Let $K$ be a field of characteristic 0, and consider $d$ generic $n\times n$ matrices $X_1,\ldots,X_d$ where $X_k = (x_{ijk})_{ij}$ and $ K[x_{ijk}]$ is the polynomial algebra in $n^2 \cdot d$ variables. Then $\textrm{GL}_n$ acts on $K[x_{ijk}]$ by simultaneous conjugation $$ (X_1,\ldots,X_d) \mapsto (gX_1g^{-1},\ldots,gX_dg^{-1}) , \ \ \ \ \ \ \ \ g \in \textrm{GL}_n $$ I am just beginning to try to understand the theory of matrix invariants $K[x_{ijk}]^{\textrm{GL}_n}$, for instance with the Drenksy's survey COMPUTING WITH MATRIX INVARIANTS and Procesi's The invariant theory of n × n matrices. In the latter we have
- The algebra $K[x_{ijk}]^{\textrm{GL}_n}$ is generated by $\textrm{tr}(X_{k_1}\cdots X_{k_m})$ for products of $X_1,\ldots,X_d$
- The algebra $K[x_{ijk}]^{O(n)}$ for $O(n) \subseteq \textrm{GL}_n$ is generated by the trace $\textrm{tr}$ evaluated on products involving $X_k$ and $X_k^t$.
- The algebra $K[x_{ijk}]^{U(n)}$ for $U(n) \subseteq \textrm{GL}_n$ is generated by the trace $\textrm{tr}$ evaluated on products involving $X_k$ and $X_k^*$.
and many more results involving minimal generators and relations.
My question is, what is the analogous scenario for $S_n \subseteq \textrm{GL}_n$ where $S_n$ is the finite subgroup of permutation matrices? Perhaps the analysis is much easier than the other cases, but I do not have a good reference.
An explicit description of these invariants can be found in Male's work on permutation-invariant random matrices using an operad, but there is no reference to this problem there. There seems toThe invariants can be some connection with random matrices invariantgenerated by a subgroup ofwhat are called there the injective generalized trace $\textrm{GL}_n$, and$\tau^0$ defined on edge-labeled graphs.
What's interesting is that the evaluationequivariant maps on permutation-invariant (in expectation and in the limitrandom) matrices form an operad, so there is a nice view of itssuch maps and invariants. In particular, it would be interesting to know when these invariants are described by an operadA broader question is whether equivariant maps/invariants for other finite subgroups of $\textrm{GL}_n$ also have some kind of operadic description, or if this is unique to $S_n$.