0% found this document useful (0 votes)
9 views

Lecture 5

Uploaded by

enlightenedep
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
9 views

Lecture 5

Uploaded by

enlightenedep
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

General Relativity Fall 2019

Lecture 5: manifolds and tangent vectors


Yacine Ali-Haı̈moud
September 17th 2019

MANIFOLDS

In simple words, a manifold is a set that locally “looks like” Rn . A little more precisely, it is a set made of
patches that look like Rn , smoothly sewn together. Before giving the formal mathematical definition, we need to
define a few preliminary concepts:
• An open ball of Rn around a point y ∈ Rn , with radius r, is the set O(y; r) ≡ {x ∈ Rn such that |x − y| < r}.
• An open set of Rn is the union of an arbitrary number of open balls: U ⊂ Rn is open if ∀ y ∈ U, ∃ r > 0 such that
O(y; r) ⊂ U .
• A map φ : A → B is said to be one-to-one or injective (these are interchangeable) is no two elements of A have
the same image: φ(x) = φ(y) if and only if x = y.
• A map φ : A → B is said to be onto or surjective if all the elements of B can be written as the image of (at least)
one element of A: ∀ y ∈ B, ∃ x ∈ A such that y = φ(x).
• A map φ : A → B is bijective if it is both injective and surjective, in other words, if it is invertible. We
denote the inverse map φ−1 : B → A. Note that an injective map φ : A → B is always bijective when viewing it as a
map from A to φ(A) = {φ(x) such that x ∈ A}.

Let us now consider a general set M which is not Rn .


• A chart or coordinate system on a subset U ⊂ M is a one-to-one map φ : U → Rn such that the image set,
φ(U ) ⊂ Rn is open. The restricted map φ : U → φ(U ) is thus bijective, and we call φ−1 : φ(U ) ⊂ Rn → M its
inverse..
• An atlas on a set M is a collection of charts and associated domains, {(Ui , φi )}, such that:
(i) The domains Ui cover M, i.e. Union(Ui ) = M,
i
(ii) Overlapping charts are compatible, i.e. if two charts have overlapping domains U1 ∩ U2 6= ∅, then the maps
φ2 ◦ φ−1 −1 n n ∞
1 and φ1 ◦ φ2 , both from R to R , are infinitely differentiable (symbol: C ), see Fig. 1.

This allows us to finally define a manifold, as a set M with a maximal atlas, i.e. one that contains every possible
compatible chart. Note that n (the dimension of Rn ) must be the same everywhere on the manifold.

Examples of sets which are not manifolds (see Fig. 2): a 1-dimensional line attached to a 2-dimensional
surface, as this would require 2 different dimensions. A self-intersecting line, which would not “look like Rn at the
intersection point.

You can think of manifolds as smooth n−dimensional surfaces of Rp , with p > n. Indeed, Whitney’s embedding
theorem states that a manifold can always be mapped to a smooth n-dimensional surface of R2n (“embedded” in
R2n ). This is not necessarily helpful, however, unless you have good visualisation skills in 4 dimensions or higher.
Moreover, a manifold has well-defined intrinsic properties regardless of its embedding.

In general, a manifold cannot be covered with a single chart. Consider for instance a circle: you can
describe all of it except for a point by an angle θ ∈ (0, 2π). To cover the full circle, you need at least 2 charts. Note
that you cannot just use [0, 2π), as this would not be an open set of R. The same holds for a shpere.

We know what a C ∞ function of Rn → R, but it is not a priori obvious how to define this notion for functions
f : M → R. We can do so as follows: at any given point p ∈ M, find a chart {(U, φ)} such that p ∈ U – such a chart
must exist from condition (i). We can then define the function f ◦ φ−1 : Rn → R, see Fig. 3. The function f is said
to be C ∞ (or smooth) if these functions are all smooth – because of property (ii), the choice of φ does not matter in
the definition.
I will denote by F the set of all smooth functions from M → R (or, more generally, form a subset of M to R):

F ≡ {f : M → R such that f is C ∞ } .
oyally
round
in0kWh
Mn
Union
It
( Ui
) =
[ the Ui cover oh ]

¥
charts overlap ¢
2
wo
,
U ,
A Ua # ,
then

Rn
-0 ctfu Nz ) I loan
q
,

a
o
.

a .

pin

FIG. 1. Domains of various maps used to defined compatiblity of overlapping charts.

FIG. 2. Example of sets which are not manifolds: a line attached to a sphere, or a self-intersecting line.

TANGENT VECTORS

When we think of Rn , we often blur the line between points and vectors: after all, vectors are differences between
two points, so we can identify a point and a vector attached to the origin of coordinates.
On a general manifold M, there is no such notion of addition or subtraction of points. Intuitively, we can easily
visualize that vectors live in a space tangent to M, at any given point p ∈ M. In Rn , the tangent spaces at each
point are unambiguously mapped onto one another, and onto the original space Rn itself. On a general curved surface,
tangent spaces at each point are not parallel, and there is no obvious mapping between tangent spaces at
distant points – consider for instance the tangent plane at the pole of a sphere, and that at one point on the equator.
A function It ( )
f
-

:D
•• .
.
→ IR is smooth e.g E if fo¢* →
3R in e

Because c
of

infer
?
of atlas
IR
.

¥
fo ¢
-
I

Def
{ f.
F- Nsk ,
-
.

Rn

Tangent
• at
space of different
FIG. 3. Domains EM a smooth function on M.
pointused pto defined
maps a

In special relativity ,
we often blink
of the space of and
events
the
space of
Tangent vectors convey the notion of infinitesimal displacements along a given direction, at a given point
the
and familiar
manifold
p. For instance,asyou are probably with
some the unit
In vectors
G. Rêr , êθ ,events
êϕ in spherical
live polar coordinates:
the they indicate
which d
oe -

on
.

the direction towards which a point p(r, θ, ϕ) would move upon an infinitesimal increase of a given coordinate. This
.

, ,

infinitesimal business means that vectors have something to do with derivatives, and indeed, their formal definition
have vector . space structure ( it is not meaningful to Subtract two events
any
should remind you of derivatives.

Definition: aBut vector V illat alooks


locally point plike
#Vp
u1
'
tangent ∈ MIRis a an operator V : F → R which
.

, ,
(i) is linear: for any two smooth functions f, g, and any two real numbers α, β, V (αf + βg) = αV (f ) + βV (g),
.

(ii) satisfies Leibniz’ rule: V (f × g) = f (p) × V (g) + g(p) × V (f ).


I will denote by Vp the set of all tangent vectors at a point p ∈ M. This forms a vector space in the
mathematical sense of the term.

In the homework, you will show that any vector V acting on a constant function gives zero: V (constant) = 0.

A simple (and important) example of a vector at p is as follows: pick a coordinate system at p, i.e. a map
φ : U ⊂ M → Rn . For any point q ∈ u, φ(q) = {x1 , ..., xn } ∈ Rn , where the n real numbers {xµ } are the coordinates
of q. Then define the following operator on F:
(
F →R
∂(µ) |p : ◦φ−1 )
f 7→ ∂(f∂x µ (φ(p)).

This operator is linear, and satisfies Leibniz’ rule, hence is a tangent vector at p. From now on I will just write ∂(µ)
rather than ∂(µ) |p , but it should be implicit that this is defined at a specific point p. Rather than writing all the φ’s,
we will often write, for short,

∂f
∂(µ) f = .
∂xµ

COORDINATE BASES

Let us now show that the n vectors ∂(µ) form a basis of the tangent space Vp , meaning that (i) they are linearly
independent and (ii) they span Vp , i.e. any vector V can be written as V = V µ ∂(µ) .
As a preiminary, let’s define the n functions φµ (q) ≡ xµ , for q in the neighborhood of a point p ∈ M. These
functions are clearly smooth, and by construction, they are such that ∂(µ) φν = δµν .
Let us start by proving the linear independence of the ∂(µ) ’s. Suppose there exist n coefficients {αµ } such that
α ∂(µ) = 0, meaning that, for any function f ∈ F, αµ ∂(µ) f = 0. Applying this to any of the f ν functions, we see that
µ

all coefficients αν must be zero, which proves linear independence.


4

Now, given a fucntion f ∈ F, consider the function F (x) ≡ f ◦ φ−1 : Rn → R. This function is C ∞ , thus there exists

C functions Hµ (x) such that

F (x) = constant + (xµ − aµ )Hµ (x).

If we wanted, we could explicitly write the full function Hµ (x) in terms of the derivatives of F , but that’s not needed.
∂F
All we need to know is that Hµ (a) = ∂x µ (a).

Let’s apply this at a = φ(p) and x = φ(q), define hµ (q) ≡ Hµ (φ(q)), and recall that xµ = φµ (q): we get

f (q) = constant + (φµ (q) − φµ (p)) × hµ (q),

which we can also write as an equality between functions:

f = constant + (φµ − φµ (p)) × hµ .

Let us now apply a vector V ∈ Vp to this function, using linearity and Leibniz’ rule:

V (f ) = V (constant) + (φµ (p) − φµ (p)) × V (hµ ) + hµ (p) × V (f µ ) = hµ (p) × V (φµ ).

On the other hand, we have

∂(f ◦ φ−1 )
hµ (p) = Hµ (a) = (φ(p)) = ∂(µ) |p f.
∂xµ
Thus we found that

V (f ) = V µ ∂µ f, V µ ≡ V (φµ ) = V (xµ ) ⇒ V = V µ ∂(µ) .

In other words, we have showed that any tangent vector can be written as a linear combination of the partial derivative
operators. The components V µ are obtained by applying V to the coordinate functions xµ .

In summary, we have exhibited a basis of Vp : the n vectors {∂(µ) }. This is called a coordinate basis of the tangent
space. The fact that these n vectors form a basis implies that the tangent space is n-dimensional.

You might also like