A Brief Introduction to Manifolds

Luigi T. Sousa
May 2020

1
Contents
1 Motivation 3

2 A bit of mathematical background 6

2.1 Topological Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Homeomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Dual Vector Space . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Definitions 10
3.1 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Coordinates, Charts and Atlases . . . . . . . . . . . . . . . . . . 11
3.3 Differentiable Manifolds . . . . . . . . . . . . . . . . . . . . . . . 13

4 References 15

2
1 Motivation
Manifolds are, roughly speaking, a generalization of the notion of a one-
dimensional curve and a two-dimensional surface and of the notion of spaces in
general, they are basically the general idea of a n-dimensional space that can
be or not be curved (to be explained further). They are the object of study
in the field of Differential Geometry, and are essential for much of the modern
understanding of theoretical elementary physics, in particular when studying
gravity (space-time Einstein manifolds) and String Theory (Calabi-Yau and
Kähler manifolds).
Einstein’s theory of Special Relativity seemed very good at explaining the
dynamics of moving charges and of very fast-moving bodies, but it had a prob-
lem: it was incompatible with gravity (at least Newtonian gravity, which had an
instant effect, contradicting SR causality). Also, non-inertial frames were an-
other problem since there the postulates wouldn’t apply, needing some deeper
investigation.
Non-inertial frames of reference are, in a way, accelerated frames when com-
pared to an inertial frame, so there seems to be two sources of problem: accel-
eration and gravity. It turns out we actually only have one source of problem,
since by the Equivalence Principle, being accelerated in free space is actually
equivalent to being in free fall on a gravitational field.
To get a visual understanding, let us first consider a test particle moving at
constant speed through space-time:

Figure 1: Test particle p moving at speed v with respect to inertial observer O

Now, consider the same particle moving with a constant acceleration in the

3
negative x direction, or equivalently, under the influence of a gravitational field
generated by another body:

Figure 2: Test particle p moving with constant acceleration with respect to O

This is fine if we want to consider p frame of reference as ”invalid” since it’s

not inertial, but if we want to try and extend the postulates of SR to p reference
frame, things will get a bit messier. Let us first note that in our reference frame
(O), p’s coordinate lines seem to be curved:

Figure 3: Coordinate lines of p from O perspective

4
 But if we now switch perspective to p’s reference frame, now it’s O’s coor-
dinate lines that seem to be curved!

Figure 4: Coordinate lines of O from p perspective

Thus, it seems to be reasonable that we need some notion of curvature of

space-time itself to deal with this problems, and if we don’t want to evoke
unnecessary extra dimensions such that we can embed our space-time in some
higher dimensional space-time, then we need to use manifolds.

5
2 A bit of mathematical background
To study manifolds and Differential Geometry in a formal manner, some
mathematical background will be needed such that definitions and proofs can
make sense, so let’s see some of the needed information.

2.1 Topological Spaces

-Definition(2.1): A topological space is a ordered pair (X,τ ) where X is
a set and τ is a collection of subsets of X satisfying:
i) the empty set ∅ and X itself belong to τ ;
ii) the union of any arbitrary (finite OR infinite) members of τ is still a
member of τ ;
iii) the intersection of any f inite number of members of τ is still a member
of τ .
The elements of τ are called open sets of X and τ is said to give a topology
to X.

-Example: The real line R with the usual topology (i.e the collection of all
the open intervals (a, b) and their unions) form a topological space (R,τ ):
i) ∅ is automatically in τ (all sets contain ∅) and if you take a → −∞ and
b → ∞ you get the entirety of R;

ii) we defined τ such that we already include the union of the intervals;

iii) the intersection of any (finite) number of open intervals (and unions of)
will either be ∅ (already in τ ) or will be another possible open interval (also in
τ ).

2.2 Homeomorphism
-Definition(2.2): A homeomorphism is a function from a set X to a set
Y
f :X→Y
that satisfies:
i) f is a bijection;
ii) f is continuous;
iii) the inverse function f −1 is also continuous.

-Example: the function f (x) = x3 is a homeomorphism from R to itself since

√
it’s a continuous bijection and it’s inverse f −1 (y) = 3 y is also continuous.

2.3 Metric
-Definition(2.3): A metric on a set X is a function
d:X ×X →R

6
satisfying: (let x, y, z ∈ X)
i) d(x, y) ≥ 0, with the equality holding ⇔ x = y (positive definite);
ii) d(x, y) = d(y, x) (symmetric);
iii) d(x, y) ≤ d(x, z) + d(z, y) (triangle inequality).
A metric space is a pair (X, d) where X is a set equipped with the metric
d.

-Example: let X = R2 with the usual metric, d(−

→, −→ p
x 1 x2 ) = (x1 − x2 )2 + (y1 − y2 )2 ,
then we have:
i)
d(−
→, −
x →
1 x2 ) = 0 ⇒

p
⇒ (x1 − x2 )2 + (y1 − y2 )2 = 0 ⇔

⇔ (x1 − x2 )2 + (y1 − y2 )2 = 0 ⇔

⇔ x1 = x2 and y1 = y2 ⇒ −
→=−
x 1
→;
x 2

ii)

d(−
→, −→ p p
x 1 x2 ) = (x1 − x2 )2 + (y1 − y2 )2 = (x2 − x1 )2 + (y2 − y1 )2 =

= d(−
→, −
x →
2 x1 ) ;

iii) let α = x1 − x3 , β = x3 − x2 , γ = y1 − y3 , δ = y3 − y2 . Note that

(d(−
→, −
x → 2 2 2 2 2
1 x2 )) = (x1 − x2 ) + (y1 − y2 ) = (α + β) + (γ + δ) =

= α2 + β 2 + γ 2 + δ 2 + 2(αβ + γδ). (1)

Now let’s compare with (d(−
→, −
x → −
→ − → 2
1 x3 ) + d(x3 , x2 )) :

(d(−
→, −
x → −
→ − → 2 −
→ − → 2 −
→ − → 2 −
→ − → − → − →
1 x3 ) + d(x3 , x2 )) = (d(x1 , x3 )) + (d(x3 , x2 )) + 2d(x1 , x3 )d(x3 , x2 ) =

p
= α2 + β 2 + γ 2 + δ 2 + 2 (α2 + γ 2 )(β 2 + δ 2 ). (2)
So, to show the triangle inequality, then we must show that
p
(αβ + γδ) ≤ (α2 + γ 2 )(β 2 + δ 2 ).

Just note that

0 ≤ (αδ − βγ)2 = α2 δ 2 + β 2 γ 2 − 2αβγδ = (α2 + γ 2 )(β 2 + δ 2 ) − (αβ + γδ)2 ⇒

7
 ⇒ 0 ≤ (α2 + γ 2 )(β 2 + δ 2 ) − (αβ + γδ)2 ⇔

⇔ (αβ + γδ)2 ≤ (α2 + γ 2 )(β 2 + δ 2 ) ⇔

p
⇔ (αβ + γδ) ≤ (α2 + γ 2 )(β 2 + δ 2 ) ,
and hence the triangle inequality holds, so d is a valid metric.

*Note 1: a metric induces a topology on a set, called the metric topology, let’s
call it in this case τd , where the open sets are the ”open discs” given by

U (x) = {y ∈ X|d(x, y) < },

where x is a fixed point in X:

i) every set always contains ∅ and, if X is finite, we can take

 → max{d(x, y)|x, y ∈ X}

or if X is infinite we can take  → +∞ such that we can recover the entirety of

X, making itself a member τd ;
ii) since x is fixed, all discs are concentric, thus their union will just be the
largest of the discs involved, which is itself member of τd ;
iii) again, since x is fixed and all discs are concentric, their intersection will
be the smallest involved, which is a member of τd .
Every metric space is a topological space, but not all topological spaces are
metric spaces.

*Note 2: a inner product defined in a vector space V induces a metric in V ,

where the distance function d(x, y) is given by
p
d(x, y) = kx − yk = hx − y, x − yi.

2.4 Dual Vector Space

-Definition(2.4): given a vector space V , we can define the dual vector
space of V , denoted V ∗ , as the set of all the linear functions which maps an
element of V into the field K
f : V → K.
Since f is linear, we have that (if v = v 1 e1 + v 2 e2 + ... + v n en )

f (v) = v 1 f (e1 ) + v 2 f (e2 ) + ... + v n f (en ),

so if we know what f does to the basis vectors ei , we know what it does to any
vector v ∈ V .

8
 Since functions form a linear space, a linear combination of functions is still
a function
(a1 f1 + a2 f2 )(v) = a1 f1 (v) + a2 f2 (v),
so we can expand a member f ∈ V ∗ as fi ei , where the {ei } are the dual basis
induced by a basis {ej } of V such that

ei (ej ) = δji ,

where δji is the Kronecker delta defined as

(
i 1, if i = j
δj =
0, if i 6= j.

which, in mathematical physics language means that ei and ej are reciprocate

vectors, since we can interpret the action of f on v as an inner product:

f (v) = fi ei (v j ej ) = fi v j ei (ej ) = fi v j δji = fi v i .

If dim(V ) is finite and V is equipped with a inner product, then dim(V ) =

dim(V ∗ ) and there exists a isomorphism between V and V ∗

g :V →V∗
v j 7→ vi = gij v j ,
where g ∈ GL(n, K), that is, the group of linear transformations of dimension
n over the field K (basically a n × n matrix whose elements are in K), allowing
us to write the inner product as

g(v, u) = hgv, ui

which, in component form is

vi ui = gij v j ui = v j gij ui = v j gji ui (3)

where we swapped the indices because g 1 must be symmetric such that we
preserve the symmetry of inner products.

1g is to be later interpreted as the metric tensor

9
3 Definitions
Now that we’ve seen some of the mathematical background needed to grasp
what are manifolds, let’s begin our study.

3.1 Manifolds
As stated before, a manifold can be described heuristically as a generalization
of the notion of a curve (one dimensional manifold), a surface (two dimensional
manifold) and of geometric spaces in general. What makes manifolds interesting
is that we can define some notion of curvature that is intrinsic to them, without
needing to embed them in another higher dimensional space to be able to study
curved spaces.
One of the main characteristics of a manifold is that it is locally homeomor-
phic to a open subset of Rn , i.e, every point p ∈ M , M being a manifold, will
have a (sufficiently small) neighbourhood Up such that there exists a homeo-
morphism from M to Rn ,

φ : Up ⊂ M → Up0 ⊂ Rn

φ(M )
M
Up φ
φ(Up )
p

φ(p)

Figure 5: Homeomorphism from a 2-manifold to R2 .

but globally it may differ a bit from Rn . What this means is that at sufficiently
small scales, you can continuously deform a piece of the manifold and turn it
into a subset of Rn without tearing it apart or intersecting itself.
With that, we can now get our first definition:
-Definition(3.1): a manifold is a metric space (more generally, a topolog-
ical space) M such that:
i) if p ∈ M then there exists a neighbourhood (open set) U of p and some
(unique) n ∈ Z+ such that U is homeomorphic to some open subset of Rn , i.e,
there is a homeomorphism

φ : U ⊂ M → U 0 ⊂ Rn

10
for all p ∈ M . The dimension of M , denoted dim(M ), is the n ∈ Z+ to
which the subsets of M are homeomorphic, and we say that M is an n −
dimensional manif old or an n − manif old, and often denote it as M n or
just M , if the dimension is already understood previously.
-Example: a classical example of a manifold (apart from Rn itself because
that’s quite obvious given the definition) is the unit 2-sphere (or just the sphere,
the 2 is just because the term sphere is also used for higher dimensional ana-
logues) denoted S 2 , usually defined as

S 2 = {(x, y, z) ∈ R3 |x2 + y 2 + z 2 = 1},

since if we look at a small portion of the sphere it will look just like a piece of
R2 (in the same way the Earth looks flat at small scales):

φ φ(Up )

Up
p

φ(p)

Figure 6: If you look sufficiently close, a piece of the sphere looks just like R2 .

3.2 Coordinates, Charts and Atlases

As we’ve seen before, manifolds are locally homeomorphic to some open
subset of Rn , and from that we can see our next definitions:
-Definition(s)(3.2): given a manifold M and a arbitrary neighbourhood
Ui ⊂ M , a local coordinate system (or just local coordinates) is the
homeomorphism
φi : Ui → Ui0 ⊂ Rn
where since given a point p ∈ Ui , for some i, it’s counterpart φi (p) is an element
of Rn , so it will have the form

φi (p) = (x1 (p), x2 (p), ..., xn (p)),

where each of the n functions {xµ (p)} is a local coordinate and we might as
well denote p as x = (x1 (p), ..., xn (p)) unless multiple coordinate systems are at
use. It’s also important to realise that the point p exist independently of any
coordinate system, and we should only evoke any system of coordinates when
it’s necessary.

11
 A chart of M is a pair (Ui , φi ), i.e, a coordinate neighbourhood Ui together
with a coordinate function (homeomorphism) φi from Ui to Rn , and we say that
two charts are smoothly compatible, smoothly related or C ∞ − related if the
transition map

ψij = φi ◦ φ−1
j : φj (Ui ∩ Uj ) → φi (Ui ∩ Uj )

and it’s inverse

−1
ψij = ψji = φj ◦ φ−1
i : φj (Ui ∩ Uj ) → φi (Ui ∩ Uj )

are themselves smooth, or in other words, infinitely differentiable (ψij and ψji
are examples of what is to be defined as dif f eomorphisms).

M
Ui

Uj
φ−1
i φj

φi φ−1
j
Ui0
ψij

ψji
Uj0

Figure 7: Visual representation of ψij and ψji

Thus, we now define an atlas2 A of M as the family of charts {(Ui , φi )}

such that the union of all the Ui covers M , i.e,
[
Ui = M
i

and that the charts are C ∞ -related, which basically guarantees that the transi-
tion from one coordinate system to the next will be smooth, without any abrupt
or discontinuous changes.
2 As a matter of curiosity, the names map, chart, atlas, geodesic, etc reflect the origins of

differential geometry by the hands of Carl F. Gauss, who was also a cartographer

12
3.3 Differentiable Manifolds
Before we can talk about differentiable manifolds themselves, it would be
nice to talk a bit about the differentiable structure which arises in M by our
atlases.
If the union of two atlases A1 and A2 is again another atlas A3 on M , then
A1 and A2 are said to be compatible atlases of M , where this compatibility
forms an equivalence relation from which the equivalence class is called the
differentiable structure on M , and two mutually compatible atlases define
the same differentiable structure on M . We call this structure differentiable
because, as we shall see later, this is what allow us to def ine differentiation and
tangent spaces for our manifolds.
-Definition(3.3): a n-dimensional differentiable manifold or smooth
manifold is a pair (M, A) of a n-manifold M and a atlas3 A of M (more
formally, a C r differentiable n-manifold would be what we just established but
with an atlas where the charts where not C ∞ and our definition is the special
case for C ∞ charts, but we’ll be interested mostly on C ∞ manifolds since if r
is finite we won’t get much interesting structure).
-Example: once again, one of the most classical examples of a manifold
(which is also nicely a differentiable manifold) is the sphere S 2 , where the charts
are given by polar coordinates
p
−1 x2 + y 2 y
ψ1 : (x, y, z) 7→ (φ = tan , θ = tan−1 )
z x
with x, y, z ∈ (−1, 1) (because of the restriction x2 + y 2 + z 2 = 1) and φ ∈ (0, π)
and θ ∈ (0, 2π), where the inverse map ψ1−1 is given by

ψ1−1 : (φ, θ) = (x = sin φ cos θ, y = sin φ sin θ, z = cos φ)

and an auxiliary chart ψ2 which is almost identical, except now φ ∈ (− π2 , π2 )

and θ ∈ (−π, π).

U1 U2
φ=0
θ=π

π π
φ=− φ=
2 2
θ=0

φ=π

Figure 8: Charts U1 and U2 from the last example

3 Some authors (mainly mathematicians) define this to be with a maximal atlas

13
 The second chart is needed since we have some troublesome points at θ = 2π,
where it discontinuously jumps back to 0 even though the points corresponding
to that neighbourhood of θ don’t have any discontinuity, and also at the poles we
find another problem since there x AND y are equal to 0, so we can’t even solve
this through limits, hence we just use another chart, and our (differentiable)
manifold is then (S 2 , P), with P = {(U1 , ψ1 ), (U2 , ψ2 )} and the Ui are our
sphere minus the trouble points in the usual orientation (U1 ) and the same but
rotated 180o along the θ ”axis” and 90o along the φ ”axis” (U2 ).

14
4 References
1. Geometry, Topology and Physics, Second Edition, by Mikio Nakahara
2. A Comprehensive Introduction to DIFFERENTIAL GEOMETRY, Vol-
ume One, Third Edition, by Michael Spivak
3. INTRODUCTION TO DIFFERENTIAL GEOMETRY, by Joel W. Rob-
bin and Dietmar A. Salamon

4. https://en.wikipedia.org/wiki/Topological_space#Definition_via_
open_sets
5. https://en.wikipedia.org/wiki/Homeomorphism#Definition

6. http://www.macs.hw.ac.uk/~jim/F13YE2/analysis_notes.pdf

15