You are on page 1of 16

Differential Forms and Exterior Derivatives

Luigi T. Sousa
May 2022

1
Contents
1 Tying up some loose ends 3
1.1 Lie Derivative of One-Forms . . . . . . . . . . . . . . . . . . . . . 3
1.2 Lie Derivative of Scalar Fields . . . . . . . . . . . . . . . . . . . . 3
1.3 Lie Derivatives of Tensor Fields . . . . . . . . . . . . . . . . . . . 4

2 The strange world of Differential Forms 5


2.1 Differential Forms . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Wedge or Exterior Product . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 A quick sneak peak (say that fast) into Exterior Algebras 7
2.3 The Space of r-forms . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4 Exterior Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 References 16

2
1 Tying up some loose ends
In our last discussion we had some loose ends about the Lie Derivative, so
let’s tie everything.
The Lie Derivative isn’t exclusive to vectors, we can also consider it’s action
on one-forms or even more general tensors, although always along a vector field.
Let’s first see how it acts on one-forms.

1.1 Lie Derivative of One-Forms


-Definition(1.1): let X ∈ X(M ) be a vector field in M with flow given by
d µ
σ (s, x) = X µ (σ(s, x)) ,
ds
and ω ∈ Ω1 (M ) be a one-form in M , then we define the Lie Derivative of ω
along the flow of X by
1
LX ω := lim (σε )∗ ω|σε (x) − ω|x =

ε7→0 ε

1 ∗ ∗ d
(σs )∗ ω|σs (x) .

= lim (σε ) ω|σε (x) − (σ0 ) ω|σ0 (x) =
ε7→0 ε ds s=0
Note that here, instead of the pullback (σ−s )∗ used for a vector field, for one-
forms we used the pullback of covectors (σs )∗ ((σ−s )∗ brings the vector from
Tσε (x) M to Tx M and (σs )∗ brings covectors back since it acts in the opposite
direction). Given a local chart, we can repeat a similar analysis as we did with
vectors:
(σε )∗ ω|σε (x) = ωµ (x)dxµ + ε X ν (x)∂ν ωµ (x) + ∂µ X ν (x)ων (x) dxµ ,
 

which when compared to the analysis for a vector has a sign difference due to the
sign of the pushforward in definition (1.1). After putting this on said definition,
we get
LX ω = (X ν ∂ν ωµ + ∂µ X ν ων )dxµ
As we can see, LX ω ∈ Tx∗ M , since it’s defined as the difference of two one-
forms living in the same vector space at x.
Another interesting case of the Lie Derivative is that of a scalar field f ∈
F(M ):

1.2 Lie Derivative of Scalar Fields


-Definition(1.2): again, let X ∈ X(M ) and f ∈ F(M ), the Lie Derivative
of f along the flow of X is
1 
LX f := lim f (σε (x)) − f (x) =
ε7→0 ε

1 d ∂
f (σs (x)) = 1 X µ (x) µ f = X[f ] ,

= lim f (σε (x)) − f (σ0 (x)) =
ε7→0 ε ds ∂x
s=0
1 here d
we expanded ds
in the coordinate basis using the chain rule

3
which is the usual directional derivative of f along X.

1.3 Lie Derivatives of Tensor Fields


-Definition(1.3): let T ∈ T qr (M ), and let we construct the map that brings
q r
T |σε (x) to T |x as ⊗ (σ−ε )∗ ⊗ (σε )∗ as seen before (although not explicitly like
this). Then, we define the Lie Derivative of T along the vector field X ∈ X(M )
as
1 q r
LX T := lim ⊗ (σ−ε )∗ ⊗ (σε )∗ T |σε (x) − T |x =

(1)
ε7→0 ε

1 q r q r
lim ⊗ (σ−ε )∗ ⊗ (σε )∗ T |σε (x) − ⊗ (σ0 )∗ ⊗ (σ0 )∗ T |σ0 (x) =

ε7→0 ε

d q r
= ⊗ (σ−s )∗ ⊗ (σs )∗ T |σs (x) .
ds
s=0

From eq(1) it’s possible (although not necessarily easy and is beside the
focus of this discussion) to see two important properties that will be helpful
when actually calculating the Lie Derivative:

i) additivity (let T1 and T2 be two tensor fields of the same type):

LX (T1 + T2 ) = LX T1 + LX T2 ;

ii) Leibniz Rule for tensor products (let T1 and T2 be tensor fields of any type):

LX (T1 ⊗ T2 ) = (LX T1 ) ⊗ T2 + T1 ⊗ (LX T2 ) .

4
2 The strange world of Differential Forms
Up until now in our discussions we’ve been calling covectors “one-forms”,
but what in the world is a “form”? That’s what we’re going to answer now, but
first, let’s consider the symmetry operation on a tensor ω ∈ T 0r,p (M ):

P ω(V1 , ..., Vr ) ≡ ω(VP (1) , ..., VP (r) ) ,

where naturally Vi ∈ Tp M and P is an element of Sr , which is the symmetric


group of order r, meaning we just permute all entries of ω. Given a local chart,
the components of P ω are given by

P ω(eµ1 , ..., eµr ) = ω(eµP (1) , ..., eµP (r) ) = ωµP (1) ,...,µP (r) .

For more general (q, r)-type tensors we do the same procedure for the contravari-
ant and covariant indices separately.
Next, we have the symmetrizer S defined as (let ω ∈ T 0r,p (M ))

1 X
Sω = Pω ,
r!
P ∈Sr

while the anti-symmetrizer (or alternator) A is


1 X
Aω = sgn(P )P ω
r!
P ∈Sr

where the sgn(P ) is +1 for even permutations and -1 for odd permutations. It
follows from this constructions that Sω is totally symmetric (i.e P Sω = Sω)
and Aω is totally anti − symmetric (i.e P Aω = sgn(P )Aω).
-Example: take a generic (0, 2) tensor ωµν dxµ ⊗ dxν . It’s components sym-
metric part is
1
Sωµν = (ωµν + ωνµ ) =: ω(µν) , (2)
2
where the indices within () is commonly used to denote the symmetric part of
a tensor, while it’s components anti-symmetric part is
1
Aωµν = (ωµν − ωνµ ) =: ω[µν] , (3)
2
where again, the indices within [] denote the anti-symmetric part of the tensor.
Note that any (0, 2) tensor can be decomposed into it’s symmetric and anti-
symmetric parts:
ωµν = ω(µν) + ω[µν]
ωνµ = ω(µν) − ω[µν]
by the above construction.

5
2.1 Differential Forms
-Definition(2.1): a differential form of order r or simply an r-form is
a totally anti-symmetric tensor field of type (0, r).
-Example: take M = R2 , and suppose f ∈ F(R2 ). Then, a 2-form in R2
would look like  
∼ 0 −f
f dx ⊗ dy − f dy ⊗ dx = .
f 0
Note that one could factorize the f in the above expression and get

f dx ⊗ dy − f dy ⊗ dx = f (dx ⊗ dy − dy ⊗ dx) ,

motivating us to define the wedge or exterior product:

2.2 Wedge or Exterior Product


-Definition(2.2): the wedge or exterior product of r 1-forms is defined
as a totally anti-symmetric tensor product:
X
dxµ1 ∧ dxµ2 ∧ ... ∧ dxµr := sgn(P )dxµP (1) ⊗ dxµP (2) ⊗ ... ⊗ dµP (r)
P ∈Sr

-Example 1:
dxµ ∧ dxν = dxµ ⊗ dxν − dxν ⊗ dxµ ,
which, in the particular case where r = 2, we have the basis element in the
example for a 2-form and we can simplify the expression:

f (dx ⊗ dy − dy ⊗ dx) ≡ f dx ∧ dy .

-Example 2:

dxα ∧ dxµ ∧ dxν = dxα ⊗ dxµ ⊗ dxν + dxν ⊗ dxα ⊗ dxµ +

+dxµ ⊗ dxν ⊗ dxα − dxα ⊗ dxν ⊗ dxµ −


−dxν ⊗ dxµ ⊗ dxα − dxµ ⊗ dxα ⊗ dxν
and so on for higher order products.
There are 3 important properties that the exterior product satisfies which
are:
i) dxµ1 ∧ ... ∧ dxµr = 0 if some index µ appears at least twice:

dxµi ∧ dxµi = dxµi ⊗ dxµi − dxµi ⊗ dxµi ≡ 0 ;

ii) dxµ1 ∧ ... ∧ dxµr = sgn(P )dxµP (1) ∧ ... ∧ dxµP (r)

dxµi ∧ dxµj = dxµi ⊗ dxµj − dxµj ⊗ dxµi

−(dxµj ⊗ dxµi − dxµi ⊗ dxµj ) = −dxµj ∧ dxµi

6
iii) dxµ1 ∧ ... ∧ dxµr is linear in each dxµi , which follows from the fact that it’s
a linear combination of tensor products.
The exterior product also has applications in other areas such as Geometric
(or Clifford) Algebra, where it is used on vectors to construct oriented k-planes
and is related, together with the inner product, with the geometric product of
two vectors, the central operation in Geometric Algebra.

2.2.1 A quick sneak peak (say that fast) into Exterior Algebras
There is a modern, more rigorous way of defining the exterior product in the
context of exterior algebras:
-Definition(2.2.1): let V be a vector space over a field K and T (V ) denote
the tensor algebra of V . Then, if I is the two-sided ideal generated by all
elements of the form x⊗x, for x ∈ V , then the exterior or Grassman algebra
of V is given by ^
(V ) := T (V )/I ,
whose product is the exterior product seen before. This definition already
implies that x ∧ x = 0 which was the first property, and from this we can also
derive the second one:

0 = (x + y) ∧ (x + y) = x ∧ x + x ∧ y + y ∧ x + y ∧ y ,

where again, x ∧ x = y ∧ y = 0 and thus we get

x ∧ y + y ∧ x = 0 ⇐⇒ x ∧ y = −y ∧ x ,

and the bilinearity is again inherited by the tensor product in definition (2.2.1).

2.3 The Space of r-forms


If we denote the vector space of r-forms at p ∈ M as Ωrp (M ), then the set of
r-forms {dxµ1 ∧ ... ∧ dxµr } constitutes a basis for Ωrp (M ) and we can expand an
element ω ∈ Ωrp (M ) as

1
ω= ωµ ...µ dxµ1 ∧ ... ∧ dxµr ,
r! 1 r
where the components ωµ1 ...µr are taken totally anti-symmetric to reflect to
anti-symmetry of the basis. The r! factor is there since after summing, many
terms will repeat in the sum, so we normalize by r!. For example, take M = R4
and let ω ∈ Ω2p (R4 ), then if we didn’t normalize we would have

ω = ω12 dx1 ∧ dx2 + ω13 dx1 ∧ dx3 + ω14 dx1 ∧ dx4 +

+ω21 dx2 ∧ dx1 + ω23 dx2 ∧ dx3 + ω24 dx2 ∧ dx4 +


+ω31 dx3 ∧ dx1 + ω32 dx3 ∧ dx2 + ω34 dx3 ∧ dx4 +

7
+ω41 dx4 ∧ dx1 + ω42 dx4 ∧ dx2 + ω43 dx4 ∧ dx3 =
= (ω23 − ω32 )dx2 ∧ dx3 + (ω31 − ω13 )dx3 ∧ dx1 + (ω12 − ω21 )dx1 ∧ dx2 +
+(ω14 − ω41 )dx1 ∧ dx4 + (ω24 − ω42 )dx2 ∧ dx4 + (ω34 − ω43 )dx3 ∧ dx4 ,
where we used the anti-commutativity of ∧ and factored like terms, and now
let’s recall that ωµν is also anti-symmetric (also, see the eerie similarities within
the first 3 terms and the cross or vector product? That’s not a mere coincidence
and we’ll explore more about this later):

= (ω23 − ω32 )dx2 ∧ dx3 + (ω31 − ω13 )dx3 ∧ dx1 + (ω12 − ω21 )dx1 ∧ dx2 +

+(ω14 − ω41 )dx1 ∧ dx4 + (ω24 − ω42 )dx2 ∧ dx4 + (ω34 − ω43 )dx3 ∧ dx4 =
= (ω23 + ω23 )dx2 ∧ dx3 + (ω31 + ω31 )dx3 ∧ dx1 + (ω12 + ω12 )dx1 ∧ dx2 +
+(ω14 + ω14 )dx1 ∧ dx4 + (ω24 + ω24 )dx2 ∧ dx4 + (ω34 + ω34 )dx3 ∧ dx4 =
= 2!(ω23 dx2 ∧ dx3 + ω31 dx3 ∧ dx1 + ω12 dx1 ∧ dx2 +
+ω14 dx1 ∧ dx4 + ω24 dx2 ∧ dx4 + ω34 dx3 ∧ dx4 ) ,
thus showing why we need the r! factor in the expansion of ω.
Note also that in (2) ω(µν) dxµ ∧ dxν = 0 while in (3), ω[µν] dxµ ∧ dxν =
ωµν dxµ ∧ dxν .
Clearly, since we have 1 ≤ r ≤ m, where dim(M )=m, and a coordinate can’t
happen twice in the basis of Ωrp (M ), we have that the dimension of Ωrp (M ) is
 
m m!
= .
r (m − r)!r!

For convenience purposes, we define Ω0p (M ) = F(M ). Also, note that Ω1p (M ) =
Tp∗ M . If r where to exceed m, it would just go to 0 since some coordinate would
appear twice in the basis, and we also have a nice relation thanks to the equality
   
m m
=
r m−r

of the binomial coefficient, showing that dim(Ωrp (M )) = dim(Ωm−r


p (M )), and
since both are vector spaces, then this shows there is a natural vector space
isomorphism between Ωrp (M ) and Ωpm−r (M ).
We can extend our definition for the exterior product to arbitrary q- and
r-forms as follows (let ω ∈ Ωqp (M ) and ξ ∈ Ωrp (M )):

∧ : Ωqp (M ) × Ωrp (M ) −→ Ωq+r


p (M )

(ω, ξ) 7−→ ω ∧ ξ := ΛA(ω ⊗ ξ) ,


where the coefficient Λ can be defined as (q+r)!
q!r! , but isn’t necessary and is just
for convenience, so for lack of a consensus in the literature, we’ll be adopting

8
Λ = 1 for simplicity, and A is the alternator. Again, we see that if q + r > m
then the product vanishes identically.
With this more general product we define the algebra

Ω∗p (M ) ≡ Ω0p (M ) ⊕ Ω1p (M ) ⊕ ... ⊕ Ωm


p (M ) ,

which is the space of all differential forms at p ∈ M closed under the exterior
product.
V ∗ This is just the exterior algebra of the cotangent space at p ∈ M ,
(Tp M ).

Space Basis Dimension


Ω0p (M ) = F(M ) {1} 1
Ω1p (M ) = Tp∗ M { dxµ } m
m!
Ω2p (M ) { dxµ1 ∧ dxµ2 } (m−2)!2!
. . .
. . .
. . .
Ωmp (M ) { dx1 ∧ dx2 ∧ ... ∧ dxm } 1

Table 1: All the different sub-spaces in Ω∗p (M )

The exterior product of general r-forms follows some... Interesting properties


(let ξ ∈ Ωqp (M ), η ∈ Ωrp (M ) and ω ∈ Ωsp (M ) ):
i) ξ ∧ ξ = 0, if q is odd, but not necessarily if q is even;
   
1 1
ξ∧ξ = ξµ1 ...µq dxµ1 ∧ ... ∧ dxµq ∧ ξν1 ...νq dxν1 ∧ ... ∧ dxνq =
q! q!

1
= ξµ ...µ ξν ...ν dxµ1 ∧ ... ∧ dxµq ∧ dxν1 ∧ ... ∧ dxνq =
q!2 1 q 1 q
1 2
= ξν ...ν ξµ ...µ (−1)q dxν1 ∧ ... ∧ dxνq ∧ dxµ1 ∧ ... ∧ dxµq ,
q!2 1 q 1 q
where the q 2 came from the fact that we had to move each dxµi q times through
2
the product, and there are q of them, hence (−1)q , and since q is odd, q 2 is
2
also odd, so (−1)q = −1, and thus
1 2
ξν ...ν ξµ ...µ (−1)q dxν1 ∧ ... ∧ dxνq ∧ dxµ1 ∧ ... ∧ dxµq =
q!2 1 q 1 q
1
=− ξν ...ν ξµ ...µ dxν1 ∧ ... ∧ dxνq ∧ dxµ1 ∧ ... ∧ dxµq =
q!2 1 q 1 q
   
1 1
=− ξν1 ...νq dxν1 ∧ ... ∧ dxνq ∧ ξµ1 ...µq dxµ1 ∧ ... ∧ dxµq = −ξ ∧ ξ ⇐⇒
q! q!
⇐⇒ ξ ∧ ξ = 0 .

9
2
Note that if q was even, then (−1)q = 1 and ξ ∧ ξ wouldn’t necessarily be 0, in
fact, here’s an example of the exterior product of one 2-form with itself that’s
nonzero: take ξ = dx1 ∧ dx2 + dx3 ∧ dx4 , then we have

ξ ∧ ξ = (dx1 ∧ dx2 + dx3 ∧ dx4 ) ∧ (dx1 ∧ dx2 + dx3 ∧ dx4 ) =

= dx1 ∧ dx2 ∧ dx3 ∧ dx4 + dx3 ∧ dx4 ∧ dx1 ∧ dx2 =


= 2dx1 ∧ dx2 ∧ dx3 ∧ dx4 ̸= 0 .
ii) ξ ∧ η = (−1)qr η ∧ ξ;
The demonstration for this one follows very similarly from the previous one,
except each dxµi from ξ would need to move through the product r times since
there are r dxνj from η.
iii) (ξ ∧ η) ∧ ω = ξ ∧ (η ∧ ω).
This last one follows from the associativity of the tensor product.
-Example: let M = R3 , then given two generic vectors V = V µ eµ and
U = U ν eν let’s find their exterior product:

V ∧ U = (V µ eµ ) ∧ (U ν eν ) = V µ U ν eµ ∧ eν =

= V 1 U 2 e1 ∧ e2 + V 1 U 3 e1 ∧ e3 +
+V 2 U 1 e2 ∧ e1 + V 2 U 3 e2 ∧ e3 +
+V 3 U 1 e3 ∧ e1 + V 3 U 2 e3 ∧ e2 =
= (V 2 U 3 − V 3 U 2 )e2 ∧ e3 + (V 3 U 1 − V 1 U 3 )e3 ∧ e1 + (V 1 U 2 − V 2 U 1 )e1 ∧ e2 ,
which should look familiar: if we identify e2 ∧ e3 → e1 , e3 ∧ e1 → e2 and
e1 ∧ e2 → e3 , what we get is precisely the cross or vector product of V and U ,
V × U , which is the reason some authors denote the vector product with a ∧
instead of ×. This is no coincidence, it’s due to the fact that the first exterior
power of Tp R3 also has dimension 3 and thus there is a natural isomorphism
V1
between Tp M and (Tp M ).

2.4 Exterior Derivative


-Definition(2.4): the exterior derivative d of a r-form
1
ω= ωµ ...µ dxµ1 ∧ ... ∧ dxµr
r! 1 r
is a linear map
d : Ωrp (M ) −→ Ωr+1
p (M )

ω 7−→ dω
with action on ω ∈ Ωrp (M ) defined as
 
1 ∂
dω := ωµ ...µ dxν ∧ dxµ1 ∧ ... ∧ dxµr ,
r! ∂xν 1 r

10

where the exterior product already anti-symmetrizes the coefficients ∂xν ωµ1 ...µr .
-Example: let M = R3 , then the possible r-forms are:
(i) ω(0) = f ;
(ii) ω(1) = ωx dx + ωy dy + ωz dz;
(iii) ω(2) = ωyz dy ∧ dz + ωzx dz ∧ dx + ωxy dx ∧ dy;
(iv) ω(3) = ωxyz dx ∧ dy ∧ dz.
The action of d then is
∂f ∂f ∂f
(i) dω(0) = dx + dy + dz ,
∂x ∂y ∂z
which, under the identification dx → ex , dy → ey and dz → ez is just ∇f ;
     
∂ωz ∂ωy ∂ωz ∂ωx ∂ωy ∂ωx
(ii) dω(1) = − dy∧dz+ − dz∧dx+ − dx∧dy ,
∂y ∂z ∂x ∂z ∂x ∂y
and now identifying dy ∧ dz → ex , dz ∧ dx → ey and dx ∧ dy → ez we have
∇×→ −ω;
 
∂ωyz ∂ωzx ∂ωxy
(iii) dω(2) = + + dx ∧ dy ∧ dz ,
∂x ∂y ∂z

and this time around if we identify dx∧dy∧dz → 1 we get ∇· →



ω , thus recovering
all important differential operators from vector calculus;

(iv) dω(3) = 0 ,

which is obvious since there is no coordinate left, so the exterior product goes
to 0.
The exterior derivative follow a (sort of) Leibniz rule with respect to exterior
products (let ξ ∈ Ωq (M ) and ω ∈ Ωr (M )):

d(ξ ∧ ω) = dξ ∧ ω + (−1)q ξ ∧ dω ,

which we can be seen as follows (given a local chart and expanding ξ and ω in
the coordinate basis):
   
1 1
d(ξ ∧ ω) = d ξµ1 ...µq dxµ1 ∧ ... ∧ dxµq ∧ ων1 ...νr dxν1 ∧ ... ∧ dxνr =
q! r!
 
1 µ1 µq ν1 νr
=d ξµ ...µ ων ...ν dx ∧ ... ∧ dx ∧ dx ∧ ... ∧ dx =
q!r! 1 q 1 r
1 ∂
= (ξµ ...µ ων ...ν )dxα ∧ dxµ1 ∧ ... ∧ dxµq ∧ dxν1 ∧ ... ∧ dxνq =
q!r! ∂xα 1 q 1 r
1 ∂ξµ1 ...µq
= ων1 ...νr dxα ∧ dxµ1 ∧ ... ∧ dxµq ∧ dxν1 ∧ ... ∧ dxνq +
q!r! ∂xα
1 ∂ων1 ...νr α
+ ξµ ...µ dx ∧ dxµ1 ∧ ... ∧ dxµq ∧ dxν1 ∧ ... ∧ dxνq =
q!r! 1 q ∂xα

11
   
1 ∂ξµ1 ...µq α µ1 µq 1 ν1 νr
= dx ∧ dx ∧ ... ∧ dx ∧ ω ν ...ν dx ∧ ... ∧ dx +
q! ∂xα r! 1 r
   
1 1 ∂ων1 ...νr α
+(−1)q ξµ1 ...µq dxµ1 ∧ ... ∧ dxµq ∧ dx ∧ dx ν1
∧ ... ∧ dx νr

q! r! ∂xα
≡ dξ ∧ ω + (−1)q ξ ∧ dω
-Proposition 1: let X, Y ∈ X(M ) and ω ∈ Ω1 (M ), then we have the identity

dω(X, Y ) = X[ω(Y )] − Y [ω(X)] − ω([X, Y ]) .

-Proof: given a local chart, from the R.H.S we have:

X µ ∂µ [ων Y ν ] − Y µ ∂µ [ων X ν ] − ων (X µ ∂µ Y ν − Y µ ∂µ X ν ) =

= X µ Y ν ∂µ ων + X µ ων ∂µ Y ν − X ν Y µ ∂µ ων − Y µ ων ∂µ X ν −
−ων X µ ∂µ Y ν + ων Y µ ∂µ X ν = X µ Y ν ∂µ ων − X ν Y µ ∂µ ων =
∂ων
= (X µ Y ν − X ν Y µ ) ,
∂xµ
and from the L.H.S,
∂ων µ
dx ∧ dxν (X α ∂α , Y β ∂β ) =
∂xµ
∂ων
= [dxµ ⊗ dxν − dxν ⊗ dxµ ](X α ∂α , Y β ∂β ) =
∂xµ
∂ων α β µ ν
= [X Y δα δβ − X α Y β δαν δβµ ] =
∂xµ
∂ων
= (X µ Y ν − X ν Y µ ) =⇒
∂xµ
=⇒ dω(X, Y ) = X[ω(Y )] − Y [ω(X)] − ω([X, Y ]) .

The expression above can be generalized for any r-form as in
r
X
dω(X1 , ..., Xr+1 ) = (−1)i+1 Xi ω(X1 , ..., X i , ..., Xr+1 )+
i=1
X
+ (−1)i+j ω([Xi , Xj ], X1 , ..., X i , ..., X j , ..., Xr+1 ) ,
i<j

where the entries with an over line, X i , are omitted in the sum.
-Proposition 2: the exterior derivative d is 2nd-order nilpotent, i.e,

d2 = 0 .

12
-Proof: let ω ∈ Ωr (M ) and expand ω in the local chart
1
ω= ωµ ...µ dxµ1 ∧ ... ∧ dxµr .
r! 1 r
Consider now the action of d2 on ω:
1 ∂ 2 ωµ1 ...µr β
d2 ω = dx ∧ dxα ∧ dxµ1 ∧ ... ∧ dxµr ,
r! ∂xβ ∂xα
∂2ω
which vanishes since ∂xµβ1∂x
...µr
α is symmetric, while dxβ ∧ dxα is anti-symmetric.
-Example: let’s reformulate Electrodynamics. First of all, let’s define the
electromagnetic potential one-form as A = Aµ dxµ where A0 = −ϕ, the electric


scalar potential, and Ai = A , the magnetic vector potential. Then we define the
electromagnetic or Faraday tensor as F = dA. Let’s calculate it’s components:

F = d(Aµ dxµ ) = ∂ν Aµ dxν ∧ dxµ =⇒

=⇒ Fµν = ∂µ Aν − ∂ν Aµ ,
by the anti-symmetric nature of F , we already see that Fµµ = 0, and now for
the rest, let’s start with F0i = −Fi0 :

F0i = ∂0 Ai − ∂i A0 = ∂t Ai + ∂i ϕ ≡ −Ei ,

the components of the electric field, and finally for Fij = −Fji , i ̸= j,

Fij = ∂i Aj − ∂j Ai ≡ Bk ,

for i ̸= k ̸= j, and thus we also have the magnetic field encoded in F . Summa-
rizing F in a matrix gives us
 
0 −Ex −Ey −Ez
Ex 0 Bz −By 
[Fµν ] = 
Ey −Bz
 .
0 Bx 
Ez By −Bx 0

Now that we know the components of F , we can find Maxwell’s equations as


follows:
dF = d(dA) = 0
d(Fµν dxµ ∧ dxν ) = ∂λ Fµν dxλ ∧ dxµ ∧ dxν = 0 =⇒
=⇒ ∂λ Fµν + ∂ν Fλµ + ∂µ Fνλ = 0 , (4)
let’s choose µ = 0 and ν, λ = i, j = 1, 2, 3, then eq (4) gives us

∂j F0i + ∂i Fj0 + ∂0 Fij = −∂j Ei + ∂i Ej + ∂t Bk = 0 =⇒





− ∂B
=⇒ ∂i Ej − ∂j Ei = −∂t Bk ⇐⇒ ∇ × E = − .
∂t

13
Now let λ = 1, ν = 2 and µ = 3:


∂1 F32 + ∂2 F13 + ∂3 F21 = −∂x Bx − ∂y By − ∂z Bz = 0 ⇐⇒ ∇ · B = 0 ,

and so we got the source-free Maxwell equations from F . The other two equa-
tions involving sources can be derived via variation of the Maxwell Lagrangian
given by LM = − 14 F µν Fµν , where we’ll make sense of the raised indices in F µν
at future discussions.
-Proposition: let ξ, ω ∈ Ωr (N ) and let φ : M −→ N . Then we have
(i) d(φ∗ ω) = φ∗ (dω);
(ii) φ∗ (ξ ∧ ω) = (φ∗ ξ) ∧ (φ∗ ω).
-Proof:
(i) given two local charts at p ∈ M and φ(p) ∈ N , (U, x) and (V, y), respec-
tively, then we have, from the L.H.S
 
∗ 1 µ1 µr
d φ ωµ1 ...µr dy ∧ ... ∧ dy =
r!

∂y α1 ∂y αr
 
1
d ωα1 ...αr µ1 ... µr dxµ1 ∧ ... ∧ dxµr =
r! ∂x ∂x
1 ∂ωα1 ...αr ∂y α1 ∂y αr λ
= ... dx ∧ dxµ1 ∧ ... ∧ dxµr , (5)
r! ∂xλ ∂xµ1 ∂xµr
and from the R.H.S,
 
1
φ∗ d ωµ1 ...µr dy µ1 ∧ ... ∧ dy µr =
r!
 
∗ 1 ∂ωµ1 ...µr λ µ1 µr
=φ dy ∧ dy ∧ ... ∧ dy =
r! ∂y λ
1 ∂y γ ∂ωα1 ...αr ∂y α1 ∂y αr λ
= ... dx ∧ dxµ1 ∧ ... ∧ dxµr =
r! ∂xλ ∂y γ ∂xµ1 ∂xµr
1 ∂ωα1 ...αr ∂y α1 ∂y αr λ
= ... dx ∧ dxµ1 ∧ ... ∧ dxµr ,
r! ∂xλ ∂xµ1 ∂xµr
which is precisely the same as eq(5), so the equality indeed holds.
(ii)
   
1 1
φ∗ (ξ ∧ ω) = φ∗ ξµ1 ...µr dy µ1 ∧ ... ∧ dy µr ∧ ων1 ...νr dy ν1 ∧ ... ∧ dy νr =
r! r!
 
∗ 1 µ1 µr ν1 νr
=φ ξµ ...µ ων ...ν dy ∧ ... ∧ dy ∧ dy ∧ ... ∧ dy =
r!2 1 r 1 r
1 ∂y α1 ∂y αr ∂y β1 ∂y βr µ1
= ξα 1 ...αr
ωβ 1 ...βr ... ... dx ∧...∧dxµr ∧dxν1 ∧...∧dxνr =
r!2 ∂xµ1 ∂xµr ∂xν1 ∂xνr
∂y α1 ∂y αr µ1
 
1 µr
= ξα ...α ... dx ∧ ... ∧ dx ∧
r! 1 r ∂xµ1 ∂xµr

14
∂y β1 ∂y βr ν1
 
1
∧ ωβ ...β ... dx ∧ ... ∧ dx νr
≡ (φ∗ ξ) ∧ (φ∗ ω) .
r! 1 r ∂xν1 ∂xνr

The exterior derivative d induces the sequence

0 −→i Ω0 (M ) −→d0 Ω1 (M ) −→d1 ... −→dm−1 Ωm (M ) −→dm 0

where i is the inclusion map i : 0 ,→ Ω0 (M ). This sequence is called the


de Rham complex. The fact that d2 = 0 implies that im(dr ) ⊂ ker(dr+1 ).
Elements from ker(dr ) are called closed r-forms while those from im(dr−1 ) are
named exact r-forms. In practice, closed r-forms are ones such that dω = 0
and exact r-forms are those where there exists a (r − 1)-form, ψ, such that
ω = dψ. The quotient space ker(dr )/im(dr−1 ) is called the rth de Rham co-
homology group and is the dual space to the homology group, which we will
study at later discussions.

15
3 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. An Introduction to Manifolds, Second Edition, by Loring W. Tu

4. https://en.wikipedia.org/wiki/Lie_derivative
5. https://en.wikipedia.org/wiki/Exterior_algebra

16

You might also like