Let $K$ be a field of characteristic zero, and $V$ a finite dimensional vector space over $K$. Consider the action of the algebraic group $G:=PGL(V)$ on the vector space $W:=End_K(V^{\otimes 2})$ by conjugation of the diagonal action.

My question is: what is known about the orbits of $G$ in $W$? what is the dimensions of the stabilizer of the generic point? What about if we restrict $W$ to the linear automorphisms of $V^{\otimes 2}$ ? Is there any specific sort of points for which we get finite stabilizers for example?


Some rough ideas: -$\small{End(V\otimes W)\cong (V\otimes W)^*\otimes (V\otimes W)\cong (V^*\otimes V)\otimes (W^*\otimes W) \cong End(V)\otimes End(W)}$ and this remains true as $G$-modules. So your object is almost $T^2(\mathfrak g):=\mathfrak g \otimes \mathfrak g$ with $\mathfrak g=Lie(G)$. (in fact $(\mathbb K\oplus \mathfrak g)\otimes (\mathbb K\oplus \mathfrak g)\cong \mathbb K\oplus \mathfrak g \oplus \mathfrak g\oplus T^2(\mathfrak g)$ as $G$-modules)

-$T^2(\mathfrak g)=S^2(\mathfrak g)\oplus\Lambda^2(\mathfrak g)$ where elements of $S^2(\mathfrak g)$ are polynomial functions of degree $2$ on $\mathfrak g^*\cong\mathfrak g$, so it is well understood.

-If you want to look at invariants on $T^2(\mathfrak g)$, you will consider $S(T^2(\mathfrak g))\subset T(T^2(\mathfrak g))\subset T(\mathfrak g)$, the tensor algebra on $\mathfrak g$. I think that some people know something in type A for such invariants. Maybe are there some hints in Procesi's paper: http://arxiv.org/abs/1501.05190

-for specific elements, you can consider stabilizers of a pure tensor $x_1\otimes x_2$ with $x_1$ and $x_2$ are semisimple. Then $G^{(x_1\otimes x_2)}=G^{(x_1,x_2)}$ (the second is the simultaneous stabilizer of $x_1$ and $x_2$ in $G$). I can't ensure yet, since I have to leave, but I think that this stabilizer is generically finite, even in this very specific case.

Edit: of course, if $x_1$ an $x_2$ are regular semisimple element lying in different tori, the simultaneous centralizer is trivial.

  • $\begingroup$ We can use another work of Procesi, to describe explicitly all the invariants: They can be described in the following way: if $T:V\otimes V\rightarrow V\otimes V$, then take $Tr(\sigma T^{\otimes n}):V^{\otimes 2n}\rightarrow V^{\otimes 2n}$, where $n$ is some number and $\sigma\in S_{2n}$ acts by permuting the tensor factors. By considering the partial traces $T_1,T_2:V\rightarrow V$ I am also convinced now that the generic stabilizer is finite. However, I will be happy to have a better description of some possible quotient spaces. $\endgroup$ – Ehud Meir Feb 3 '16 at 16:35
  • 1
    $\begingroup$ It smells like a wild problem to me. Indeed, just in $(\mathbb K\otimes\mathfrak g)\oplus(\mathfrak g\otimes \mathbb K)\cong\mathfrak g \oplus\mathfrak g$, you have some closed orbits arising from nilpotent elements. E.g. in $\mathfrak{sl}_2\times\mathfrak{sl}_2$, $(e,f)$ has a closed orbit. The same phenomenon should also arise in $\mathfrak g\otimes\mathfrak g$ $\endgroup$ – Bulois Michael Feb 4 '16 at 8:23
  • $\begingroup$ what if we restrict somehow the variety? that is: if instead of taking the entire $End(V\otimes V)$ we just take some constructible subset? Are there any nice quotients which one can describe? One of the things I am interested in is the subvariety of all solutions for the Young-Baxter Equation: $(T\otimes 1) (1\otimes T)(T\otimes 1) = (1\otimes T) (T\otimes 1) (1\otimes T)$ $\endgroup$ – Ehud Meir Feb 4 '16 at 12:02
  • $\begingroup$ Sorry, no idea for your particular equation. But generally, I would say that it is indeed usually possible to describe a quotient in either of the two following cases: -a relatively small closed subvariety where nilpotents do not behave very badly (e.g. commuting variety in $\mathfrak g\times \mathfrak g$) -an open locus which keeps only the very general points $\endgroup$ – Bulois Michael Feb 4 '16 at 12:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.