# Decomposition into irreducibles of symmetric powers of irreps.

Suppose I have an irreducible representation of a simple Lie algebra, say $\mathfrak{sl}(n)$ or $\mathfrak{so}(n)$ i.e., $A$ and $D$ type. Given such a representation, $\Gamma_\lambda$, indexed by its highest weight $\lambda = (\lambda_1, \dots, \lambda_n)$, what is known about the decomposition of the symmetric powers of $\Gamma_\lambda$?

To be more specific, suppose I am considering a Lie algebra inside $\mathfrak{gl}({n^2})$, and I have a representation $(\rho, \Gamma_\lambda)$ of one of the above two Lie algebras of equal rank to that of $\mathfrak{gl}(n^2)$ i.e., of rank $n^2$. My interest is specifically concerned with the occurrence of the trivial representation in the $n$th symmetric power of such a representations i.e., $\mathrm{Sym}^n \circ \rho$.

As a concrete example, consider the rank 9 representation of $\mathfrak{so}_4$ with highest weight $8 \omega_1$ where $\omega_i$ denotes the $i$th fundamental weight. Upon applying $\mathrm{Sym}^3$ to $\Gamma_{8\omega_1}$, this representation decomposes as a direct sum of subrepresentations, one of which is the trivial representation (I computed this with Sage).

Is there any literature on the occurrence of the trival representation inside $\mathrm{Sym}^n$ powers of such representations as above?

• You might check out the publications of Roger Bryant (sometimes with coauthors) May 23 '13 at 22:21
• Hmmm, I am looking through his publications here: manchester.ac.uk/research/Roger.bryant/publications and he seems to treat with Lie algebras in characteristic $p$, but I should have mentioned that I am working over $\Bbb C$ here. And the symmetric powers he deals with are mostly with finite groups? Was there a particular paper you had in mind? Many thanks for your time. May 23 '13 at 23:31

First I'd recomend that you use more careful language in describing the groups and representations, including the precise notion of rank. The subject is complicated enough as is, so for instance you need to refer to the Lie algebras of special orthogonal groups as having Lie type $B_n$ or $D_n$ in the respective cases when $n$ is odd, even. (Then $n$ is the rank, while the natural representation of the group has dimension $2n+1$ or $2n$.)

Your question is actually extremely difficult to answer in detail, as people understood in the era around 1900 when invariant theory was being intensively developed. Eventually, in work of Hilbert, Weyl, and others, an emphasis was developed on ring-theoretic methods here. Rather than just concentrate on the multiplicity of the trivial module in each symmetric power of a given representation, one should look at the entire graded algebra of polynomials (or polynomial functions) and try to describe the (graded) subalgebra of invariants. The starting point is the "natural" representation of the Lie algebra or Lie group, where it's already quite challenging to make the answers explicit.

Modern sources include the Springer GTM volumes 129 by Fulton-Harris and 255 by Goodman-Wallach (a second edition of an earlier book published elsewhere). Relevant work by Howe and others is covered here, along with helpful examples and references.

• This is quite helpful as it gives me a new direction in which to read. I will look through Goodman-Wallach and continue with my study of Fulton and Harris. Thank you for your response Dr. Humphreys May 24 '13 at 2:46

For what it's worth, here's a small general observation for representations of compact groups $G$. As Humphreys mentioned the idea is to consider the entire symmetric algebra $S(V)$ of a representation rather than a particular symmetric power, then look at its invariant subalgebra. The symmetric algebra is a graded representation and accordingly it has a graded character

$$\chi_{S(V)}(g) = \sum_{n \ge 0} \chi_{S^n(V)}(g) t^n.$$

The graded character of $S(V)$ can be computed using symmetric function identities; if $\lambda_1, ... \lambda_k$ are the eigenvalues of $g$ acting on $V$ then the graded character evaluated at $g$ is

$$\sum_{n \ge 0} h_n(\lambda_1, ... \lambda_k) t^n = \frac{1}{(1 - \lambda_1 t)...(1 - \lambda_k t)} = \frac{1}{\det(1 - g t)}.$$

With the graded character of $S(V)$ in hand, you can write down the graded dimension of its invariant subalgebra by averaging over $G$:

$$\sum_{n \ge 0} \dim S^n(V)^G t^n = \int_G \frac{1}{\det(1 - gt)} d \mu$$

where $\mu$ is normalized Haar measure. So in principle the problem reduces to computing this integral (or differentiating under the integral sign with respect to $t$ some number of times, setting $t = 0$, then computing the resulting integral). The integrand is conjugation-invariant so this reduces to a simpler integral over the conjugacy classes of $G$ but I don't know what form it takes.

For $\mathfrak{s}\mathfrak{l}_n$ and $\rho$ a symmetric power you can have a look at this article by Bedratyuk published in Lin. Multlin. Algebra. It has a formula for the multiplicity of the trivial representation but it is quite scary.