Symmetric algebras

Motivation


The tensor algebra construction created, from each R-module M, a "minimal" R-algebra T(M). In other words, beginning from the additive operation in M and the R-scaling on M, it created a structure that also had an internal multiplication (compatible with those structures). Our universal property formally encoded the "minimality" of the construction, in that all R-module morphisms from M to R-algebras A "lifted" to R-algebra morphisms from T(M). There were no additional properties imposed on T(M) other than those required to have an R-algebra. In particular, the R-algebra T(M) was not guaranteed to be commutative (and rarely ever is).

Question

Can we modify our construction so that the R-algebra we obtain is also commutative?

Yes. Let's do just that.

A desired universal property


As with the tensor algebra functor, there is a functor from the category of R-modules to the category of commutative R-algebras. It is analogous to any free construction and the tensor algebra construction, in that it is left adjoint to a forgetful functor:

A universal property of the symmetric algebra

Let U:R-CAlg→R-Mod be the forgetful functor from the category of commutative R-algebras to the category of R-modules. Then there is a functor S:R-Mod→R-CAlg together with a natural bijection

τM,A:HomR-CAlg(S(M),A)→∼HomR-Mod(M,U(A)).

In other words, the functor S is a left adjoint of the forgetful functor U.

As with any object satisfying a universal property, we can now deduce many properties of S(M):

The construction


We already have a construction that takes an R-module M and creates an R-algebra T(M) with most of the properties we want. To obtain a commutative R-algebra, it's reasonable to consider a quotient of T(M) that forces a commutativity relation in the quotient ring.

Definition of symmetric algebra

Suppose R is a commutative ring (with unity) and M is an R-module. Let C(M)⊆T(M) be the ideal generated by elements of the form m1⊗m2−m2⊗m1 for m1,m2∈M. The symmetric algebra of M is quotient

S(M)=T(M)/C(M).

Some notes are in order. First, the ideal C(M) is generated by homogeneous elements of degree 2, which implies C(M) is a graded ideal (with degree i the submodule denoted Ci(M)) and the quotient ring S(M) is a graded ring. The homogeneous component of degree i is

Si(M)=Ti(M)/Ci(M).

This R-module is called the ith symmetric power of M. One can show that the submodule Ci(M) is generated by all elements of the form

(m1⊗m2⊗⋯⊗mi)−(mσ(1)⊗mσ(2)⊗⋯⊗mσ(i)),

where mj∈M and σ∈Si.

Note that since C(M) is generated (as an ideal) by degree 2 homogenous elements, we have C0(M)=C1(M)=0 and hence also have S0(M)=R and S1(M)=M.

Optional notation

It is common to drop the tensor symbol between elements when working in S(M); e.g., to write simply m1m2 rather than m1⊗m2+C(M) for the image in S(M) of the element m1⊗m2∈T(M).

Example


Let V be an n-dimensional vector space over a field k. Then S(V) is isomorphic as a graded k-algebra to k[x1,…,xn]. If B={v1,…,vn} is a basis for V as an k-vector space, then a basis for Sm(V) is

{v1a1v2a2⋯vnan∣ai≥0,a1+a2+⋯+an=m}.

In particular, the dimension of the F-vector space Sm(V) is (n−1+mn−1).

Suggested next notes


Exterior algebras
Symmetric and alternating tensors