3 - Tangent - Bundles - and - Vector - Fields
3 - Tangent - Bundles - and - Vector - Fields
3 - Tangent - Bundles - and - Vector - Fields
Luigi T. Sousa
July 2020
1
Contents
1 A brief talk about Tangent Bundles 3
1.1 Tangent Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Cotangent Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Higher order (co)tangent bundles . . . . . . . . . . . . . . . . . . 6
1.4 Tensor Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3 References 11
2
1 A brief talk about Tangent Bundles
Vector fields are mathematically defined in terms of spaces called tangent
bundles, which we are going to briefly define and talk about here, but are to be
explored more deeply at a later discussion.
[
= {p} × Tp M =
p∈M
[
= {(p, v)|v ∈ Tp M } = {(p, v)|p ∈ M, v ∈ Tp M }. (1)
p∈M
It’s important to realise that even though equation (1) looks like a Carte-
sian Product of M and Tp M , it actually might have a quite different structure
globally, and this is a defining characteristic of what will later be defined as
f ibre bundles, which is the fact that they are spaces that locally look like a
direct product of spaces but globally may have a different structure to them.
Another important thing to note here is that although in a rigorous mathemat-
ical way the tangent bundle is defined as a disjoint union, it’s kind of redundant
to do it so and S it could just as well be defined as a regular union of the tangent
spaces, T M = p∈M Tp M since by construction all tangent spaces are already
disjoint, i.e, they don’t “overlap” at any point,
M = S1
p
vp
vq
q
3
which in the picture can be seen by the fact that even though the tangent lines to
the circle cross at some point, the actual space is made from the tangent vectors
at any given point, and the vectors themselves are quite different1 even if the
lines they span actually intersect, and this intersection only happens because of
the embedding of S 1 into R2 .
Since an element (p, v) ∈ T M is fully characterized by the point p ∈ M
because v ∈ Tp M is constructed from p, we give rise to a sort of natural map
called the projection and denoted by π (no, not 3.1415...), defined as follows:
π : T M −→ M
(p, v) 7→ p
which maps any element from T M to a point p ∈ M , but we lose all the
information about the vector v ∈ Tp M by doing so. Note that since π −1 (p) only
fixes p, the vector v is free to go through the entirety of Tp M and we might as
well identify π −1 (p) = Tp M and we say that Tp M is the fibre of T M at p.
-Example: let M = S 1 then we have
n d d o
T S1 = θ, v | θ ∈ S1, v ∈ Tθ S 1 ,
dθ dθ
and since Tθ S 1 is canonically isomorphic to R we get
n d d o
θ, v | θ ∈ S1, v ∈ Tθ S 1 ∼= {(θ, v) | θ ∈ S 1 , v ∈ R} = S 1 × R,
dθ dθ
which is basically the cylinder with infinite height.
S1
θ
Tθ S 1
1 The vectors start at different points, point in different directions and might even have
4
Tθ S 1 = π −1 (θ)
θ= π(θ, v)
and we have basically the same notion of projection and fibre, i.e, π(p, ω) = p
and π −1 (p) = Tp∗ M , but since we’re now dealing with the cotangent space rather
then the tangent space, it might be convenient to distinguish the projections
and fibres, so let’s just label this ones with a *:
π(p, ω) = p −→ π ∗ (p, ω) = p
5
1.3 Higher order (co)tangent bundles
Since the tangent and cotangent bundles generate a differentiable manifold
from another one, it seems plausible to re-apply this operator again and again,
generating higher order bundles. For example, given a manifold M , one could
build the tangent bundle of the tangent bundle to M
T (T M ) = T 2 M = {(p0 , v 0 ) | p0 ∈ T M, v 0 ∈ Tp0 T M },
T ∗ (T M ) = T ∗ T M = {(p0 , ω 0 ) | p0 ∈ T M, ω 0 ∈ Tp∗ T M }
T (T ∗ M ) = T T ∗ M = {(p∗ , v ∗ ) | p∗ ∈ T ∗ M, v ∗ ∈ Tp T ∗ M }.
Thus we can recursively write a formula for a k-th order tangent bundle as
T k M = T (T k−1 M )
T ∗k M = T ∗ (T ∗k−1 M ).
T ∗ T M = {(p, v, ω ∗ ) | p ∈ M, v ∈ Tp M, ω ∗ ∈ T(p,v)
∗
TM}
2 some authors may use the notation T m (M ) and reserve the T m (M ) notation specifically
n n
for smooth tensor bundles, but since we’re really only interested in C ∞ stuff, we might as
well not distinguish it at all
6
which since an element of T m n (M ) will be of the form (p, T ), with p ∈ M and
T ∈ Tm n,p (M ), will in most cases be isomorphic to the (tensor) product of m
copies of T M and n copies of T ∗ M :
Tm ∼ m n ∗
n (M ) = ⊗ T M ⊗ ⊗ T M.
T 10 (M ) = T M,
T 01 (M ) = T ∗ M.
Another example that will appear later is the (0, 2), T 02 (M ), bundle since it’s
the range for the tensor field known as the metric tensor, a central object in
Riemannian geometry and General Relativity.
7
2 A nice Field trip
Now that we’ve had a taste of what are tangent bundles, there’s only one
more thing we need to define before discussing about vector fields:
-Definition(2): a section (or cross section) of a general tensor bundle of
M is a (smooth) mapping
s:M →Tm n (M )
such that the composition π ◦ s = idM is the identity map in M , i.e, the “do
nothing” function.
π(p, X|p ) = p ⇒
⇒ (π ◦ X)(p) = p ⇒ π ◦ X = idM ,
which should look familiar since this is precisely the definition of a section for a
(1, 0) bundle, a.k.a, the tangent bundle of M . Hopefully the next example will
help illustrate the reasoning behind the name “section”.
8
-Example: let M = S 1 , then if we let our vector field X(p) have a constant
orientation (X µ > 0, for instance) and a constant magnitude (Xµ X µ = α ∈ R),
then in the same pictorial view from example (1.1) we would get something like
X S1
S1
which is literally the cross section of the cylinder. We must keep in mind,
however, that this is an embedding of a space which is isomorphic to T S 1 , not
T S 1 itself.
ω(p) : M −→ T ∗ M
p 7→ (p, ω|p ),
and the set of all covector fields is denoted by Ω1 (M ) in the context of differential
forms. In a sense, the set of covector fields is the dual of the set of vector fields,
since each element of Ω1 (M ) is a one-form field that maps, at all points in the
intersection of the domain of both fields, a vector field X ∈ X(M ) to a scalar
field F ∈ F(M ), since point-wise a one-form maps a vector to a scalar:
9
2.3 Tensor Fields
-Definition(2.3): again, let M be a C ∞ m-manifold, then a tensor field
of type (m, n) is, surprise surprise, a section of T m
n (M ),
T (p) : M −→ T m
n (M )
p 7→ (p, T |p ).
Quite confusingly (or maybe not), the notation for the set of all tensor
fields of M is exactly the same as the notation for the tensor bundle of M
itself, namely, T m 1 0 1
n (M ). As always, T 0 (M ) = X(M ), T 1 (M ) = Ω (M ) and
0 0
T 0 (M ) = F(M ), also denoted as Ω (M ) in the context of differential forms.
Also, as stated before, T 02 (M ) will be the range for a important symmetric,
non-degenerate bi-linear form, i.e the metric tensor.
In the same way that a regular (q, r)-tensor is a object that maps q covectors
and r vectors to a scalar and a covector field maps a vector field to a function,
a (m, n)-tensor field will be a multi-linear mapping of m covector f ields and n
vector f ields to a scalar f ield, but can, just as a regular point-defined tensors,
map covectors fields into vector fields and similar stuff.
10
3 References
1. INTRODUCTION TO DIFFERENTIAL GEOMETRY, by Joel W. Rob-
bin and Dietmar A. Salamon
2. Geometry, Topology and Physics, Second Edition, by Mikio Nakahara
3. An Introduction to Manifolds, Second Edition, by Loring W. Tu
4. https://en.wikipedia.org/wiki/Tangent_bundle
5. https://unapologetic.wordpress.com/2011/07/06/tensor-bundles/
11