Symmetric and alternating tensors

An action of the symmetric group on Tk(M)


For any R-module M there is a left action of the symmetric group Si on the i-fold direct product M××M, given by permuting the factors:

σ(m1,,mi)=(mσ1(1),,mσ1(i)).

(The reason for σ1 is to make this a left group action.) This balanced, multilinear map corresponds to a left group action of Si on Ti(M) which is defined on simple tensors by

σ(m1mi)=mσ1(1)mσ1(i).

For example,

(1,3,2)(m1m2m3)=m2m3m1.
Definition of symmetric and alternating tensors

Suppose R is a commutative ring and M is an R-module. An element zTi(M) is called:

  • a symmetric i-tensor if σz=z for all σSi
  • an alternating i-tensor if σz=sign(σ)z for all σSi.

For example, the elements mm and m1m2+m2m1 are symmetric 2-tensors, while the element m1m2m2m1 is an alternating 2-tensor. The 2-tensor m1m2 is neither symmetric nor alternating.

The collection of symmetric i-tensors forms a submodule of Ti(M), as does the collection of alternating i-tensors. These submodules are denoted ... nothing. At least as far as I can tell, Dummit & Foote doesn't use any notation for them.

One can prove there is also an induced action on the quotient modules Si(M) and i(M). Moreover, we have:

These actions seem identical to that of Si on the submodules of Ti(M) consisting of the symmetric and alternating tensors, respectively. Let's investigate this a bit further.

Symmetrization and skew-symmetrization


For any i-tensor zTi(M), define

Sym(z)=σSiσzAlt(z)=σSisign(σ)σz.

It is straightforward to verify these i-tensors are symmetric and alternating, respectively. We call them the symmetrization and skew-symmetrization of z.

One can verify that we have actually defined R-module morphisms Sym:Ti(M)Ti(M) and Alt:Ti(M)Ti(M) whose images lie in the submodule of symmetric and alternating tensors, respectively.

Note that if z is a symmetric i-tensor, then

Sym(z)=i!z.

Similarly, if z is an alternating i-tensor, then

Alt(z)=i!z.

From these observations, it is not too difficult to prove the following:

Proposition

Suppose R is a commutative ring and M is an R-module. For each nonnegative integer i, if i! is a unit in R then there is an R-module isomorphism between Si(M) (respectively, i(M)) and the submodule of Ti(M) consisting of all symmetric i-tensors (respectively, alternating i-tensors).

As an upshot, so long as i! is a unit in R (e.g., when R is a field), we can write Sym(Ti(M)) and Alt(Ti(M)) for the submodules of symmetric and alternating k-tensors, respectively.

Suggested next note


Noetherian modules