the map
Φ(t, t0 ) : Tγ(t0 ) M −→ Tγ(t) M
by
Φ(t, t0 )(v0 ) := Xγ (t),
where Xγ is the unique parallel vector field along γ satisfying Xγ (t0 ) = v0 . The
collection of maps Φ(t, t0 ) for t, t0 ∈ I is called parallel transport along γ.

5. Orientability
Let V be a finite dimensional vector space. Two ordered bases (u1 , ..., un ) and
(v1 , ..., vn ) are said to be equally oriented if the transition matrix S, whose columns
are the coordinates of the vectors (u1 , ..., un ) with respect to the basis (v1 , ..., vn ),
has positive determinant. Being equally oriented is an equivalence relation among
bases, for which there are precisely two equivalence classes. The space V is said to
be oriented if a specific class has been chosen, this class is then called the orienta-
tion of V , and its member bases are called positive. The Euclidean spaces Rn are
usually oriented by the class containing the standard basis (e1 , ..., en ). For the null
space V = {0} we introduce the convention that an orientation is a choice between
18 2. DIFFERENTIAL MANIFOLDS

the signs + and −.

Let now a (φ, U ) be a chart on M with φ(p) = q, we obtain a basis for Tp M

by taking the pre-images (ei ) of each of the standard basis vectors e1 , . . . . , em for
Rm (in that order), that is, the basis vectors will be the equivalence classes of the
curves t → σ(x0 + tei ). This basis for Tp M is called the standard basis with respect
to σ. The compatibility condition between charts (Oi , φi ) and (Oj , φj ) on a set
M is that the map φj ◦ φ−1 j : φj (Oi ∩ Oj ) → φi (Oi ∩ Oj ) is a diffeomorphism. In
particular, the Jacobian matrix J(φj ◦ φ−1 j ) of the transition map is invertible, and
hence has nonzero determinant. If the determinant is > 0 everywhere, then we say
(Oi , φi ), (Oj , φj ) are oriented-compatible. An oriented atlas on M is an atlas such
that any two of its charts are oriented-compatible; a maximal oriented atlas is one
that contains every chart that is oriented-compatible with all charts in this atlas.
Definition 5.1. A manifold is called orientable if it admits an oriented atlas.
If an orientation has been chosen we say that M is an oriented manifold. The
notion of an orientation on a manifold is crucial, since integration of differential
forms over manifolds is only defined if the manifold is oriented.

Let (φ, Oi ) be a chart on an abstract manifold M , then the tangent space is

equipped with the standard basis with respect to φ. For each p ∈ Oi we say that
the orientation of Tp M , for which the standard basis is positive, is the orientation
induced by φ.
Proposition 5.2. Let M be a connected manifold. Then the following are
equivalent
(i) M is orientable.
(ii) Orientation is preserved moving along loops.
(iii) M admits a nowhere vanishing n−form.
Proof. 
n
Example 5.3. The spheres S are orientable. To see this, consider the atlas
with the two charts (U+ , φ+ ) and (U, φ− ), given by stereographic projections. (see
Example 1.7). Here
φ− : (U+ ∩ U) = φ+ (U+ ∩ U) = Rn r {0}.
The entries of the Jacobian matrix of φ− ◦ φ−1
+ at x are given by
∂ xi 1 2xi xj
(4) 2
= 2
δij − .
∂xj kxk kxk kxk4
Its determinant is −kxk−2n . (exercise).
If M is a smooth manifold of dimension n then for each p ∈ M the tangent space
Tp M is a vector space of dimension n, and hence has two choices of orientation.
This gives a two-sheeted covering space OM called the orientation covering of M .
So one sees that M is orientable if and only if its covering space is disconnected.
An advantage of this covering space point of view is that we immediately have
the following result.
Proposition 5.4. If M is a smooth connected manifold with π1 (M ) = 0 then
M is orientable.
 6. TENSORS 19

Proof. Each covering space M̃ → M is trivial since if p ∈ M then π1 (M̃ , p̃) ⊂

π1 (M, p) = 0. In particular the orientation covering must then consist of two simply-
connected components, each diffeomorphic to M. 
Recall That the fundamental group π1 (X; x0 ) of a pointed topological space
(X; x0 ) is the set of all homotopy classes of loops in X with base point x0 together
with the multiplication induced by concatenation as follows
[λ] · [η] = [λ ∗ η],
(
λ(2t) if t ∈ [0, 1/2]
where λ ∗ η[t] = .
η(2t − 1) if t ∈ [1/2, 1]
Example 5.5. The Möbius strip is a surface with only one side.

It is maybe the simplest example of a non-orientable manifold. To see this, one can
remark that the orientation is not preserved along loops, thus it is impossible to
make a consistent choice of orientations, because the band is one-sided. Choosing
an orientation in one point forces it by continuity to be given in neighboring points,
and eventually we are forced to the opposite choice in the initial point.
Example 5.6. The real projective space P2 is another example of a non ori-
entable surface. One way to visualise P2 is as the semisphere with points on the
equator identified with their antipodal points.
Example 5.7. In general, the real projective space Pn is orientable if and only
if n is odd or n = 0 .

6. Tensors
6.1. Exterior product. Let k > 0, and let V be a finite dimensional vector
space over R. The exterior product Λk V of V is defined by
k
O
Λk V = V / ∼,
where the equivalence relation is given by
⊗ki=1 xi ∼ ⊗ki=1 yi ⇔ ⊗ki=1 xi = sng(σ) ⊗ki=1 yσ(i) , for some permutation σ.
The class of a tensor x1 ⊗ · · · ⊗ xk is denoted x
1 ∧ · · · ∧ xk .
Note that the dimension of Λk V is given by nk , where n = dim(V ).
There is a canonical alternating k−linear map ∧k : V k → Λk V given by
∧k (x1 , . . . , xn ) = x1 ∧ · · · ∧ xn ,
such that for any alternating k−linear map f : V k → W , there is a unique linear
map g : Λn V → W such that f = g ◦ ∧k .
20 2. DIFFERENTIAL MANIFOLDS

Lemma 6.1. Assume that e1 , . . . , en is a basis of V . Then {ei1 ∧ · · · ∧ eik |i1 <
· · · < ik } is a basis of Λk V . In particular,
 
k n
dim Λ V = ,
k
where n = dim(V ).

Proposition 6.2. The vectors v1 , . . . , vk are linearly independent if and only

if v1 ∧ · · · ∧ vk 6= 0. Two linearly independent k−tuples of vectors v1 , . . . , vk and
w1 , . . . , wk span the same k−dimensional linear subspace if and only if v1 ∧· · ·∧vk =
cw1 ∧ · · · ∧ wk for some c ∈ Rn r {0}.
Corollary 6.3. The Grassmannian manifold Grk (V ) can be embedded into
the projective space P (Λk V ) by assigning to the k−dimensional subspace spanned by
the linearly independent vectors v1 , · · · , vk the 1−dimensional linear space spanned
by v1 ∧ · · · ∧ vk . This embedding is called the Plücker embedding.
Ln
The direct sum Λ∗ (V ) = k=1 Λk V is an associative algebra where the product
∧ is defined
(v1 ∧ · · · ∧ vk ) ∧ (w1 ∧ · · · ∧ wl ) = v1 ∧ · · · ∧ vk ∧ w1 ∧ · · · ∧ wl .
∗
Λ V is called the exterior algebra or Grassmannian algebra of V.

Definition 6.4. A (covariant) k−tensor on V is a multilinear map T : V k → R.

The set of k−tensors on V is denoted T k (V ).
Let φ1 , . . . , φk ∈ V ∗ . The map
φ1 ⊗ · · · ⊗ φk : (v1 , . . . , vk ) 7→ φ1 (v1 ) . . . φk (vk )
is a k−tensor on V . More generally, if S ∈ T k (V ) and T ∈ T l (V ) are tensors, we
define the tensor product S ⊗ T ∈ T k+l (V ) by
S ⊗ T (v1 , . . . , vk , vk+1 , . . . , vk+l ) = S(v1 , . . . , vk )T (vk+1 , . . . , vk + l).
Lemma 6.5. Let ξ1 , . . . , ξn be a basis of V ∗ . Then the family {ξi1 ⊗ · · · ⊗ ξik },
where i1 , . . . , ik are arbitrary number in {1, . . . , n}, form a basis of T k (V ). In
particular dim T k (V ) = nk .
So one sees that T k (V ) = ∼ V ∗ ⊗ · · · ⊗ V ∗.
∗
Definition 6.6. A contravariant
k−tensor on V is a covariant k−tensor on V .
k
More generally a tensor of type l is a k−covariant, l−contravariant multilinear
map
F : V ∗ × · · · × V ∗ × V × · · · × V → R.
| {z } | {z }
l times k times
The space of these tensors is denoted Tlk (V ).
Lemma 6.7.
(1) There is a natural identification T11 (V ) ∼
= End V .
k
(2) There is a natural identification between Tl+1 and the space of multilinear
maps
V ∗ × ··· × V ∗ ×V × ··· × V → V
| {z } | {z }
l times k times
 7. DIFFERENTIAL FORMS 21

Proof. Exercise. 

Definition 6.8. A k−tensor ϕ is called alternating if for every v1 , . . . , vk ∈ V

and any σ ∈ Σk , one have
ϕ(v1 , . . . , vk ) = sign(σ)ϕ(vσ(1) , . . . , vσ(k) ).
The space of these is denoted Ak (V ). For φ1 , . . . , φk ∈ V ∗ , the map
X
φ1 ∧ · · · ∧ φk : (v1 , . . . , vk ) 7→ φ1 (vσ(1) ) · · · φk (vσ(k) )
σ∈Σ

is alternating k−tensor.
So we get as before
Ak (V ) ∼
= Λk V ∗ .
Remark 6.9. There is a canonical map Λk V ∗ → (Λk V )∗ defined by sending
an alternating tensor ϕ1 ∧ · · · ∧ ϕk to the linear form
X
Alt(ϕ1 ∧ · · · ∧ ϕk )(x1 ∧ · · · ∧ xk ) = sgn(σ)ϕ1 (xσ(1) ) · · · ϕk (xσ(k) ),
σ∈Σk

which is an isomorphism.

7. Differential forms
Let M ⊂ R be a smooth submanifold. Let f ∈ C ∞ (M ). The differential of f
n

is a map df : T M → R, at each fiber, it gives a linear form dp f of the vector space

Tp M . Hence an element of the dual space.

Denote by T ∗ M the cotangent bundle over M ; it is the dual of the tangent

bundle. Each fiber of T ∗ M is isomorphic to (Tp M )∗ .
Definition 7.1. A differentiable 1−form on M (also called covector field ) is
a section of the cotangent bundle, that’s a smooth map ω : M → T ∗ M such that
ω(p) ∈ Tp∗ M for each p ∈ M . The space of 1−forms is denoted Ω1 (M ).
As we have seen before, each smooth function f gives rise to a 1−form df . A
differentiable 1−form is called exact if it is equal to df for some f ∈ C ∞ (M ). In
particular, if f equals one of the coordinates xi around a point p ∈ M , we get the
1−form dxi . Moreover, any 1−form ω can be written, locally around p in the form
n
X
ω= ai dxi .
i=1

Example 7.2. Let M = R2 with coordinates x, y, and let f ∈ C ∞ (R2 ). The

differential of x and y equal to dx and dy whose Jacobian matrices at any point are
equals to (1, 0) and (0, 1) respectively. Hence the differential of f is given by
∂f ∂f
df = dx + dy.
∂x ∂y
And any other covector field ω is of the form adx+bdy, with a, b ∈ C ∞ (R2 ), because
the cotangent bundle is trivial.
More generally, we have
22 2. DIFFERENTIAL MANIFOLDS

Definition 7.3. A differentiable k−form on M is a section of the exterior

product Λk T ∗ M of the cotangent bundle, that’s a smooth map ω : M → Λk T ∗ M
such that ω(p) ∈ Λk Tp∗ M for each p ∈ M . The space of k−forms is denoted Ωk (M ).
On a local coordinate chart U ⊂ M , let x1 , . . . , xd , then dxi1 ∧ · · · ∧ dxik is
k−form, where 1 6 i1 < · · · < ik 6 n. Moreover, any k−form on U can be written
uniquely in the form
X
w= aI dxi1 ∧ · · · ∧ dxik ,
I
where aI are smooth functions on U .
Example 7.4. Let M = R2 with coordinates x, y. A differential 2−form ω
on M is of the form adx ∧ dy, with a ∈ C ∞ (R2 ), because the cotangent bundle is
trivial and its second exterior power is a line bundle. If X, Y ∈ Tp M , then
ωp (X, Y ) = a(p)(x1 y2 − x2 y1 ).
Ln
We denote by Ω∗ (M ) = k=0 Ωk (M ). The wedge product ∧ gives Ω∗ (M ) the
structure of an algebra over C ∞ (M ).

7.1. Interior product. Let ω a differential k−form on a manifold M , and X

a smooth vector field. We have
Definition 7.5. The interior product of X and ω is the differential (k −
1)−form ιX ω defined by
ιX ω(X1 , . . . , Xk−1 ) := ω(X, X1 , . . . , Xk−1 ).
If k = 0, we set ιX ω = 0.
Interior product ιX with the vector field X can be thought of as a linear map
from Ω∗ (M ) into itself. This is a degree −1 map, as it decreases the degree of any
form by 1. In general, a linear map L : Ω∗ (M ) → Ω∗ (M ) is a degree d map, if
L(Ωk (M )) ⊂ Ωk+d (M ) for all k.

7.2. Exterior derivative. Let M ⊂ Rn be a submanifold. In a local chart

around a point p ∈ M , we can write a k−form ω ∈ Ωk in the form
X
ω= aI dxi1 ∧ · · · ∧ dxid ,
i1 <···<id

We get then
Definition 7.6. The exterior derivative of the k−form ω is the (k + 1)−form
dω giving locally by
X
dω = daI ∧ dxi1 ∧ · · · ∧ dxid .
i1 <···<id

It is a non trivial result that this definition can be globalized to get a morphism
d : Ωk (M ) → Ωk+1 (M ).
Definition 7.7. A k−form is called exact if it equals dα form some (k −
1)−form α. It is called closed if dω = 0.
Since d2 = 0 by definition, we deduce
 8. INTEGRATION OF DIFFERENTIAL FORMS 23

Lemma 7.8. An exact form is closed. That’s the exterior derivative defines a
complex
d d d
0 −→ Ω1 (M ) −→ Ω2 (M ) −→ · · · −→ Ωm (M ) → 0,
meaning Im(d) ⊂ Ker(d). Here m = dim(M ).
Example 7.9. Reconsider the example 7.2. The 1−form ω = adx + bdy is
exact if and only if
∂a ∂b
= .
∂y ∂x
∂f ∂f
Indeed, if ω is exact then there is an f ∈ C ∞ (R2 ) such that a = ,b = . Hence
∂x ∂y
∂a ∂2f ∂b
= = .
∂y ∂y∂x ∂x
∂a ∂b
Conversely, assume that = . Then choose a smooth two variable function f
∂y ∂x
∂f ∂a ∂2f ∂b ∂f
such that = a. Then = = . So b− doesn’t depend on x. Choose
∂x ∂y ∂y∂x ∂x ∂y
∂f
a function g on y such that g 0 (y) = b − . Then one sees that ω = d(f + g).
∂y
Note that this in not true in arbitrary manifold.
Let f : M → N a smooth map and let ω ∈ Ωk (N ). We define the pull back of
ω by f to be the k−form denoted f ∗ ω on M given by
f ∗ ω(p)(v1 , · · · , vk ) = ω(f (p))(dp f (v1 ), · · · , dp f (vk )),
for all v1 , · · · , vk ∈ Tp M and all p ∈ M .

8. Integration of differential forms

 CHAPTER 3

Vector bundles

S 6.3. [Sch08] Let Ep be a vector space of dimension r for any p ∈ M .

Theorem
Let E = p∈M Ep . Denote by π : E → M the canonical projection which sends
v ∈ Ep to p, and ι : M → E the inclusion p 7→ 0 ∈ Ep .
Definition 0.1. We say that E → M is a vector bundle of rank r if for each
p ∈ M there are charts (U, φ) around p and (Ũ , φ̃) around ι(p), with Ũ = π −1 (U ),
such that φ̃ restricts to vector space isomorphisms
Ep = π −1 (p) → φ(p) × Rr =∼ Rr
for all p ∈ M . One calls E the total space and M the base of the vector bundle.
The vector bundle charts may be pictured in terms of a diagram,

E ⊃ Ũ
φ̃
/ φ(U ) × Rr

π p
 
M ⊃U
φ
/ φ(U )

Example 0.2. Let M = Pn , and consider the line bundle T on Pn such that
its fiber Tp over p ∈ Pn is the line lp in Rn+1 defined p, i.e.
[
T = lp .
[l]∈Pn

It is a line bundle (a vector bundle of rank 1) over Pn . Indeed, let (Oi , φi ) be one of
the charts given in Example 1.11, and let pi : Rn+1 → R be the projection onto the
ith coordinate. Then for any [l] ∈ Oi , the restriction of pi to l gives an isomorphism
l → R. Let φ̃ : Õi := π −1 (Oi ) → Rn × R given by φ̃(p, t) = (φ(p), pi (t)) where
t ∈ lp .
S
Theorem 0.3. The union T M = p∈M Tp M → M is a vector bundle over M
of rank d = dim M .
Proof. Let (U, φ) be a chart on M . Recall that
Tp φ : Tp U = Tp M → Tφ(p) φ(U ) ∼
= Rd
is an isomorphism. Now, the pair (T U, T φ) is a chart on T M . Moreover, the
composition
Tp ψ ◦ (Tp φ)−1 = Tφ(p) (ψ ◦ φ−1 ),
which shows that this composition is smooth. 

25
 CHAPTER 4

Riemannian manifolds

The Riemannian manifold is the essential object studied in Riemannian geome-

try. It is a variety on which it is possible to perform length calculations. Technically,
a Riemannian manifold is a differential manifold endowed with an additional struc-
ture called a Riemannian metric allowing to calculate the scalar product of two
vectors tangent to the manifold at the same
point.
Let M be a smooth manifold. A kl −tensor field on M is section of the
tensor bundle Tlk M = qp∈M Tlk (Tp M ). A 01 −tensor field is a vector field and a


k


0 −tensor field which is alternating is k−form.
A Riemannian metric on an abstract manifold M is a 2−tensor field g ∈ T 2 (M )
that is symmetric (i.e., g(X, Y ) = g(Y, X)) and positive definite (i.e., g(X, X) > 0
if X 6= 0). A Riemannian metric thus determines an inner product on each tangent
space Tp M , which is typically written hX, Y i = g(X, Y ) for X, Y ∈ Tp M . A
manifold together with a given Riemannian metric is called a Riemannian manifold.
Lemma 0.1. Every smooth manifold can be given a Riemannian metric.
Proof. Consider a partition of unity {fα }α∈I . Let {(Ui , φi )} be a atlas on M .
Then define X
g= fα φ∗ gi ,
α
where gi is the canonical metric on φi (Ui ). 
Given a Riemannian metric g on M , we definepthe length or norm of a tangent
vector at p, as in the Euclidean case, by |X|p = gp (X, X). We define the angle
between two nonzero vectors X, Y ∈ Tp M to be the unique θ ∈ [0, π] satisfying
cos θ = gp (X, Y )/(|X|p |Y |p ).
Given a smooth curve γ : [a, b] → M . For any t ∈ [a, b], we have a tangent
vector γ 0 (t) ∈ Tγ(t) M . So get a nonnegative continuous function on [a, b] given by
t 7→ |γ 0 (t)|γ(t) (one should assume that g is continuous). So we define the length of
the curve γ to be the integral of this function
Z b
L (γ) = |γ 0 (t)|γ(t) dt.
a
Assume now that M is connected. Note that since M is locally homeomorphic
to an open of Rd , then it is locally path connected, and so it is path connected since
it is connected.
Define
d(p, q) = inf L (γ)|γ is a pieacewise C 1 curve connecting p and q .


Proposition 0.2. d is a metric on M .

27
 CHAPTER 5

Exercises

Exercise 0.1.
Prove that the collection A = {(φ± , U± )} given in Example (1.7) is an atlas on
Sn.
Exercise 0.2.
Assume that W ⊂ Rm and V ⊂ Rn are open non-empty sets, and let f : Rm →
Rn be a continuous map. Denote by M the graph of f , i.e.
M = {(x, y)| y = f (x)}.
Let U = (W × V ) ∩ M and let φ(x, y) = x be the projection of U onto W . Then
the pair (φ, U ) is a chart on M .
Exercise 0.3. Consider the cylinder C = {(x, y, z) ∈ R |x2 + y 2 = 1} ⊂ R3
with the induced topology, and let ϕ1 : R2 → C and ϕ2 : R2 → C given by
1 − u2 2u
ϕ1 (u, v) = ( , , v).
1 + u 1 + u2
2

u2 − 1 −2u
ϕ2 (u, v) = ( , , v).
1 + u2 1 + u2
Denote by Vi the image of ϕi (for i = 1, 2).
(1) Show that ϕ1 is injective.
(2) Is ϕ1 surjective? Justify your answer.
(3) Describe precisely V1 the image of ϕ1 , and deduce that it is an open in C.
(4) Deduce that ϕ1 : R2 → V1 is bijective.
(5) Show that the map ϕ−1 2
1 : V1 → R given by
y
(x, y, z) 7→ ( , z)
1+x
is the inverse map of ϕ1 .
Use the fact that x2 + y 2 = 1
 
2
 y 1 − x
which induces =
(1 + x)2 1+x
−1
(6) Show that ϕ1 : R2 → V1 is a homeomorphism and deduce that (ϕ1 , V1 )
is a chart on C.
(7) Prove that the collection A = {(ϕ−1 −1
1 , V1 ), (ϕ2 , V2 )} is an atlas on C (we
−1
admit that (ϕ2 , V2 ) is a chart).
(8) Deduce that C is an abstract manifold. What is its dimension?

Exercise 0.4. Let T n = S 1 ×· · ·×S 1 be the real n-torus. Let O = {(x1 , · · · , xn ) ∈

T n |∀i : xi 6= (1, 0)}.
(1) Show that O is open in T n with respect to the induced topology of R2n .
29
30 5. EXERCISES

(2) Show that the map (x1 , · · · , xn ) 7→ (φ(x1 ), · · · , φ(xn )) is a bijection : O →

b
Rn , where φ(a, b) = 1−a
(3) Show that (φ, O) is a chart on T n .
Exercise 0.5. Consider the projective space Pn given in Example (1.11). Show
that the collection of pairs (φi , Oi ), for i = 1, . . . , n + 1, is an atlas on Pn .
Exercise 0.6. Let P1 be the real projective line, and S 1 the unit circle in R2 .
Consider the relation ? on S 1 given by
v ? w ⇔ v = w or v = −w.
(1) Determine the equivalence class of v ∈ S 1 .
(2) For any v = (x, y) ∈ R2 r 0, show that there are exactly two real numbers
λ1 and λ2 such that λ1 v and λ2 v are in S 1 (λ1 and λ2 must be deter-
mined).
(3) Deduce a map ϕ : P1 → S 1 /? and show that it is bijective.
Consider now the circle S 1 as a subset of C via the natural identification R2 ∼
= C.
Let φ : S 1 /? → S 1 be the map given by φ([z]) = z 2 , where [z] is the equivalence
class of z modulo ?.
(4) Show that φ is well defined and bijective.
(5) Deduce that P1 is bijective to S 1 .
p
Exercise 0.7. Consider the subset T = {(x, y, z) ∈ R3 |( x2 + y 2 −2)2 +z 2 =
1} ⊂ R3 with the induced topology, and let f : S 1 × S 1 → R3 the map given by
given by
f ((a, b), (c, d)) = ((c + 2)a, (c + 2)b, d)
(1) Show that f is injective.
(2) Show that Im(f ) = T and deduce that f : S 1 × S 1 → T is smooth
diffeomorphism.
(3) Define an atlas on S 1 .
(4) Deduce an atlas on T .
(5) What is the dimension of T .
Exercise 0.8. If f ∈ C ∞ (M ) is a constant function and D is a derivation at
a point p ∈ M , then D(f ) = 0.
Exercise 0.9. Let M = S 2 be the unit sphere. Determine the tangent space.
Give an example of a vector field X and an integral curve for this X.
Exercise 0.10. Let M = S 2 be the unit sphere. Find the directional derivative
at p = (0, 1, 0) in the following cases : X = (2, 1, 0), Y = (1, −1, 2) and Z = (0, 1, 1).
Exercise 0.11 (Bump function). Prove that there exists a smooth function
ϕ : Rn → R with value in [0, 1] such that ϕ(x) = 1 if kxk 6 r and ϕ(x) = 0 if
kxk > s, where 0 < r < s.
Exercise 0.12. If two functions f, g ∈ C ∞ (M ) coincide on a neighborhood U
of p ∈ M and D is a derivation at p then D(f ) = D(g).
Exercise 0.13. Show that the universal embedding of an abstract manifold
M in D(M ) is injective.
Exercise 0.14. Show that for any two vector fields X an Y , we have
ιX ◦ ιY = ιY ◦ ιX .
 1. SOLUTIONS 31

Exercise 0.15. Let U ⊂ R3 be a non empty open subset, and consider the
∂
vector field X = x2 y ∂x and the 2−form ω = zdx ∧ dy + xdy ∧ dz.
(1) Compute ω̃ := dω.
(2) Compute ιX ω̃.
(3) Is ω̃ exact ? Is it closed? Justify your answer.
Exercise 0.16. Show that S n is orientable.
Exercise 0.17. Let M = Pn , and consider the line bundle T on Pn such that
its fiber Tp over p ∈ Pn is the line lp in Rn+1 defined p, i.e.
[
T = lp .
[l]∈Pn

Show that T is a line bundle (a vector bundle of rank 1) over Pn .

1. Solutions
Solution 1.1. Application exercise.
Solution 1.2. It is enough to show that φ is a homeomorphism (its inverse is
given by φ−1 (x) = (x, f (x))).

Solution 1.3.
1 − u21
(1) Let (u1 , v1 ), (u2 , v2 ) ∈ R2 s.t. ϕ1 (u1 , v1 ) = ϕ1 (u2 , v2 ), then =
1 + u21
1 − u22 2u1 2u2
, = and v1 = v2 . The first one implies u21 = u22 ,
1 + u22 1 + u21 1 + u22
replacing this in the second one we get u1 = u2 .
(2) No, because (−1, 0, v) doesn’t belong to its image, for any v.
(3) V1 = {(x, y, z)|x 6= −1, x2 + y 2 = 1}. So geometrically, it is the cylinder
deprived of the line with equations x = −1, y = 0. Since this line is
closed, V1 is open.
(4) By equation (1), ϕ1 is injective, and by definition of V1 , ϕ1 is surjective
on V1 . So it is bijective.
(5) We have for all (u, v) ∈ R2 and all (x, y, z) ∈ V1 :
1 − u2 2u
ϕ−1 −1
1 ◦ ϕ1 (u, v) = ϕ1 ( , , v)
1 + u2 1 + u2
2u
1+u2
=( 2 , v) = (u, v).
1 + 1−u
1+u2

and
y
ϕ1 ◦ ϕ−1
1 (x, y, z) = ϕ1 ( , z)
1+x
y y
1 − ( 1+x )2 2( 1+x )
=( y , y , z)
1 + ( 1+x ) 1 + ( 1+x )2
2

1− y
1 − ( 1+x ) 2( 1+x )
=( , , z)
1 + ( 1−x 1−x
1+x ) 1 + ( 1+x )
= (x, y, z).
32 5. EXERCISES

(6) By the previous questions, ϕ1 is bijective. Moreover, from the expressions

of ϕ1 and ϕ−1
1 , we see that both are continuous, since they are given by
well defined rational polynomials. Since ϕ−1 2
1 : V1 → R and both V1 and
2 −1
R are open, we deduce that (ϕ1 , V1 ) is a chart on C.
(7) One can see that V2 is the cylinder C without the line whose equations are
x = 1 and y = 0. So clearly V1 ∩ V2 = C. Now the composition ϕ−1 1 ◦ ϕ2
is given on ϕ−1
2 (V 1 ∩ V2 ) = {(u, v)|u 6
= 0} ⊂ R 2
by
−2u
1+u2 −1
(u, v) 7→ ( 2 −1 , v) = ( , v),
1 + u1+u 2
u

which is clearly smooth.

(8) We have an atlas A on C, so we have a maximal atlas A, and C is clearly
Hausdorff, so (C, A) is an abstract manifold. Its is dimension is 2.
Solution 1.4. (1) Let N = (1, 0). Then O = (R2n r {(N, · · · , N )}) ∩ T n
So it is open.
(2) It is sufficient to show that φ : S 2 r N → R is bijective.
(3) From the formula of φ and φ−1 , we see that they are continuous.
Solution 1.5. We have
φ−1
i : Rn → Oi , (x1 , · · · , xn ) 7→ [x1 : · · · : 1 : · · · : xn ],
where 1 is in ith position. Hence φj ◦ φ−1
i : Rn r {xj 6= 0} → Rn
φj ◦ φ−1
i (x1 , · · · , xn ) = (x1 /xj , . . . , 1/xj , · · · , xn /xj ),

which is smooth.
Solution 1.6.
(1) v̄ = {v, −v}
1
(2) λ1 = = −λ2 , clearly k λ1 v k=k λ2 v k= 1.
kvk
(3) given [x : y] ∈ P1 , define ϕ([x : y]) = λ1 v, where v = (x, y). Clearly
this is well defined. In fact, geometrically, it sends a line l through the
origin in R2 to the class given by the two points l ∩ S 1 , this is a bijective
correspondence.
(4) Since z̄ = {z, −z}, we see that φ([z]) = z 2 = (−z)2 = φ([−z]), so φ
is well defined. Now, since C is algebraically closed, φ is surjective. If
φ(w) = φ(z), then w2 = z 2 , so w = ±z, hence [w] = [z], hence φ is
injective, so it is bijective.
(5) The map φ ◦ ϕ : P1 → S 1 is the required bijection.
Solution 1.7.
(1) Easy computation.
(2) Easy.
y
(3) A = {(U+ , φ+ ), (U− , φ− )}, where U± = S 1 r {(±1, 0)}, φ± (x, y) =
.
1±x
(4) An atlas on T is given by {(U± × U± , φ± × φ± )} which contains four
charts.
(5) For example φ+ × φ+ : U+ × U+ → R2 , so dim T = 2.
 1. SOLUTIONS 33

Solution 1.8. Because of linearity, it is enough to show that D(1) = 0, where

1 is the constant 1 function on M . This comes from
D(1) = D(1 · 1) = D(1)1(p) + 1(p)D(1) = 2D(1).
Solution 1.9. The tangent space at p = (a, b, c) ∈ S 2 is the plan of R3 whose
normal vector is (a, b, c). Thus Tp S 2 has equation ax + by + cz = 0. Let X be
the map S 2 → R3 given by X(p) = (−b, a, 0). Clearly X is a vector field and an
integral curve for X is given by [0, 2π] 3 t 7→ (cos t, sin t, 0) ∈ S 2 .
Solution 1.10. The directional derivative DX is defined by DX : C ∞ (S 2 ) → R
as
∂f ∂f ∂f ∂f ∂f
DX (f ) = dp f (X) = ( (p), (p), (p))X = 2 (p) + (p).
∂x ∂y ∂z ∂x ∂y
Solution 1.11. The function ψ : R → R which equals ψ(x) = e−1/x if x > 0
and 0 otherwise, is a smooth function. Define
ψ(s − kxk)
ϕ(x) =
ψ(s − kxk) + ψ(kxk − r)
Solution 1.12. We can define a smooth function h on M which is zero outside
U and such that h(p) = 1. In this case h(f − g) is the constant 0 function on M .
Thus we have
0 = D(0) = D(h(f − g)) = D(h)(f (p) − g(p)) + h(p)D(f − g) = D(f ) − D(g).
Solution 1.13. Use the bump function to show that the universal embedding
of an abstract manifold M in D(M ) is injective.
Solution 1.14. We have ω(X, Y, · · · ) = −ω(Y, X, · · · ). this implies the result.
Solution 1.15.
(1) ω̃ = dz ∧ dx ∧ dy + dx ∧ dy ∧ dz = 2dx ∧ dy ∧ dz.
(2) ιX ω̃(Y, Z) = ω̃(X, Y, Z) = 2dx(X)dy(Y )dz(Z) − 2dx(X)dy(Z)dz(Y ) =
2x2 y(dy(Y )dz(Z) − dy(Z)dz(Y )), So ιX ω̃ = 2x2 ydy ∧ dz.
(3) By definition, ω̃ = dω, so it is exact. It is closed, since it is three form on
three dimensional space.
Solution 1.16. Consider the atlas with the two charts (U+ , φ+ ) and (U− , φ− ),
given by stereographic projections. (see Example 1.7). Here
φ− : (U+ ∩ U− ) = φ+ (U+ ∩ U− ) = Rn r {0}.
The entries of the Jacobian matrix of φ− ◦ φ−1
+ at x are given by
∂ xi 1 2xi xj
(5) 2
= 2
δij − .
∂xj kxk kxk kxk4
Its determinant is −kxk−2n . Indeed, one can see that this Jacobian matrix can be
written in the form  
1 1 t
In − x x .
kxk2 kxk2
Since xxt = kxk2 , one sees by multiplying by x from the right that x is a simple
eigenvector for this matrix with eigenvalue −1/kxk2 , and if y1 , · · · , yn−1 are a basis
for the orthogonal subspace {x}⊥ of x, then by multiplying the Jacobian by yi , we
see that they are eigenvector with eigenvalue 1/kxk2 . Since the determinant is the
34 5. EXERCISES

multiplication of the eigenvalues, the result follows.

Now the atlas {(U+ , φ+ ), (U− , φ̃− )}, where
−x1 x2 xn
φ̃− (x0 , · · · , xn ) = ( , ,··· , ),
1 ± x0 1 ± x0 1 ± x0
has a determinant kxk−2n > 0.
Solution 1.17.
Let (Oi , φi ) be one of the charts given in Example 1.11, and let pi : Rn+1 → R
be the projection onto the ith coordinate. Then for any [l] ∈ Oi , the restriction of
pi to l gives an isomorphism l → R. Now take φ̃ : Õi := π −1 (Oi ) → Rn × R given
by φ̃(p, t) = (φ(p), pi (t)) where t ∈ lp . This gives a local trivialization.
 Bibliography

[Csi14] Balazs Csikos. Differential geometry. Lecture notes, 2014.

[Gua18] Marco Gualtieri. Differential geometry. Lecture notes, 2018.
[Oan18] Alexandru Oancea. Notes de cours de Géométrie différentielle. Lecture notes, 2018.
[Pet] Peter Petersen. Manifold Theory. Lecture notes.
[RS18] Joel W. Robbin and Dietmar A. Salamon. Introduction to differential geometry. Lecture
notes, 2018.
[Sch08] Henrik Schlichtkrull. Lecture Notes for Geometry 2. Lecture notes, 2008.
[Spi99] Michael Spivak. A comprehensive introduction to differential geometry, volume 1. Publish
or Perish, Inc, 1999.

35