Differential geometry

Lecture 4: Tangent spaces (part 2)

David Lindemann

University of Hamburg
Department of Mathematics
Analysis and Differential Geometry & RTG 1670

1. May 2020

David Lindemann DG lecture 4 1. May 2020 1 / 16

1 Tangent vectors from local coordinates

2 Differentials of smooth maps

3 Immersions, submersions, embeddings, local diffeomorphisms

David Lindemann DG lecture 4 1. May 2020 2 / 16

Recap of lecture 3:
defined vector space of smooth functions C ∞ (M)
defined bump functions (w.r.t. given data)
recalled tangent vectors & tangent space of Rn
defined tangent vectors for general smooth manifolds
showed that tangent vectors are local objects
erratum: nope!

David Lindemann DG lecture 4 1. May 2020 3 / 16

 Tangent vectors from local coordinates

How can we write down tangent vectors explicitly? Use local

coordinates!
Definition
Let ϕ = (x 1 , . . . , x n ) be local coordinates on
 a smooth
manifold M defined on U ⊂ M. Then ∂x∂ i p ∈ Tp M, p ∈ U, is
defined as
∂(f ◦ ϕ−1 )

∂  ∂f
i
(f ) := i
(p) := (ϕ(p)) ∀f ∈ C ∞ (M).
∂x p  ∂x ∂u i

∂ 

we can think of ∂x i p
as partial derivative at p with
respect to the chosen local coordinates (x 1 , . . . , x n )
alternative notations: ∂i |p , ∂x i |p
need to check the following:

Lemma
∂ 

∂x i p
is a well-defined tangent vector.

Proof: (next page)

David Lindemann DG lecture 4 1. May 2020 4 / 16
 Tangent vectors from local coordinates

need to check linearity and Leibniz rule:

linearity follows from linearity of partial derivatives in Rn
Leibniz rule: for f , g ∈ C ∞ (M) we calculate

∂(f · g )
(p)
∂x i
∂((f · g ) ◦ ϕ−1 )
= (ϕ(p))
∂u i
∂((f ◦ ϕ ) · (g ◦ ϕ−1 ))
−1
= (ϕ(p))
∂u i
−1
∂(f ◦ ϕ ) ∂(g ◦ ϕ−1 )
= g (p) (ϕ(p)) + f (p) (ϕ(p))
∂u i ∂u i
∂f ∂g
= g (p) i (p) + f (p) i (p)
∂x ∂x

Example
∂x j
(p) = δij ∀1 ≤ i ≤ n, 1 ≤ j ≤ n,
∂x i
follows from (x j ◦ ϕ−1 )(u 1 , . . . , u n ) = u j .

David Lindemann DG lecture 4 1. May 2020 5 / 16

 Tangent vectors from local coordinates

It turns out that any tangent

 vector can be written as a linear
combination of the ∂x∂ i p ’s in a unique way:
Proposition
(ϕ = (x 1 , . . . , x n ), o
For all p ∈ M and any local chart n U) with
∂ 

p ∈ U, the set of tangent vectors i , 1 ≤ i ≤ n is basis
∂x p
of Tp M.

Proof: 
∂ 
Linear independence of the ∂x i p
:
assume ∃ (c 1 , . . . , c n ) ∈ Rn 6= 0, such that
n 
X ∂ 
v0 := ci =0
i=1
∂x i p

as a linear map
then ∃ at least one 1 ≤ j ≤ n, such that c j 6= 0, and we
obtain
v0 (x j ) = c j 6= 0
which is a contradiction to v0 = 0
David Lindemann DG lecture 4 1. May 2020 6 / 16
 Tangent vectors from local coordinates

n  o
∂ 
Tp M = spanR ∂x i p
, 1≤i ≤n :
assume wlog ϕ(U) = Br (0) [after possibly shrinking U
and translating ϕ via constant vector −ϕ(p) ∈ Rn ]
for any g ∈ C ∞ (ϕ(U)) obtain using the fundamental
theorem of calculus and
Z1
∂g
gj (q) := (tq)dt ∀ q ∈ ϕ(U), 1 ≤ j ≤ n,
∂u j
0

the identity
n
X
g = g (0) + gj u j
j=1

∀ f ∈ C ∞ (U) get via g := f ◦ ϕ−1

n
X
f = g ◦ ϕ = f (p) + fj x j
j=1

David Lindemann DG lecture 4 1. May 2020 7 / 16

 Tangent vectors from local coordinates

(continuation of proof)

acting with ∂x∂ i p on both sides of above eqn. yields

∂f
fi (p) = (p) ∀1≤i ≤n
∂x i
for any v ∈ Tp M fixed get by using x i (p) = 0 for all
1 ≤ i ≤ n and the Leibniz rule
n
X ∂f
v (f ) = i
(p)v (x i )
i=1
∂x

hence
n 
X ∂ 
v= v (x i )
,
i=1
∂x i p
n  o
meaning that ∂x∂ i p , 1 ≤ i ≤ n spans Tp M as claimed

David Lindemann DG lecture 4 1. May 2020 8 / 16

 Tangent vectors from local coordinates

Immediate consequence:

Corollary
The dimensions of a smooth manifold M and its tangent
space Tp M coincide for all p ∈ M.

ci ∂ 
P
Question: How do the tangent vectors of the form ∂x i p
i
behave under a change of coordinates?

to answer this, we need differentials of smooth maps

David Lindemann DG lecture 4 1. May 2020 9 / 16

 Differentials of smooth maps

Definition
Let M be a smooth manifold of dimension m and N be a
smooth manifold of dimension n.
The differential at a point p ∈ M of a smooth function
f ∈ C ∞ (M) is defined as the linear map

dfp : Tp M → R, v 7→ v (f ).

In a given local coordinate system ϕ = (x 1 , . . . , x m ) on M

that covers p, dfp is of the form

∂  ∂f
dfp : 7→ (p).
∂x i p ∂x i

(continued on next page)

David Lindemann DG lecture 4 1. May 2020 10 / 16

 Differentials of smooth maps

Definition (continuation)
The differential at a point p ∈ M of a smooth map
F : M → N in given local coordinate systems
ϕ = (x 1 , . . . , x m ) on M and ψ = (y 1 , . . . , y n ) on N
covering p ∈ M and F (p) ∈ N, respectively, is defined as
the linear map
n
∂F j
 
∂  X ∂ 
dFp : Tp M → TF (p) N, →
7 (p) ,
∂x i p j=1
∂x i ∂y j F (p)

where we have used the notation F j := y j ◦ F . The rank

of F is the rank of the linear map dFp : Tp M → TF (p) N,
which coincides with the rank of the Jacobi matrix of F
at p in the local coordinate systems ϕ, ψ,
 j 
∂F
(p) ∈ Mat(n × m, R).
∂x i ji

[ j = row, i = column ]

David Lindemann DG lecture 4 1. May 2020 11 / 16

 Differentials of smooth maps

Similar to real analysis, differentials fulfils a chain rule:

Lemma
Let M, N, P be smooth manifolds and F : M → N,
G : N → P, smooth maps. Then

d(G ◦ F )p = dGF (p) ◦ dFp

for all p ∈ M.

Proof: For any v ∈ Tp M and f ∈ C ∞ (P) we have

d(G ◦ F )p (v )(f ) = v (f ◦ G ◦ F )
= dFp (v )(f ◦ G ) = dGF (p) (dFp (v ))(f ).

David Lindemann DG lecture 4 1. May 2020 12 / 16

 Differentials of smooth maps

Example
For any local coordinate system ϕ = (x 1 , . . . , x n ) on M
covering p ∈ M, dϕp : Tp M → Tϕ(p) Rn ∼ = Rn is of the form

dϕp = (dx 1 , . . . , dx n )p = (dxp1 , . . . , dxpn ),

 !
∂ 
dxpj = δij .
∂x i p

Viewed as n × n-matrix [by identifying dϕp with its Jacobi

matrix], dϕp = 1n .

David Lindemann DG lecture 4 1. May 2020 13 / 16

 Differentials of smooth maps

Lemma
For two local coordinate systems ϕ = (x 1 , . . . , x n ) and
ψ = (y 1 , . . . , y n ) with overlapping domain on M we obtain for
all p contained in the overlap the identity
n
∂y j
 
∂  X ∂ 
= (p) .
∂x i p j=1
∂x i ∂y j p

Proof:
For any f ∈ C ∞ (M), 1 ≤ i ≤ n, p in the overlap we calculate

∂f ∂(f ◦ ϕ−1 ) ∂(f ◦ ψ −1 ◦ ψ ◦ ϕ−1 )

i
(p) = i
(ϕ(p)) = (ϕ(p))
∂x ∂u ∂u i
n
X ∂(f ◦ ψ −1 ) ∂(u j ◦ ψ ◦ ϕ−1 )
= j
(ψ(p)) (ϕ(p))
j=1
∂u ∂u i
n
X ∂f ∂y j
= j
(p) i (p).
j=1
∂y ∂x

David Lindemann DG lecture 4 1. May 2020 14 / 16

 Immersions, submersions, embeddings, local diffeomorphisms

Properties of differentials are used to classify smooth maps as

follows:
Definition
A smooth map between smooth manifolds F : M → N is
called an immersion if dFp : Tp M → TF (p) N is injective
for all p ∈ M.
F : M → N is called a submersion if
dFp : Tp M → TF (p) N is surjective for all p ∈ M.
An immersion F : M → N is called an embedding if F is
injective and an homeomorphism onto its image
F (M) ⊂ N equipped with the subspace topology.
A smooth map F : M → N between smooth manifolds of
the same dimension is called a local diffeomorphism if for
all p ∈ M there exists an open neighbourhood of p,
U ⊂ M, such that F |U : U → N is a diffeomorphism onto
its image.

David Lindemann DG lecture 4 1. May 2020 15 / 16

 Immersions, submersions, embeddings, local diffeomorphisms

END OF LECTURE 4

Next lecture:
smooth manifold analogue to inverse function theorem
more examples of smooth maps
submanifolds of smooth manifolds

David Lindemann DG lecture 4 1. May 2020 16 / 16