7
$\begingroup$

Pardon if this is well known. Suppose I have a (say complex) connected reductive group $G^{\vee}$ with the $\tilde{\Delta}=\Delta\cup\{\alpha_0\}$ being the simple roots plus the negative highest root of $G^{\vee}$. Any proper subset of $\tilde{\Delta}$ defines a connected reductive subgroup $H^{\vee}\subset G^{\vee}$. Let $H$ and $G$ be the corresponding dual groups. For any special nilpotent orbit $u_H$ in $H$ we can define the Lusztig-Spaltenstein induction $u$ as a nilpotent orbit in $G$.

(that is, $(u_H,triv)$ corresponds to some representation of $W_H$ by Springer correspondence, we do the $j$-induction on this to get a representation of $W$, which correspond to $(u,triv)$ for some $u$)

Question: Is it true that the codimension of $u$ in the nilpotent cone of $G$ is the same is the codimension of $u_H$ in the nilpotent cone of $H$?

Thanks!

Edited: Just to say what I understand / guess so far: For any $u_H$, say the Springer fiber $\mathcal{B}_{e_H}$ at $u_H$ has dimension $e_H$. Then the corresponding Weyl group representation appears as the component of $H^{2e_H}(\mathcal{B}_{e_H})$ on which $A_{e_H}$ acts trivially. This component receives a surjection from $H^{2e_H}(H/B_H)$, which is a quotient of $\text{Sym}^{2e_H}(std_H)$, where $std_H$ is the standard representation of $W_H$. The Lusztig-Spaltenstein $j$-induction then, I think (I cannot find a reference for why this works for $H^{2e_H}(H/B_H)$ instead of $\text{Sym}^{2e_H}(std_H)$), gives a $W$-representation in $H^{2e_H}(G/B)$, which then corresponds to some orbit in $G$ whose Springer fiber has dimension $e_H$.

$\endgroup$
6
  • 1
    $\begingroup$ I don't think you want to call the highest root $\rho$ (except in $SL_3$) because that's standardly used for half the sum of the positive roots. $\endgroup$ Oct 7, 2015 at 1:50
  • 2
    $\begingroup$ Theorem 1.3(b) of Lusztig-Spaltenstein (Induced unipotent class) says the springer fiber of $u$ in $G$ has the same dimension as that of $u_H$ in $H$ if $H$ is a Levi of $G$. Is it going to help? $\endgroup$
    – wky
    Oct 7, 2015 at 2:04
  • $\begingroup$ Definitely! I think their construction showed that the assertion is true when $H$ is a Levi (e.g. via the theorem you pointed out). $\endgroup$ Oct 7, 2015 at 3:25
  • 1
    $\begingroup$ I learned from Prof. Lusztig that this is (essentially) true, and really follows from that in the definition of $j$-induction, the minimum number $e$ so that the interested rep'n of the Weyl group appears in $\mathrm{Sym}^e(std)$ is the same as the dimension for the Springer fiber. I'll try to add reference later when I read them better. $\endgroup$ Oct 9, 2015 at 17:03
  • $\begingroup$ @Tsai: As Jay points out (and Lusztig suggests), there are relevant details in Lusztig's papers. The extra groups $H$ are called "pseudo-Levi subgroups" by his former student Eric Sommers, though Eric and I disagree about whether that label should include the genuine Levi subgroups of parabolic subgroups. $\endgroup$ Oct 18, 2015 at 17:08

1 Answer 1

4
$\begingroup$

What you're asking for can be phrased entirely in terms of nilpotent orbits. If $\mathcal{O}$ is the nilpotent orbit of $H$ then the nilpotent orbit of $G$ you obtain via your process is the induced nilpotent orbit $\mathrm{Ind}_H^G(\mathcal{O})$. It is well known that the codimensions of these orbits coincide. See Proposition 7.1.4 of the book "Nilpotent Orbits in Semisimple Lie Algebras" by Collingwood and McGovern.

EDIT: Sorry, I just saw that you want this for a more general class of subgroups. The statement is still true. See 13.3 in Lusztig's book "Characters of reductive Groups over a finite field" and also Lusztig's paper "Unipotent classes and special Weyl group representations", J. Algebra (321), no. 11, 3418-3449, where more details are provided concerning some of the statements in the book.

$\endgroup$
1

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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