Differential Geometry*

A BHILASH S AHA
Spring 2025

Contents
1 Smooth Manifolds 2
1.1 Locally Euclidean Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Smooth Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Smooth Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Projective Spaces and Grassman Manifolds . . . . . . . . . . . . . . . . . 11

2 Tangent Vectors 12
2.1 Geometric Tangent Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Derivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Bump Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Pushforward of Smooth Map . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Submanifolds 17
3.1 The Rank Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Immersions, Submersions and Embeddings . . . . . . . . . . . . . . . . . 18
3.3 Regular Submanifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4 The Level Set Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4.1 Tangent Space of a Regular Level Set . . . . . . . . . . . . . . . . . 25
3.5 Matrix Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

4 The Tangent Bundle 25

4.1 Topology on the Tangent Bundle . . . . . . . . . . . . . . . . . . . . . . . 26
4.2 Vector Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

5 Vector Fields 29
5.1 Smoothness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 The Lie Bracket . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.3 Relatedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

* based on a course by Dr. Sayan Bagchi, IISER Kolkata

1
1 S MOOTH M ANIFOLDS 2

5.4 Lie Groups and Lie Algebras . . . . . . . . . . . . . . . . . . . . . . . . . 32

5.5 Pushforward of Left-Invariant Vector Fields . . . . . . . . . . . . . . . . . 35

6 Partitions of Unity 36
6.1 Definition and Existence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2 Whitney Embedding Theorem . . . . . . . . . . . . . . . . . . . . . . . . 36

§1 Smooth Manifolds
§1.1 Locally Euclidean Spaces

Definition 1.1
A topological space E is said to be locally Euclidean of dimension n if for every
p ∈ E, there exist an open subset U ⊂ E with U containing p, an open subset
V ⊂ Rn and a homeomorphism ϕ : U → V.

The pair (U, ϕ) is called a chart about p. Any open subset U ⊂ M , and a homeomorphism
ϕ : U → ϕ(U ) ⊂ Rn is called a chart.

Definition 1.2
A topological space M is said to be topological manifold of dimension n if it is
Hausdorff, second-countable, and locally Euclidean of dimension n.

Manifold Dimension

Any discrete topological space 0

Any open subset of Rn n
Mm×n (R) mn
GLn (R) n2

Table 1: Examples of Topological Manifolds

Example 1.3 (Graph of a continuous function)

Let F : U → Rk be continous, where U ⊂ Rn is an open set. Then the graph

Γ(F ) = {(x, F (x)) : x ∈ U } ⊂ Rn × Rk

is a manifold of dimension n.
1 S MOOTH M ANIFOLDS 3

Example 1.4 (n-sphere)

Let S n = {x ∈ Rn+1 : ni=1 x2i = 1}. Define
P

Ui+ = {x ∈ S n : xi > 0}
Ui− = {x ∈ S n : xi < 0}.

Then,
n
[
Sn = Ui+ ∪ Ui− .
i=1

We define the maps ϕ+

i : Ui+ → Bn (0, 1) and ϕ− −
i : Ui → Bn (0, 1) as
q
ϕ+
i (x1 , x2 , . . . , xi−1 , 1 − x21 − . . . x2n , xn , xi+1 , . . . , xn ) = (x1 , . . . , xn )
q
ϕ− 2 2
i (x1 , x2 , . . . , xi−1 , − 1 − x1 − . . . xn , xn , xi+1 , . . . , xn ) = (x1 , . . . , xn ).

Theorem 1.5 (Product manifold)

Let M, . . . , Mk be topological manifolds of dimension n1 , . . . , nk respectively. Let

M = M × . . . × Mk .

Then, M is a topological manifold of dimension n1 + . . . + nk .

Proof. Let p = (p1 , . . . , pn ) ∈ M . There exist charts (Ui , ϕi ) about pi for 1 ≤ i ≤ k.

Let U = U1 × . . . × Uk , and note that it is an open subset of M containing p. Define
ϕ : U → ϕ(U1 ) × . . . × ϕ(Uk ) as

ϕ(p1 , . . . , pn ) = (ϕ(p1 ), . . . , ϕ(pn )).

It is a homeomorphism since each ϕi is, and the Hausdorffness and second countability
are inherited by the product topology.

§1.2 Smooth Structures

Let M be a topological manifold, and let V be an open subset of M . The goal of this
section is to set up the stage to define what it means for a function f : V → R to be
smooth. If M is covered by just one chart (M, ϕ), then we may declare f : M → R to
be smooth if the function f ◦ ϕ−1 : ϕ(V ) → R is smooth. Unfortunately, this strategy
doesn’t work, since it depends on the choice of charts.
1 S MOOTH M ANIFOLDS 4

Example 1.6
Let M = R be the one-dimensional manifold covered by a two different global
charts.

1. ϕ1 (x) = x for each x ∈ R, and

2. ϕ2 (x) = x3 for each x ∈ R.

According to the above naive definition, the function f : R → R, f (x) = x is smooth

at x = 0 with respect to ϕ1 but not with respect to ϕ2 .

To make it work, we need to ensure that if two different charts (U, ϕ) and (V, ψ) are
about p, then ϕ ◦ ψ −1 is smooth at the point ψ(p). To do this, we introduce the notion of
compatibililty of charts.

Definition 1.7
Two charts (U, ϕ) and (V, ψ) of a topological manifold are said to be compatible if
either one of the following conditions hold true.

1. U ∩ V = ∅, or

2. the functions

ϕ ◦ ψ −1 : ψ(U ∩ V ) → ϕ(U ∩ V ) and

ψ ◦ ϕ−1 : ϕ(U ∩ V ) → ψ(U ∩ V )

are smooth.

Definition 1.8
A smooth atlas on a locally Euclidean space M is a collection U = {(Uα , ϕα ) : α ∈ I}
of pairwise compatible charts such that the collection {Uα : α ∈ I} is an open cover
of M .

If we only consider charts that belong to a smooth atlas, then of course checking smooth-
ness for any one chart suffices, since

f ◦ ψ −1 = f ◦ ϕ−1 ◦ ϕ ◦ ψ −1 .

Next we would like to determine if compatibililty of charts is an equivalence relation.

The answer turns out to be no: indeed, a slightly stronger hypothesis is required.
1 S MOOTH M ANIFOLDS 5

Definition 1.9
We say that a chart (U, ϕ) is compatible with an atlas U = {(Uα , ϕα ) : α ∈ I} if it is
compatble with all charts in U.

Lemma 1.10
Let U = {(Uα , ϕα ) : α ∈ I} be an atlas on a locally Euclidean space. If (V, ψ) and (W, σ)
are both compatible with U, then (V, ψ) and (W, σ) are compatibile with each other.

Proof. If V ∩ W = ∅, we are done. Otherwise, let p ∈ V ∩ W . Since U covers M , there

exists α ∈ I such that p ∈ Uα . Since (Uα , ϕα ) is compatibile with both (V, ψ) and (W, σ),
the maps

ϕα ◦ ψ −1 : ψ(Uα ∩ V ) → ϕα (Uα ∩ V ),
ψ ◦ ϕ−1
α : ϕα (Uα ∩ V ) → ψ(U ∩ V ),

ϕα ◦ σ −1 : σ(Uα ∩ W ) → ϕα (Uα ∩ W ),
σ ◦ ϕ−1
α : ϕα (Uα ∩ W ) → ψ(U ∩ W )

are all smooth. Now, let N be a neighbourhood of p that is contained in U ∩ V ∩ W . On

σ(N ), the map ψ ◦ σ −1 is expressed as

ψ ◦ σ −1 = ψ ◦ ϕ−1 −1
α ◦ ϕα ◦ σ ,

which is smooth. Similarly, on ψ(N ), the map σ ◦ ψ −1 is expressed as

σ ◦ ψ −1 = σ ◦ ϕ−1 −1
α ◦ ϕα ◦ ψ ,

which is smooth as well. Thus, any point σ(p) ∈ σ(U ∩ V ) has a neighbourhood
σ(N ) containing it on which ψ ◦ σ −1 is smooth, and any point ψ(p) ∈ ψ(U ∩ V ) has a
neighbourhood ψ(N ) containing it on which the map ψ ◦ σ −1 is smooth. Thus, the maps

ψ ◦ σ −1 : σ(V ∩ W ) → ψ(V ∩ W ) and

σ ◦ ψ −1 : ψ(U ∩ V ) → σ(U ∩ V )

are smooth, which proves the lemma.

Definition 1.11
An atlas M of a locally Euclidean space M is said to be maximal if it is not properly
contained in any other atlas of M .
1 S MOOTH M ANIFOLDS 6

Definition 1.12
A C ∞ atlas of dimension n is a pair (M, M), where M is a topological manifold of
dimension n and M is a maximal atlas of M .

The data of M is called the differentiable structure on M .

Theorem 1.13
Any atlas U = {(Uα , ϕα ) : α ∈ I} of a locally Euclidean space M is contained in a unique
maximal atlas.

Proof. Let
[
M= {(V, ψ) : (V, ψ) is a chart about p that is compatible with U}.
p∈M

Clearly, the open sets in M generously cover M , since U ⊂ M. By Lemma 1.10, the charts
added to M are compatible with U so compatble with each other. Any chart compatible
with the M must be compatible with the original atlas U and so by construction belongs
to the new atlas. This proves that the new atlas is maximal.

Let M′ be a maximal atlas containing U . Since M′ is an atlas, all of the charts in M′ are
compatibile with U, hence contained in M. Conversely, any chart in M must lie in M′ by
maximality of M′ . Hence, M = M′ .

In summary, to show that a topological space M is a smooth manifold, it suffices to

check that
(i) M is Hausdorff and second countable, and

(ii) M has a C ∞ atlas (not necessarily maximal).

Lemma 1.14
Let U = {(Uα , ϕα ) : α ∈ I} be the maximal atlas on a manifold M . For any open set U in
M and a point p ∈ U , there α ∈ I such that p ∈ Uα ⊂ U .

Proof. Since U is an atlas, there exists some α′ ∈ I such that p ∈ Uα′ . The set U ∩ Uα′ is
open, and let ϕ be homeomorphism ϕα′ |U ∩Uα′ . Then, (U ∩ Uα′ , ϕα ) is a chart about p that
is compatibile with U. By maximality, U must contain (U ∩ Uα′ , ϕ).

Example 1.15
The graph Γ(f ) of a C ∞ function f : U → Rm has an atlas with a single chart
(Γ(f ), f ), and is therefore a C ∞ manifold.
1 S MOOTH M ANIFOLDS 7

§1.3 Smooth Functions

Definition 1.16
Let M be a smooth manifold of dimension n. A function f : M → R is said to be
C ∞ or smooth at a point p in M if there is a chart (U, ϕ) about p in M such that f ◦ ϕ−1 ,
a function defined on the open subset ϕ(U ) of Rn , is C ∞ at ϕ(p). The function f is
said to be C ∞ on M if it is C ∞ at every point of M .

Remark 1.17. The definition of the smoothness of a function f at a point is inde-

pendent of the chart (U, ϕ), for if f ◦ ϕ−1 is C ∞ at ϕ(p) and (V, ψ) is any other chart
about p in M , then on ψ(U ∩ V ),

f ◦ ψ −1 = (f ◦ ϕ−1 ) ◦ (ϕ ◦ ψ −1 ),

which is C ∞ at ψ(p).

Lemma 1.18
Let U be an open subset of a smooth manifold M . If f : U → R is smooth at p ∈ U then f
is continuous at p.

Proof. The map f ◦ ϕ−1 : ϕ(U ) → R is a smooth, hence continuous at ϕ(p). Now, express
f = f ◦ ϕ ◦ ϕ−1 , which is continuous as a composition of continuous maps.

Here is an alternate characterization of a smooth map, without resorting to points.

Theorem 1.19 (Smoothness in terms of atlas)

Let M be a manifold of dimension n, and f : M → R a real-valued function on M . The
following are equivalent:

1. The function f : M → R is C ∞ .

2. The manifold M has an atlas such that for every chart (U, ϕ) in the atlas,

f ◦ ϕ−1 : ϕ(U ) → R is C ∞ .

3. For every chart (V, ψ) on M , the function

f ◦ ψ −1 : ψ(V ) → R is C ∞ .

Proof. 3 =⇒ 2 is obvious.
1 S MOOTH M ANIFOLDS 8

1 =⇒ 3: Let (V, ψ) be a chart, and let p ∈ V . By Remark 1.17, the function f ◦ ψ −1

is smooth at p. Since p was arbitrary, f ◦ ψ −1 is smooth on all of V .

2 =⇒ 1: Follows from the definition.

Definition 1.20
Let F : N → M be a map and h : M → R. The pullback of h by F , denoted by F ∗ h,
is the composite function h ◦ F .

In this terminology, a function f on M is C ∞ on a chart (U, ϕ) if and only if its pullback

(ϕ−1 )∗ f by ϕ−1 is C ∞ on the subset ϕ(U ) of Rn .

Theorem 1.21
Let M , N , P be smooth manifolds, and let F : M → N , and G : N → P be smooth maps.
Let f, g ∈ C ∞ (N ) and a, b ∈ R. Then:

1. F ∗ (af + bg) = aF ∗ (f ) + bF ∗ (g).

2. F ∗ (f g) = (F ∗ f )(F ∗ g).

3. (G ◦ F )∗ = F ∗ ◦ G∗ .

In particular, F ∗ : C ∞ (N ) → C ∞ (M ) is a homomorphism of R-algebras.

Definition 1.22
Let M and N be manifolds of dimension m and n, respectively. A continuous map
F : M → N is C ∞ at a point p in N if there are charts (V, ψ) about F (p) in N and
(U, ϕ) about p in M such that the composition ψ ◦ F ◦ f −1 , a map from the open
subset f (F −1 (V ) ∩ U ) of Rm to Rn , is C ∞ at ϕ(p). A continuous map F : M → N is
said to be smooth on M if it is smooth at every p ∈ M .

We assume F : M → N continuous to ensure that F −1 (V ) is an open set in M .

Lemma 1.23 (Independence of the choice of charts)

Suppose F : M → N is C ∞ at p ∈ M . If (U, ϕ) is any chart about p in M and (V, ψ) is
any chart about F (p) in N , then ψ ◦ F ◦ ϕ−1 is C ∞ at ϕ(p).

Proof. Since F is C ∞ at p ∈ M , there are charts (Uα , ϕα ) about p in M and (Vβ , ψβ ) about
F (p) in N such that ψβ ◦ F ◦ ϕ−1
α is C
∞ at ϕ (p). By the C ∞ compatibility of charts in a
α
maximal atlas, both ϕα ◦ ϕ−1 and ψ ◦ ψβ−1 are C ∞ . Hence, the composite
1 S MOOTH M ANIFOLDS 9

ψ ◦ F ◦ ϕ−1 = (ψ ◦ ψβ−1 ) ◦ (ψβ ◦ F ◦ ϕ−1 −1

α ) ◦ (ϕα ◦ ϕ )

is C ∞ at ϕ(p).

Theorem 1.24 (Smoothness in terms of atlases)

Let M and N be smooth manifolds. Let F : M → N be a continuous map. Then the
following statements are equivalent.

1. The map F : M → N is C ∞ .

2. There exist atlases U of M and V of N such that for every (U, ϕ) ∈ U and (V, ψ) ∈ V,
the map
ψ ◦ F ◦ ϕ−1 : ϕ(U ∩ F −1 (V )) → Rn

is smooth.

3. For every chart (U, ϕ) of M and (V, ψ) of N , the map

ψ ◦ F ◦ ϕ−1 : ϕ(U ∩ F −1 (V )) → Rn

is smooth.

Theorem 1.25 (Compostion of Smooth Maps is Smooth)

F G
Let M −
→N −
→ P be a sequence of smooth manifolds and maps. If F and G are smooth,
then so is G ◦ F.

Proof. We shall use the third part of the last theorem. Let (U, ϕ) be a chart of M , and
(W, σ) be a chart of P . For any chart (V, ψ) of N , we may write

σ ◦ G ◦ F ◦ ϕ−1 = σ ◦ G ◦ ψ −1 ◦ ψ ◦ F ◦ ϕ−1 ,

which makes the composite map smooth by smoothness of F and G.

Definition 1.26
Let M , N be smooth manifolds. A map F : M → N is called a diffeomorphism if F
is smooth, invertible and if F −1 : N → M is smooth.

Coordinate maps are diffeomorphisms, and conversely, every diffeomorphism of an

open subset of a manifold with an open subset of Rn can serve as a coordinate map.
1 S MOOTH M ANIFOLDS 10

Theorem 1.27
If (U, ϕ) is a chart on a manifold M of dimension n, then the coordinate map ϕ : U →
ϕ(U ) ⊂ Rn is a diffeomorphism.

Proof. ϕ is a homeomorphism, so it sufficies to check the smoothness of ϕ and ϕ−1 . We

use the atlas {(U, ϕ)} on U , and {(ϕ(U ), idϕ(U ) )} on ϕ(U ). Now note that

idϕ(U ) ◦ ϕ ◦ ϕ−1 = idϕ(U )

which is smooth ϕ(U ) → ϕ(U ). Thus ϕ is smooth. For the other direction, we note that

ϕ ◦ ϕ−1 ◦ id−1
ϕ(U ) = idϕ(U )

which is smooth ϕ(U ) → ϕ(U ). Thus, ϕ−1 is smooth.

Theorem 1.28
Let U be an open subset of a manifold M of dimension n. If F : U → F (U ) ⊂ Rn is
a diffeomorphism onto an open subset of Rn , then (U, F ) is a chart in the differentiable
structure of M .

Proof. Clearly, F : U → F (U ) is a homeomorphism. Let M be the maximal atlas of M ,

and let (Uα , ϕα ) be a chart in M. Then ϕα and ϕ−1
α are both smooth maps. By smoothness
of F , F ◦ ϕ−1
α and ϕα ◦ F
−1 are both smooth. Hence, (U, F ) is compatible with the

maximal atlas. By the maximality of the atlas, the chart (U, F ) is in the atlas.

Theorem 1.29 (Smoothness in terms of coordinates)

Let F : M → N be a continuous map between two manifolds of dimensions n and m
respectively. The following are equivalent:

(i) The map F : M → N is C ∞ .

(ii) The manifold N has an atlas such that for every chart (V, ψ) = (V, y 1 , . . . , y m ) in
the atlas, the vector-valued function ψ ◦ F : F −1 (V ) → Rm is C ∞ .

(iii) For every chart (V, ψ) = (V, y 1 , . . . , y m ) on N , the vector-valued function ψ ◦ F :

F −1 (V ) → Rm is C ∞ .

Proof. (ii) ⇒ (i): Let V be the atlas for N in (ii), and let A = {(U, ϕ)} be an arbitrary atlas
for M . For each chart (V, ψ) in the atlas U, the collection

{(U ∩ F −1 (V ), ϕ|U ∩F −1 (V ) )}
1 S MOOTH M ANIFOLDS 11

is an atlas for F −1 (V ). Since ψ ◦ F : F −1 (V ) → Rm is C ∞ ,

ψ ◦ F ◦ ϕ−1 : ϕ(U ∩ F −1 (V )) → Rm

is C ∞ . It then follows that F : M → N is C ∞ .

(i) ⇒ (iii): Being a coordinate map, ψ is C ∞ . As the composite of two C ∞ maps, ψ ◦ F is
C ∞.
(iii) ⇒ (ii): Obvious.

§1.4 Projective Spaces and Grassman Manifolds

Definition 1.30
An equivalence relation ∼ on a topological space X is said to be open if the quotient
map π : X → X/ ∼ is an open map.

Lemma 1.31
Let ∼ be an open equivalence relation. Then X/ ∼ is Hausdorff if and only if R = {(x, y) ∈
X × X : x ∼ y} is closed in X × X.

Proof. Suppose R is closed in X × X, and let x, y ∈ X/ ∼ with x ≁ y. Hence, (x, y) ∈

X × X \ R. By openness of X × X \ R, there exist open neighbourhoods U of x and V of
y such that (x, y) ∈ U × V ⊂ X × X \ R. By openness of π, π(U ) and π(V ) are open sets.
If π(U ) ∩ π(V ) is non-empty, there exist z1 ∈ U , z2 ∈ V with (z1 , z2 ) ∈ R, contradicting
the fact that R ∩ (U × V ) = ∅.

Conversely, assume that X/ ∼ is Hausdorff. We shall show that X × X \ R is open. Let

(x, y) ∈ X × X \ R. Then x and y are distinct points in X/ ∼. Let U
‹ and V
‹ be disjoint
open neighbourhoods of x and y respectively, and let U = π −1 (U
‹) and V = π −1 (V
‹).
Since U ‹ are disjoint, (U × V ) ∩ R = ∅.
‹ and V

Lemma 1.32
Let f : X → Y be an open continuous map, and let B be a basis of X. Then Be = {f (U ) :
U ∈ B} is a basis of Y .

Proof. Let O be an open subset of Y . Then f −1 (O) is open in X, and is thus expressed as
[
f −1 (O) = Uα .
α
S
Now, O = α f (Uα ).
2 TANGENT V ECTORS 12

Corollary 1.33
Let ∼ be an open equivalence relation on a topological space X, and let π : X → X/ ∼ be
the quotient map. If X is second countable, so is X/ ∼ .

Definition 1.34
Let X = Rn+1 \ {0}, and on X define the equivalence relation x ∼ y if x = ty for
some non-zero t ∈ R. The quotient space X/ ∼ obtained is called the real projective
space of dimension n, denoted by P n (R).

§2 Tangent Vectors
§2.1 Geometric Tangent Vectors

Definition 2.1
A differentiable curve onto Rn is a differentiable function γ : I → Rn for some open
interval I containing 0.

Let p ∈ Rn . The tangent space of Rn at p is the collection

Dp = {γ ′ (0) : γ is a differentiable curve onto Rn with γ(0) = p}.

Dp is an R-linear space, with

γ ′ (0) + η ′ (0) = ξ ′ (0), where ξ(t) = γ(t) + η(t) − p, and

cγ ′ (0) = ζ ′ (0), where ζ(t) = γ(ct).

The tangent space of Rn helps us define directional derivatives in the following way.
Let U be an open subset of Rn , and V be an open subset of R. Let f : U → V be a C ∞
function. For γ ′ (0) ∈ Tp (Rn ), let us compute

d
(f ◦ γ(t)) .
dt t=0

By the chain rule of multivariable calculus, we have

f ′ (p)γ ′ (0),

which is precisely the directional derivative of f at p in the direction of γ ′ (0). Conversely,

for any v ∈ Rn , let η : (−ε, ε) → Rn be the straight line η(t) = p + tv, so that

d
(f ◦ η(t)) = f ′ (p)v.
dt t=0
2 TANGENT V ECTORS 13

Thus, the set Dp is in a one-one correspondence to directional derivatives, which are

maps Dp,v : C ∞ (Rn ) → R, sending

Dp,v (f ) = f ′ (p)v.

Directional derivatives satisfy

(i) Linearity: Dp,v (f + g) = Dp,v (f ) + Dp,v (g).

(ii) Leibniz Rule: Dp,v (f g) = g(p)Dp,v (f ) + f (p)Dp,v (g).

§2.2 Derivations

Definition 2.2
A linear map X : C ∞ (Rn ) → R is called a derivation at p if it satisfies

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

Let Tep Rn denote the collection of all derivations at p. It is clear that Tep admits a natural
R-linear space structure. We shall show that it is isomorphic to Dp .

Lemma 2.3
For any constant function c : Rn → R, X(c) = 0 for any derivation X at p ∈ Rn .

Proof. Let 1 denote the constant function 1. X(1) = X(1 · 1) = 1 · X(1) + 1 · X(1) =⇒
X(1) = 0 =⇒ X(c) = cX(1) = 0.

Theorem 2.4
Tep Rn is isomorphic to Dp as an R-linear space.

Proof. We had already seen that any directional derivative Dp,v is a a derivation. Now,
for any Xp ∈ Tep Rn , we shall show that Xp = Dp,v for some v ∈ Rn .

For any vector v = v i ei in Rn , and f ∈ C ∞ (M ), Dp,v (f ) may be expressed as

n
X ∂f
Dp,v (f ) = vi .
∂xi
i=1

Thus, it suffices to find n real numbers v i . Let v i = X(xi ), where xi : Rn → R is the map
(u1 , . . . , un ) 7→ ui . We shall show that X = Dp,v . For any smooth function f : Rn → R
2 TANGENT V ECTORS 14

and x in a sufficiently small neighbourhood of p, we may write by Taylor’s theorem

n Z 1
X
i ∂f
f (x) = f (0) + x (tx + (1 − t)p)dt.
0 ∂xi
i=1

Applying X, we obtain

n Z 1 n
X
i ∂f X ∂f
X(f ) = X(x ) (tx + (1 − t)p)dt = v i i (p).
∂xi ∂x
i=1 |0 {z } i=1
x=p

Definition 2.5
Let M be a smooth manifold of dimension n. A linear map X : C ∞ (M ) → R is
called a derivation at p ∈ M if it satisfies

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

for all f, g ∈ C ∞ (M ).

The collection of all derivations at p is a vector space, called the tangent space of M at p.
Any element of the tangent space is called a tangent vector at p.

Let (U, ϕ) = (U, x1 , . . . , xn ) be a chart of M . The tangent vector to the map x1 : U → R

∂
at p ∈ U is denoted by .
∂xi p

§2.3 Bump Functions

Definition 2.6
Let M be a smooth manifold. Suppose q ∈ M , and U is a neighbourhood of q. A
bump function at q supported on U is a smooth nonnegative function ρ : M → R
that is equal to 1 in a neighbourhood of q and supp f ⊂ U.

Theorem 2.7
Let M be a smooth manifold, let p ∈ M , and let U be a neighbourhood of q. Then a bump
function at q supported on U exists.

Proof. Let (a, b) be a bounded open interval in R.

2 TANGENT V ECTORS 15

Step 1: Let f : R → R,

e−1/x x>0
f (x) =
0 x ≤ 0.

f is smooth function supported on (0, ∞).

Step 2: Define g : (0, 1) → R,

f (x)
g(x) = .
f (x) + f (1 − x)

Step 3: Define ρ(a,b) : R → R,

b2 − x2
h(x) = .
b2 − a2

ρ is a smooth function supported on (a, b).

Step 4: A smooth function supoported on B(p, r) in Rn can be constructed as

σ(x) = ρ(0,r) (∥x − p∥).

U is contained in some chart (U e U . Then ϕ : U → ϕ(U ) is a

e about p. Let ϕ = ϕ|
‹, ϕ)
homeomorphism. Let σ be a smooth function at ϕ(p) supported on ϕ(U ). Then

σ ◦ ϕ(q) q ∈ U
ρ(q) =
0 q∈/ U.

is a bump function at p supported in U .

Theorem 2.8
Suppose f is a smooth function defined on a neighbourhood U of a point p in a manifold.
Then, there exists a C ∞ function fe on M that agrees with f in some (possibly smaller)
neighbourhood of p.

Proof. Let ρ : M → R be a bump function at p supported on U , and consider


f (x)ρ(x) x∈U
fe(x) =
0 x∈
/ U.
2 TANGENT V ECTORS 16

Lemma 2.9
Let M be a smooth manifold, p a point in M and let X ∈ Tp M be fixed. If f and g are two
smooth functions on M that agree on some neighbourhood U of p, then X(f ) = X(g).

Proof. Let h = f − g, then h ≡ 0 inside U . Let ρ be a bump function at p supported on U ,

and let η = 1 − ρ. Let V be an open subset of U such that η = 0 on V . Clearly, h = hη on
M . It is clear that X(h) = 0 =⇒ X(f ) = X(g).

§2.4 Pushforward of Smooth Map

Definition 2.10
Let M and N be smooth manifolds. Let F : M → N be a smooth map. The
pushforward of F is defined as

F∗ : Tp M → Tf (p) N, F∗ (X)(f ) := X(F ∗ f ) = X(f ◦ F ).

Theorem 2.11
Let F : M → N and G : N → P be smooth maps and let p ∈ M . Then

(i) F∗ is linear.

(ii) (G ◦ F )∗ = G∗ ◦ F∗ .

(iii) (idM )∗ = idTp M .

(iv) F is a diffeomorphism =⇒ F∗ is an isomorphism, and (F∗ )−1 = (F −1 )∗ .

Proof. (ii): (G◦F )∗ (X)(f ) = X((G◦F )∗ f ) = X(F ∗ (G∗ f )) = F∗ (X)(G∗ f ) = G∗ (F∗ X)(f ) =
(G∗ ◦ F∗ )(f ).
(iii): (idM )∗ (X)(f ) = X(f ◦ idM ) = X(f ) =⇒ (idM )∗ (X) = X.
(iv): F ◦ F −1 = idN =⇒ F∗ ◦ (F −1 )∗ = idTp M .

Theorem 2.12
Let U be an open subset of a smooth manifold M and let ι : U → M be the inclusion map.
Then ι∗ : Tp U → Tp M is an isomorphism.

Proof. To show injectivity, it suffices to show ker ι∗ = 0. Let X ∈ Tp U such that ι∗ X = 0.

Let f be a smooth function on U . By Lemma 2.9, we can extend f to all of M to some fe.
Since ι∗ X = 0, we have
X(fe ◦ ι) = 0 =⇒ X(f ) = 0.
3 S UBMANIFOLDS 17

Since f is arbitrary, X = 0.

To show surjectivity, let Y ∈ Tp M be given. For every f ∈ C ∞ (U ), find fe ∈ C ∞ (M )

such that fe ◦ ι = f. Defining X ∈ Tp U , X(f ) = Y (fe), we are done.

Definition 2.13
Let M, N be smooth manifolds. A function f : M → N is said to be a local
diffeomeorphism if for each p ∈ M , there exists an open set U containing p such that
f : U → f (U ) is a diffeomorphism.

Theorem 2.14
Let M, N be smooth manifolds. Let f : M → N be a local diffeomorphisn. Then f∗ :
Tp M → Tp N is an isomorphism.

Theorem 2.15
Let M be a smooth manifold of dimension n. Let (U, ϕ) be a chart in M . Then ϕ : U → Rn
is a local diffeomorphism.

Corollary 2.16
The dimension of Tp M is equal to the dimension of M .

§3 Submanifolds
§3.1 The Rank Theorem

We first state a version of the theorem for Euclidean spaces.

Theorem 3.1 (Euclidean Rank Theorem)

Let A0 ⊂ Rm , and B0 ⊂ Rn be open sets, and let F : A0 → B0 be a smooth map, and
suppose DF (x) has rank k for every x ∈ A0 . Let a ∈ A0 , and b = F (a). Then there exist
open sets A ⊂ A0 containing a, B ⊂ B0 containing b, open sets U ⊂ Rm and V ⊂ Rn ,
and diffeomorphisms G : A → U and H : B → V such that F (A) ⊂ B and

H ◦ F ◦ G−1 (x1 , . . . , xk , xk+1 , . . . , xm ) = (x1 , . . . , xk , 0, . . . , 0).

| {z }
n−k times

Proof. Let F = (f1 , . . . , fn ). Since rank DF (a) = k, without loss of generality, the
3 S UBMANIFOLDS 18

differential of F at a is of the form

 
∂f1 ∂f1 ∂f1 ∂f1
··· ···
 ∂x1 ∂x2 ∂xk ∂xm 
 .. .. .. .. 
 . . ··· . ··· . 
 ∂f ∂fk ∂fk ∂fk 

DF (a) = 
 ∂x1
k
∂x2 ··· ∂xk ··· ∂xm 
 . .. .. .. 
 .. . ··· . ··· . 
 
∂fn ∂fn ∂fn ∂fn
∂x1 ∂x2 ··· ∂xk ··· ∂xm

where the top left k × k block is invertible. Let G : A0 → Rm as

G(u1 , . . . , um ) = (f1 (u1 , . . . , um ), . . . , fk (u1 , . . . , um ), uk+1 , . . . , um ).

The differential of G at a is given by

∂(f1 ,...,fk )
" #
∂(x1 ,...,xk ) 0
DG(a) = .
∗ I(m−k)×(m−k)

Clearly, the rank of DG(a) of m and hence is invertible. By the inverse function theorem,
there exist open sets A ⊂ A0 containing a and V ⊂ Rm such that G : A → V is a
diffeomorphism. By making A a sufficiently small neighbourhood of p

Theorem 3.2 (Rank Theorem)

Let F : M → N be a smooth map and suppose dim M = m, dim N = n. Let p be a point
in M and let rank F∗ = k on a neighbourhood of p. Then there exist charts (U, ϕ) about
p and (V, ψ) about F (p) such that F (U ) ⊂ V and F̂ = ψ ◦ F ◦ ϕ−1 : ϕ(U ) → ψ(V ) is
given by
F̂ (u1 , . . . , um ) = (u1 , . . . , uk , 0, . . . , 0).
| {z }
n−k times

§3.2 Immersions, Submersions and Embeddings

Definition 3.3
A smooth map F : M → N is called an immersion if F∗p is injective for every
p ∈ M.

Definition 3.4
A smooth map F : M → N is called a submersion if F∗p is surjective for every
p ∈ M.
3 S UBMANIFOLDS 19

Example 3.5 (Canonical Submersion)

Let F : Rn+k → Rn be defined as F (x1 , . . . , xn , xn+1 , . . . , xn+k ) = (x1 , . . . , xn ). This
map is a submersion, and is known as the canonical submersion.

Example 3.6 (Canonical Immersion)

Let F : Rn → Rn+k be defined as F (x1 , . . . , xn ) = (x1 , . . . , xn , 0, . . . , 0). This map is
| {z }
k times
an immersion, and is known as the canonical immersion.

Example 3.7 (Regular curves)

Let γ : (a, b) → Rm be a regular curve, i.e., let γ ′ (t) ̸= 0 for every t ∈ (a, b). Then
γ∗ : R → Rn is of rank 1, hence injective. Therefore, γ is an immersion.

Example 3.8 (Inclusion)

Let U ⊂ M be an open set, and let ι : U → M be the inclusion map. Then ι is both
an immersion and a submersion.

Example 3.9 (Graph)

Let f : Rn → R be a smooth function, and let Γ(f ) = {(x, F (x)) : x ∈ Rn }. Then the
map x 7→ (x, F (x)) is an immersion.

In the last example, the map is both an immersion and injective by its own right. The
image of such a map turns out to be interesting. We could always, for example, give a
smooth manifold structure on the image.

Let F : M → N be an injective immersion. Then F (M ) is said to be an immersed

submanifold, and it is equipped with the following topology: a subset V of F (M ) is
said to be open if F −1 (V ) is open in M . If (U, ϕ) is a chart about p, then (F (U ), ϕ ◦ F −1 )
is a chart about F (p), with respect to which F : M → F (M ) is a smooth map.
3 S UBMANIFOLDS 20

Example 3.10 (Figure Eight)

Let F : R → R2 be defined as
  π  π 
F (t) = 2 cos t − , sin 2 t − .
2 2

F is a regular curve and hence an immersion. Find a strictly increasing smooth

function g : R → R such that g(0) = π, limt→−∞ g(t) = 0, and limt→∞ g(t) = 2π.
Letting G = F ◦ g, we obtain that G(R) = F (R) is an immersed submanifold.

However, F (R) is not locally Euclidean when equipped with the subspace topology
of R2 .

Definition 3.11
An embedding is a smooth map F : M → N which is an injective immersion and a
homeomorphism M → F (M ) where F (M ) is equipped with the subspace topology
of N .

Recall that a map f : X → Y between topological spaces is said to be proper if f −1 (K)

is compact for every compact K ⊂ Y.

Theorem 3.12
Suppose F : M → N is a proper injective immersion. Then F is an embedding.

Proof. We prove that F is a closed map. Then F : M → F (M ) is closed bijective map,

hence a homeomorphism.
Let K ⊆ M be a closed subset. To show that F (K) is closed in N , we show that it
contains all of its limit points. Let y be a limit point of F (K), and let U be a precompact
neighbourhood of y. Then y is also a limit point of F (K) ∩ U . Because F is proper,
F −1 (U ) is compact, which implies that K ∩ F −1 (U ) is compact. Because F is continuous,
F (K ∩ F −1 (U )) = F (K) ∩ U is compact and therefore closed in N . In particular,
y ∈ F (K) ∩ U ⊆ F (K), so F (K) is closed.

Here is a useful criterion to determine when a map is proper. A sequence (pn )n∈N in a
topological space X is said to escape to infinity if for every compact K ⊂ X, there are at
most finitely many n ∈ N for which pn ∈ K.

Theorem 3.13
Suppose M and N are topological manifolds. A continuous map F : M → N is proper
if and only if for every sequence (pn )n∈N in M that escapes to infinity, (F (pn ))n∈N also
escapes to infinity.
3 S UBMANIFOLDS 21

Proof. By the Urysohn’s metrization theorem1 , a topological manifold is metrizable. For

any metric space, this can be seen to be true.

Remark 3.14. An embedding is not necessarily proper. Consider the map ι :

B(0, 1) → R. This is an embedding since ι∗ is an isomorphism, and ι is injec-
tive. But, ι−1 (B[0, 1]) = B(0, 1) which is not compact.

§3.3 Regular Submanifolds

The goal of this section is to establish that the set of zeroes of a smooth real-valued func-
tion on a manifold constitute a submanifold of a particularly well-behaved kind.

Definition 3.15
Let U ⊂ Rn be an open subset. A k-slice of U is a set S of the form

S = {(x1 , . . . , xk , xk+1 , . . . , xn ) ∈ U : xk+1 = ck+1 , . . . , xn = cn }

where ck+1 , . . . , cn are constants.

Definition 3.16
Let M be an n-dimensional manifold. Let (U, ϕ) be a smooth chart on M . If S ⊂ U
is such that ϕ(S) is a k-slice of ϕ(U ), then S is said to be a k-slice of U .

Definition 3.17
A subset S ⊂ M is called a regular submanifold if for each point p ∈ S, there exists
a smooth chart (U, ϕ) of M such that p ∈ U and U ∩ S is a k-slice of U .

We call such a chart (U, ϕ) in N a slice chart relative to S. On U ∩ S, note that ϕ =

(x1 , . . . , xk , ck+1 , . . . , cn ). Let

ϕS : U ∩ S → Rk

be the restriction of the first k components of ϕ to U ∩ S, that is, ϕS = (x1 , . . . , xk ). Note

that (U ∩ S, ϕS ) is a chart for S in the subspace topology.

1
a second countable regular space is metrizable
3 S UBMANIFOLDS 22

Theorem 3.18
Let S be a regular submanifold of N and U = {(U, ϕ)} a collection of compatible slice
charts of N that covers S. Then {(U ∩ S, ϕS )} is an atlas for S. Therefore, a regular
submanifold is itself a manifold. If N has dimension n and S is locally defined by the
vanishing of n − k coordinates, then dim S = k.

Proof. Let (U, ϕ) = (U, x1 , . . . , xn ) and (V, ψ) = (V, y 1 , . . . , y n ) be two slice charts in the
given collection. Assume that they intersect. In any slice chart relative to a submanifold
S it is possible to renumber the coordinates so that the last n − k coordinates vanish on
points of S. Then for p ∈ U ∩ V ∩ S,

ϕ(p) = (x1 , . . . , xk , 0, . . . , 0) and ψ(p) = (y 1 , . . . , y k , 0, . . . , 0),

so

ϕS (p) = (x1 , . . . , xk ) and ψS (p) = (y 1 , . . . , y k ).

Therefore,

(ψS ◦ ϕ−1 1 k 1 k
S )(x , . . . , x ) = (y , . . . , y ).

Since y 1 , . . . , y k are C ∞ functions of x1 , . . . , xk (because ψ ◦ ϕ−1 (x1 , . . . , xk , 0, . . . , 0) is

C ∞ ), the transition function ψS ◦ ϕ−1 ∞ 1 k
S is C . Similarly, since x , . . . , x are C
∞ functions

of y 1 , . . . , y k , ϕS ◦ ψS−1 is also C ∞ . Hence, any two charts in {(U ∩ S, ϕS )} are C ∞

compatible. Since {U ∩ S}U ∈U covers S, the collection {(U ∩ S, ϕS )} is a C ∞ atlas on
S.

Theorem 3.19
If f : M → N is an embedding, then its image f (M ) is a regular submanifold.

Proof. Let p ∈ M and f (p) ∈ N . By the rank theorem, there exist charts (U, ϕ) at p and
(V, ψ) at f (p) such that

ψ ◦ f ◦ ϕ−1 (x1 , . . . , xn ) = (x1 , . . . , xm , 0, . . . , 0).

Clearly, ψ(f (U ) ∩ V ) is a k-slice. Note that since f : M → f (M ) is a homeomorphism,

f (U ) is open in f (M ). There exists an open neighbourhood V ′ ⊂ N such that f (U ) =
′
f (M ) ∩ V ′ , which means f (U ) ∩ V = f (M ) ∩ V
| ∩
{zV} . Now take (V , ψ|V‹ ) to be the slice
‹
V
‹
chart at f (p), hence f (M ) is a regular submanifold.
3 S UBMANIFOLDS 23

Corollary 3.20
Let f : M → N be an immersion. Then f is locally an embedding.

Theorem 3.21
Let M be a regular submanifold of M . The the inclusion ι : M → N is an embedding.

Thus, a subset N of M is a regular submanifold if and only if it is an embedded subman-

ifold.

§3.4 The Level Set Theorem

Theorem 3.22
Let F : M → N be a smooth map, and let dim M = m and dim N = n. If F has a constant
rank k on M and q ∈ im F , then F −1 (q) is a closed regular submanifold of dimension
n − k.

Proof. Let A = F −1 (q). A is clearly a closed set. Let p ∈ A, and apply the rank theorem
to find charts (U, ϕ) at p and (V, ψ) at q such that ψ(q) = 0 and

ψ ◦ F ◦ ϕ−1 (x1 , . . . , xm ) = (x1 , . . . , xm , 0, . . . , 0).

Now note that

A ∩ U = {p′ ∈ U : F (p′ ) = q}
= {p′ ∈ U : F (ϕ−1 ϕ(p′ )) = q}
= {p′ ∈ U : ψ ◦ F ◦ ϕ−1 (ϕ(p′ )) = q}
= {p′ ∈ U : ϕ(p′ ) = (0, . . . , 0, xk+1 , . . . , xm )}
| {z }
k times

which makes (U, ϕ) is a slice chart of A at p.

Example 3.23
Pn
Let F : Rm \ {0} → R, F (x) = 2
i=1 xi , then S n = F −1 (1) is a regular submanifold.
3 S UBMANIFOLDS 24

Definition 3.24
Let F : M → N be a smooth map.

(a) A point p ∈ M is called a regular point if F∗p is surjective, else p is said to be a

critical point.

(b) A point q ∈ F (M ) is called a regular value if every point p ∈ F −1 (q) is a

regular point.

(c) A point q ∈ F (M ) is called a critical value if there exists p ∈ F −1 (q) which is

a critical point.

If q is a regular value then F −1 (q) is called a regular level set.

Corollary 3.25
If F : M → N is a submersion then each level set is a closed regular submanifold whose
codimension is equal to the dimension of N .

Proof. Every smooth submersion has constant rank equal to the dimension of its codomain.

This result should be compared to the corresponding result in linear algebra: if L : Rm →

Rr is a surjective linear map, then the kernel of L is a linear subspace of codimension r by
the rank-nullity law. The vector equation Lx = 0 is equivalent to r linearly independent
scalar equations, each of which can be thought of as cutting down one of the degrees of
freedom in Rm , leaving a subspace of codimension r. In the context of smooth manifolds,
the analogue of a surjective linear map is a smooth submersion, each of whose (local)
component functions cuts down the dimension by one.

Corollary 3.26 (Regular Level Set Theorem)

Let F : M → N be a smooth map. Then any regular level set of F is a closed regular
submanifold of M .

Proof. Let F : M → N be a smooth map and let c ∈ N be a regular value. The set U
of points p ∈ M where rank dFp = dim N is open in M , and contains F −1 (c) because
of the assumption that c is a regular value. It follows that F |U : U → N is a smooth
submersion, and the preceding corollary shows that F −1 (c) is an embedded submanifold
of U . Since the composition of smooth embeddings F −1 (c) ,→ U ,→ M is again a smooth
embedding, it follows that F −1 (c) is an embedded submanifold of M , and it is closed by
continuity.
4 T HE TANGENT B UNDLE 25

§3.4.1 Tangent Space of a Regular Level Set

Lemma 3.27
Let F : M → N be a smooth map, let q ∈ N is a regular value, and let S = F −1 (q). Then

Tp S = ker F∗,p

for every p ∈ S.

Proof. Let ι : S → M be the inclusion map. Then F ◦ ι : S → N is a constant map and

hence F∗,p ◦ ι∗,p = 0 which means

im(ι∗p ) ⊂ ker F∗p .

Furthermore, their dimensions are equal.

§3.5 Matrix Manifolds

We shall show that SLn (R) is a closed regular submanifold of GLn (R). Consider the
map
F = det : GLn (R) → R, A 7→ det A.

Clearly, SLn (R) = det−1 (1). Let mij denote the (i, j)th minor of A. Then
X
det A = (−1)i+j aij mij .
i,j

Then
∂F
= (−1)i+j mij = (A−1 )ji .
∂aij
If F admits a critical value then mij = 0 for every 1 ≤ j ≤ n =⇒ det A = 0, a
contradiction since A ∈ GLn (R). Thus, F is a submersion, hence SLn (R) is a regular
submanifold of GLn (R) of dimension of n2 − 1.

§4 The Tangent Bundle

A vector bundle over a smooth manifold M is a pair (E, π) such where E = {Ep : p ∈ M }
is a family of vector spaces and π : E → M is a smooth map sending Ep → p. A section
of the vector bundle E is a smooth assignment p → Ep .
4 T HE TANGENT B UNDLE 26

Definition 4.1
Let M be a smooth manifold. We define the tangent bundle T M as the disjoint
union of its tangent spaces
def
a
TM = Tp M.
p∈M

§4.1 Topology on the Tangent Bundle

Let (U, ϕ) = (U, x1 , . . . , xn ) be a chart in M , a smooth manifold of dimension n. We shall

define a smooth structure on T U , where U is an open subset of M . Any v ∈ Tp M can be
expressed as
n
X ∂
v= ci . (1)
∂xi p
i=1

Let π : T M → M be the canonical projection map defined as π(vp ) = p. We shall define

a set of local coordinates on T M in the following way. Let xi : T U → R be defined as

xi = xi ◦ π.

The map ϕe : T U → ϕ(U ) × Rn is thus defined as

ϕ(v
e p ) = (x1 (p), . . . , xn (p), c1 , . . . , cn ).

Define
ϕe−1 (u1 , . . . , un , c1 , . . . , cn ) = vp

where p = ϕ−1 (u1 , . . . , un ) and v = ni=1 ci ∂x ∂

P
i p . A subset A ⊂ T U is said to be open if

ϕ(A)
e is open in ϕ(U ) × Rn . If V is an open subset of U then the subspace topology of
e V : T V → ϕ(V ) × Rn .
T V is the same as the topology induced by ϕ|

Let τ = {A ⊂ T M : A ⊂ T U for some chart (U, ϕ) of M }. We show that τ is a basis for

some topology on T M . Let ẋi (vp ) = ci where vp = ni=1 ci ∂x
∂
P
i p.

Lemma 4.2
Let U and V be coordinate open sets in a manifold. If A is open in T U and B is open in
T V then A ∩ B is open in T (U ∩ V ).

Proof. A ∩ T (U ∩ V ) and B ∩ T (U ∩ V ) are open in T (U ∩ V ). Now,

A ∩ B = (A ∩ B) ∩ T (U ∩ V )
= (A ∩ T (U ∩ V )) ∩ (B ∩ T (U ∩ V )).
4 T HE TANGENT B UNDLE 27

Theorem 4.3 (Atlas on the tangent bundle)

Let {(U, x1 , . . . , xn )} be a C ∞ atlas on M . Then {(π −1 (U ), x1 ◦π, . . . , xn ◦π, ẋ1 , . . . , ẋn )}
is a C ∞ atlas on T M .

Lemma 4.4
The topology on the tangent bundle is Hausdorff.

Proof. Let vp and wq be two distinct points in T M .

e ) is an open subset of ϕ(U ) × Rn .

If p = q, let U be a coordinate chart at p. Then ϕ(U
By Hausdorffness of Rn , we find disjoint open sets A and B in Rn such that
(ẋ1 (vp ), . . . , ẋn (vp )) ∈ A, (ẋ1 (wp ), . . . , ẋn (wp )) ∈ B. Then ϕe−1 (ϕ(U ) × A) and
ϕe−1 (ϕ(U ) × B) are disjoint open sets in T M containing vp and wp respectively.

If p ̸= q, use the Hausdorffness of M to find disjoint charts (U, ϕ) and (V, ψ) about
p and q. Now, let U ‹ = π −1 (U ), V
‹ = π −1 (V ): they are clearly disjoint and open by
continuity of π.

Lemma 4.5
A manifold M has a countable basis consisting of coordinate open sets.

Proof. Let {(Uα , φα )} be the maximal atlas on M and B = {Bi } a countable basis for M .
For each coordinate open set Uα and point p ∈ Uα , choose a basic open set Bp,α ∈ B such
that
p ∈ Bp,α ⊂ Uα .

The collection {Bp,α }, without duplicate elements, is a subcollection of B and is therefore

countable.
For any open set U in M and a point p ∈ U , there is a coordinate open set Uα such that

p ∈ Uα ⊂ U.

Hence,
p ∈ Bp,α ⊂ U,

which shows that {Bp,α } is a basis for M .

4 T HE TANGENT B UNDLE 28

Theorem 4.6
The topology on the tangent bundle is second-countable.

Proof. Let {Ui }∞

i=1 be a countable basis for M consisting of coordinate open sets. Let φi
be the coordinate map on Ui . Since T Ui is homeomorphic to the open subset φi (Ui ) × Rn
of R2n and any subset of a Euclidean space is second countable, T Ui is second countable.
For each i, choose a countable basis {Bi,j }∞ ∞
j=1 for T Ui . Then {Bi,j }i,j=1 is a countable
basis for the tangent bundle.

Theorem 4.7
Let F : M → N be a smooth map. Then F∗ : T M → T N , defined by F∗ (vp ) := F∗p (vp )
is a smooth map.

Proof. Let

(U
‹, ϕ) ‹, x1 , . . . , xn , ẋ1 , . . . , ẋn ) and (V
e = (U ‹, ψ) ‹, y 1 , . . . , y n , ẏ 1 , . . . , ẏ n )
e = (V

be coordinate charts in T M and T N respectively. Then:

ψe ◦ F∗ ◦ ϕe−1 (p1 , . . . , pn , c1 , . . . , cn ) = ψe ◦ F∗ (vp )

= (y 1 (F (p)), . . . , y n (F (p)), ẏ 1 (F∗p (vp )), . . . , ẏ n (F∗p (vp ))).

Theorem 4.8
If M is a smooth n-manifold, and M can be covered by a single smooth chart, then T M is
diffeomorphic to M × Rn .

Proof. If (U, ϕ) is a global smooth chart for M then ϕ is, in particular, a diffeomorphism
from U ⊂ M to an open subset U ‹ ⊂ Rn .

§4.2 Vector Bundles

Recall that for any map f : E → B, and any point b ∈ B, the inverse image f −1 ({b}) is
said to be the fiber at b. The fiber of b is denoted by Eb .
5 V ECTOR F IELDS 29

Definition 4.9
Given two maps π : E → M and π ′ : E ′ → M , a map ϕ : E → E ′ is said to be
fiber-preserving if the following diagram commutes

ϕ
E E′
π π′
M

i.e., if π = ϕ ◦ π ′ .

Lemma 4.10
Let π : E → M and π ′ : E ′ → M be maps. Then, the map ϕ : E → E ′ is fiber-preserving
if and only if
ϕ(Ep ) ⊂ Ep′

for all p ∈ M .

§5 Vector Fields
§5.1 Smoothness

Definition 5.1
A vector field X on a manifold M is a function that assigns a tangent vector Xp ∈
Tp M to each point p ∈ M .

A vector field is said to be smooth if it is smooth as a map from M to T M .

Lemma 5.2
Let (U, ϕ) = (U, x1 , . . . , xn ) be a chart on a manifold M . A vector field X = ai ∂/∂xi
P

on U is smooth if and only if the coefficient functions ai are all smooth on U .

∼
Proof. Since ϕe : T U → U × Rn is a diffeomorphism, X : U → T U is smooth if and only
if ϕe ◦ X : U → U × Rn is smooth. For p ∈ U ,

(ϕe ◦ X)(p) = ϕ(X

e p)

= (x1 (p), . . . , xn (p), c1 (Xp ), . . . , cn (Xp ))

= (x1 (p), . . . , xn (p), a1 (p), . . . , an (p)).

As coordinate functions, x1 , . . . , xn are smooth on U . Thus, ϕe ◦ X is smooth if and only

5 V ECTOR F IELDS 30

if the ai ’s are.

Theorem 5.3
Let X be a vector field on a manifold M . The following are equivalent:

(i) The vector field X is smooth on M .

(ii) The manifold M has an atlas such that on any chart (U, ϕ) = (U, x1 , . . . , xn ) of
the atlas, the coefficients ai of X =
P i
a ∂/∂xi relative to the frame ∂/∂xi are all
smooth.

(iii) On any chart (U, ϕ) = (U, x1 , . . . , xn ) on the manifold M , the coefficients ai of

X = ai ∂/∂xi relative to the frame ∂/∂xi are all smooth.
P

Proof. (ii) ⇒ (i): Assume (ii). By the preceding lemma, X is smooth on every chart (U, ϕ)
of an atlas of M . Thus, X is smooth on M .
(i) ⇒ (iii): A smooth vector field on M is smooth on every chart (U, ϕ) on M . The
preceding lemma then implies (iii).
(iii) ⇒ (ii): Obvious.

A vector field X on a manifold M induces a natural map on the set of real-valued

def
functions on M , defined by X(f ) = Xf , where Xf (p) = Xp (f ).

Theorem 5.4
A vector field X on M is smooth if and only if for every smooth function f on M , the
function Xf is smooth on M .

Proof. (⇒) Suppose X is smooth and f ∈ C ∞ (M ). Then, on any chart (U, x1 , . . . , xn ) on

M , the coefficients ai of the vector field X = a ∂/∂xi are C ∞ . It follows that
P i

X ∂f
Xf = ai
∂xi

is C ∞ on U . Since M can be covered by charts, Xf is C ∞ on M .

(⇐) Let (U, x1 , . . . , xn ) be any chart on M . Suppose X =
P i
a ∂/∂xi on U and p ∈ U .
Then, for k = 1, . . . , n, each xk can be extended to a C ∞ function x
ek on M that agrees
with xk in a neighborhood V of p in U . Therefore, on V ,

xk
X  X 
∂ ∂e
ek =
Xx ai ek =
x ai = ak .
∂xi ∂xi

This proves that ak is C ∞ at p. Since p is an arbitrary point in U , the function ak is C ∞

on U .
5 V ECTOR F IELDS 31

Theorem 5.5
Suppose X is a C ∞ vector field defined on a neighborhood U of a point p in a manifold
M . Then there is a C ∞ vector field X
‹ on M that agrees with X on some possibly smaller
neighborhood of p.

Proof. Choose a C ∞ bump function ρ : M → R supported in U that is identically 1 in a

neighborhood V of p. Define

ρ(q)X
q for q ∈ U,
X(q)
‹ =
0 for q ∈
/ U.

The set of all smooth vector fields on M is denoted by X(M ).

§5.2 The Lie Bracket

If X and Y are two smooth vector fields on M , we can define a map XY (f ) := X(Y f )
for all f ∈ C ∞ (M ), but this is not a vector field since (XY )p will not be a derivation. We
can, however, define a sort of product of two vector fields in the following way.

Definition 5.6 (Lie bracket)

Let M be a manifold, and let X, Y ∈ X(M ). The Lie bracket [X, Y ] is defined as

[X, X]f := X(Y f ) − Y (Xf ).

Note that [Y, X] = −[X, Y ].

Theorem 5.7 (Jacobi Identity)

For any X, Y, Z ∈ X(M ), we have

[[X, Y ], Z] + [[Y, Z], X] + [[Z, X], Y ] = 0.

§5.3 Relatedness

If F : M → N is a diffeomeorphism, and X is a vector field on M , we define the

pushforward of X as
(F∗ X)F (p) = F∗,p (Xp ).
5 V ECTOR F IELDS 32

Definition 5.8
Let F : M → N be a smooth map. A vector field Y on N is said to be F -related to a
vector field X on M if
F∗,p (Xp ) = YF (p)

for all p ∈ M .

Theorem 5.9
Let F : M → N be a smooth map. A vector field Y on N is F -related to a vector field X on
M if and only if
(Y f ) ◦ F = X(f ◦ F )

for all f ∈ C ∞ (N ).

Proof.

(Y f )(F (p)) = YF (p) [f ]

= (F∗p Xp )[f ]
= Xp [f ◦ F ]
= (X(f ◦ F ))(p).

Theorem 5.10
Let F : M → N be smooth, and let X, Y ∈ X(M ), and X ′ , Y ′ ∈ KX(N ) be vector fields
such that X and Y are F -related to X ′ and Y ′ respectively. Then [X, Y ] is F -related to
[X ′ , Y ′ ].

Proof.

F ∗ ([X ′ , Y ′ ]f ) = F ∗ (X ′ Y ′ f − Y ′ X ′ f )
= F ∗ (X ′ Y ′ f ) − F ∗ (Y ′ X ′ f )
= X(F ∗ (Y ′ f )) − Y (F ∗ (X ′ f ))
= X(Y (F ∗ f )) − Y (X(F ∗ f ))
= [X, Y ](F ∗ f ).

§5.4 Lie Groups and Lie Algebras

5 V ECTOR F IELDS 33

Definition 5.11
Let G be a group which is also a smooth manifold. G is called a Lie group if the
maps
G × G → G, (g, h) 7→ gh and G → G, g 7→ g −1

are smooth.

(R, +) and (GLn (R), ·) are examples that come to mind.

Theorem 5.12
Let G be a Lie group and let H be an abstract subgroup that is also a regular submanifold of
G. Then H is also a Lie group.

Proof. We need to show that the multiplication map H × H → H and the inversion map
H → H are smooth. Since H is an abstract subgroup, it is closed under multiplication
and inversion, so it sufficies to show that the maps H × H → G and H → G are smooth.
Since H is a regular submanifold of G, the inclusion maps H → G and H × H → G × G
are smooth. Thus, the composition

H ×H →G×G→G
(h, h) 7→ (h, h) 7→ hh′

is smooth. Similarly, H → G → G, the assignment h 7→ h 7→ h−1 is smooth.

Theorem 5.13
Let G be a Lie group. Then the map Lg : G × G, h 7→ gh is a diffeomorphism, with
L−1
g = Lg −1 .

The pushforward of a vector field X ∈ X(G) is thus well-defined, as computed as

(Lg )∗ (X)(h) = (Lg )∗,h (Xh )

for all g, h ∈ G. Note that the map (Lg )∗,h : Tg G → Tgh G is the differential of Lg .

Definition 5.14
Let G be a Lie group. A vector field X ∈ X(G) is said to be left-invariant if

(Lg )∗ (X)(h) = Xgh

for all g, h ∈ G.
5 V ECTOR F IELDS 34

Definition 5.15
Let K be a field. A Lie algebra over K is a vector space g over K, equipped with a
Lie bracket [·, ·] : g × g → g defined that satisfies the following:

(i) (bilinearity):

[aX + bY, Z] = a[X, Z] + b[Y, Z]

[Z, aX + bY ] = a[Z, X] + b[Z, Y ]

(ii) (anticommutativity):
[X, Y ] = −[Y, X]

(iii) (Jacobi identity):

[[X, Y ], Z] + [[Y, Z], X] + [[Z, X], Y ] = 0.

The collection of all vector fields on a smooth manifold M provides a natural example of
a Lie algebra.

Definition 5.16
Let h be a linear subspace of a Lie algebra g. h is called a Lie subalgebra of g if h is
closed under Lie bracket.

Definition 5.17
Let g, h be Lie algebras over K. A K-linear map A : g → h is called a Lie algebra
homomorphism if
A[X, Y ] = [AX, AY ]

for all X, Y ∈ g.

Theorem 5.18
Let G be a Lie group. The collection of left-invariant vector fields on G is a Lie subalgebra of
X(G).

Proof. We need to show that if X, Y ∈ X(M ) are left-invariant, so is [X, Y ]. This is done
by showing that if F : M → N is a diffeomorphism, then F∗ [X, Y ] = [F∗ X, F∗ Y ]. Using

• F∗ [X, Y ] is the unique vector field in N which is F -related to [X, Y ],

• F∗ X is the unique vector field in N which is F -related to X,

• F∗ Y is the unique vector field in N which is F -related to Y ,

5 V ECTOR F IELDS 35

• if X and X ′ are F -related, Y and Y ′ are F -related, so are [X, Y ] and [X ′ , Y ′ ],

we have the desired result. In particular, (Lg )∗ [X, Y ] = [(Lg )∗ X, (Lg )∗ Y ].

The collection of all left-invariant vector fields on a Lie group G is denoted by Lie(G),
and called the Lie algebra associated with G.

Theorem 5.19
Let G be a Lie group, and let g = Lie(G). Then the evaluation map eve : g → Te (G),
X 7→ Xe is an isomorphism of vector spaces.

Proof. We explicitly define an inverse for eve . Define τ : Te (G) → g as

τ (v)(g) = (Lg )∗ (v).

To show that τ (v) is left-invariant, let h ∈ G be fixed and consider

(Lh )∗ τ (v)(g) = (Lh )∗ ◦ (Lg )∗ (v)

= (Lhg )∗ (v)
= τ (v)(gh).

§5.5 Pushforward of Left-Invariant Vector Fields

Because of the correspondence between left-invariant vector fields and tangent vectors
at the identity, it is possible to push forward left-invariant vector fields under a Lie group
homomorphism.

Let F : H → G be a Lie group homomorphism. A left-invariant vector field X on H is

generated by its value A = Xe ∈ Te H at the identity, so that X = A.
e Since a Lie group
homomorphism F : H → G maps the identity of H to the identity of G, its differential
F∗,e at the identity is a linear map from Te H to Te G.

Theorem 5.20
Let G, H be Lie groups, and let g, h be the associated Lie algebras. Let F : G → H be a Lie
group homomorphism. Then the map F∗ : g → h defined by

F∗ X = τ (F∗,e Xe )

is a homomorphism of Lie algebras.

6 PARTITIONS OF U NITY 36

§6 Partitions of Unity
§6.1 Definition and Existence

If {Ui }i∈I is a finite open over of a smooth manifold M , a smooth partition of unity
subordinate to {Ui }i∈I is a collection of nonnegative smooth real-valued functions
{ρi : i ∈ I} such that supp ρi ⊂ Ui and
X
ρi = 1.
i∈I

When I is an infinite set, we want the sum to make sense, and hence impose a local
finiteness condition.

Definition 6.1
A collection {Aα } of subsets of a topological space S is said to be locally finite if
every point q in S has a neighborhood that intersects only finitely many of the sets
Aα .

In particular, every q ∈ S is contained in only finitely many of the Aα ’s.

Definition 6.2
A smooth partition of unity {Uα } is a collection of nonnegative real-valued smooth
functions {ρα } such that

(i) the collection of supports {supp ρα } is locally finite.

P
(ii) α ρα = 1.

Given an open cover {Uα } of M , we say that a partition of unity {ρα } is subordinate
to {Uα } if for each ρα , there exists some Uα satisfying supp ρα ⊂ Uα .

Theorem 6.3 (Existence of a smooth partition of unity)

Let {Uα } be an open cover of a smooth manifold M . Then

(i) There is a smooth partition of unity {φk }∞

k=1 with every φk having compact support
such that for each k, supp φk ⊂ Uα for some α ∈ A.

(ii) If we do not require compact support, then there is a smooth partition of unity {ρα }
subordinate to {Uα }.

§6.2 Whitney Embedding Theorem

6 PARTITIONS OF U NITY 37

Theorem 6.4
Any compact manifold M can be embedded into RN for sufficiently large N .

Proof. Let {(φi , Ui , Vi )}1≤i≤k be a finite set of coordinate charts on M so that U = {Ui |
1 ≤ i ≤ k} is an open cover of M . Let {ρi | 1 ≤ i ≤ k} be a partition of unity subordinate
to U. Let φ̃i (p) = ρi (p)φi (p), extended to 0 outside Ui . Define

Φ : M → Rk(m+1) , p 7→ (φ̃1 (p), . . . , φ̃k (p), ρ1 (p), . . . , ρk (p)) .

We claim that Φ is an injective map. In fact, suppose Φ(p1 ) = Φ(p2 ). Take an index i so
that ρi (p1 ) = ρi (p2 ) ̸= 0. Then p1 , p2 ∈ supp(ρi ) ⊂ Ui . It follows that φi (p1 ) = φi (p2 ). So
we must have p1 = p2 since φi is bijective.
Next let’s prove that Φ is an immersion. In fact, for any Xp ∈ Tp M ,

Φ∗,p (Xp ) = (Xp (ρ1 )φ1 (p) + ρ1 (p)(ϕ1 )∗,p (Xp ), . . . , Xp (ρk )φk (p) + ρk (p)(ϕk )∗,p (Xp ), Xp (ρ1 ), . . . , Xp (ρk )) .

It follows that if Φ∗,p (Xp ) = 0, then Xp (ρi ) = 0 for all i, and thus ρi (p)(ϕi )∗,p (Xp ) = 0
for all i. Pick an index i so that ρi (p) ̸= 0. We see (ϕi )∗,p (Xp ) = 0. Since φi is a
diffeomorphism, we conclude that Xp = 0. So Φ∗ is injective.
Since Φ is an injective immersion, and M is compact, Φ must be an embedding.