Exercise [12.

08]

First, I would like to repeat the motivation/definition of the "antisymmetrization" operation in

view of the wedge product of p different 1-forms α, … γ. Basically, we have:

α ∧ ... ∧ γ = ∑ α r ... γ u dx r ∧ ... ∧ dx u with M = {1, 2, ... n} (1)

( r ,...u )∈M p

As Penrose explains in §11.6, the above sum comprises groups of p! summands each which
have the same – but permuted – indices. The wedge products dx r ∧ ... ∧ dx u in these
summands are not independent, because they differ at most in a "±" sign. It is then desirable
to give all those equivalent terms the same coefficients. This is achieved by adding all their
coefficients as they appear in eq. (1) (with appropriate sign) and divide this sum by p!. The
unique coefficient that results from this procedure is the "antisymmetrized" value α[r …γu].
Hence we have:

α ∧ ... ∧ γ = ∑ α [ r ... γ u ] dx r ∧ ... ∧ dx u (2)

( r ,...u )∈M p

1
with α[ r ...γ u ] = ∑ sign(Π) ⋅ α Π ( r ) ... γ Π (u )
p! Π
(3)

In eq. (3), the sum runs over all permutations Π of the p indices r, … u, and sign(Π) is the
sign of the permutation Π, i.e. the factor (-1)number of transpositions of Π. Note that the whole terms in
eq. (1) and (2) have the same value, but that the coefficients appearing in the sums with the
"dxr…dxu" are (usually) different.

Now to the actual task "Justify that ϕ ∧ χ = α ∧ ... ∧ γ ∧ λ ∧ ... ∧ ν where

ϕ = α ∧ ... ∧ γ and χ = λ ∧ ... ∧ ν ":

According to Penrose, the wedge product of the p-form ϕ and the q-form χ is defined with
antisymmetrized coefficients according to:

ϕ∧χ = ∑ ϕ[ r ...u χ j ...m ] dx r ∧ ... ∧ dx m

( r ,...u , j ...m )
(4)

According to the schema of eq. (2), we have ϕ r ...u = α [ r ...γ u ] and χ j ...m = λ[ j ...ν m ] . Hence we

can rewrite eq. (4) as:

ϕ∧χ = ∑ α [[ r ... γ u ]λ[ j ...ν m ]] dx r ∧ ... ∧ dx m

( r ,...u , j ...m )
(5)

On the other hand, the immediate product of all involved 1-forms is, according to the schema
of eq. (2), defined as:

Juergen Beckmann
 α ∧ ...γ ∧ λ ∧ ...ν = ∑ α[ r ... γ u λ j ...ν m ] dx r ∧ ... ∧ dx m
( r ,...u , j ...m )
(6)

The assertion " ϕ ∧ χ = α ∧ ... ∧ γ ∧ λ ∧ ... ∧ ν " is hence justified if the coefficients in
equations (5) and (6) are equal, i.e. if

α[ r ... γ u λ j ...ν m ] = α[[ r ... γ u ]λ[ j ...ν m ]] (7)

Eq. (7) is plausible from the fact that it cannot make a difference

- if the "average" of all terms α r ... γ u λ j ...ν m that result from permutations of the indices is

calculated (left side of eq.(7)),

- or if first the "average" of the terms α r ... γ u and the "average" of the terms λ j ...ν m is

calculated, and finally the "average" of these averages is calculated (right side of
eq.(7)).

More formally, one can show that for any entities Wr…uj…m we have:

1
W[[ r ...u ] j ...m ] = ∑ sign(Π ) ⋅ WΠ ([ r ...u ] j ...m )
Π: permutation in M p + q ( p + q )!

1 1
= ∑ sign(Π ) ∑ sign(Γ) ⋅ WΠoΓ (r ...uj ...m )
Π: permutation in M p + q ( p + q )! Γ: permutation in M p p!

1 1 ⎛ ⎞
=
( p + q )! p!
∑ ⎜
⎜ ∑ sign(Π ) sign(Γ) ⋅ WΠoΓ (r ...uj ...m ) ⎟
⎟
Γ: permutation in M p ⎝ Π: permutation in M p + q ⎠

1 1 ⎛ ⎞
=
( p + q)! p!
∑ ⎜
⎜ ∑ sign(Π o Γ −1 ) sign(Γ) ⋅ W( Π o Γ −1 ) o Γ (r ...uj ...m ) ⎟
⎟
Γ: permutation in M p ⎝ Π o Γ −1 : permutation in M p + q ⎠

[Note: If Π runs through all permutations in Mp+q, then Π°Γ-1, does so, too.]

1 1 ⎛ ⎞
=
( p + q)! p!
∑ ⎜
⎜ ∑ sign(Π ) sign(Γ −1 ) sign(Γ) ⋅ WΠ (r ...uj ...m ) ⎟
⎟
Γ: permutation in M p ⎝ Π o Γ −1 : permutation in M p + q ⎠
=1

= p! terms independent of Γ

1 ⎛⎜ ⎞
= ∑
( p + q)! ⎜⎝ Π: permutation in M p+q
sign(Π ) ⋅ WΠ (r ...uj ...m ) ⎟ = W[ r ...uj ...m ]
⎟
⎠

The first and the last line of the above considerations show that, in general, "inner
antisymmetrizations" can be left out (or inserted) at will. This proves eq.(7).

Juergen Beckmann