3
$\begingroup$

Let $G$ be a split reductive group over a nonarchimedean local field $F$ (I'm particularly interested in the case of $\operatorname{GSp}_{2n}$).


Given a parahoric subgroup $K \subset G(F)$, and a parabolic subgroup $P$, is there a "nice" group-theoretic description of the orbits of $K$ on the flag variety $\mathcal{F}(F) = P(F) \backslash G(F)$?


I have the following conjecture: if $T$ is a split maximal torus in $G$ and $\widetilde{W} = N_G(T)(F) / T(\mathcal{O}_F)$ is the extended affine Weyl group, then $\widetilde{W}$ is generated by a set $S$ of simple reflections, and there is a bijection between standard parahorics and proper subsets of $S$. I think that if $K_J$ is a standard parahoric corresponding to some $J \subset S$, then $$ P \backslash G / K = W_{M_P} \backslash W_G / \pi(W_J) $$ where $W_J$ is the subgroup of $\widetilde{W}$ generated by $J$, and $\pi(W_J)$ is its image in the usual Weyl group. This conjecture is true if $K$ is contained in $G(\mathcal{O}_F)$. Does anyone know if it is true in general?

$\endgroup$
4
  • $\begingroup$ I think that what you want should follows from the following decomposition of $G$. Let $I$ be a standard Iwahori subgroup and $B$ be a standard Borel subgroup. Then $G=I\cdot N(T)\cdot B$ (cf; Bruhat-Tits IHES - I, Thm (5.1.3)(iv), page 83. From this you get that the RHS of your conjectural identity contains a system of representatives. $\endgroup$ Commented Jun 17, 2023 at 14:35
  • $\begingroup$ @PaulBroussous As I wrote in the question, I know the result holds for any parahoric contained in G(O), and in particular for the Iwahori. So it is clear that the RHS surjects onto the LHS. What is not so clear to me is how to prove injectivity. $\endgroup$ Commented Jun 18, 2023 at 15:09
  • $\begingroup$ Sorry, you're right, it doesn't help at all. $\endgroup$ Commented Jun 19, 2023 at 7:18
  • $\begingroup$ I wouldn't go that far, it's definitely helpful to have a reference, and the general setting of "double Tits systems" considered loc. cit. does sound as if it might be relevant for my general question. $\endgroup$ Commented Jun 19, 2023 at 9:53

1 Answer 1

3
$\begingroup$

Let $G$ a reductive group over a nonarchimedean local field $F$. Let $P_0$ be a minimal parabolic subgroup of $G$ and $A$ a maximal split torus contained in $P_0$. The normalizer $N_G(A)(F)$ acts on the apartment of $A$ via the extended affine Weyl group, which contains the affine Weyl group $W_\mathrm{aff}$ with finite index. Let $I$ be an Iwahori subgroup of $G(F)$ compatible with $P_0$.

By the Iwasawa decomposition, the double cosets in $P_0(F)\backslash G(F)/I$ are represented by the elements of the Weyl group $W := N_G(A)(F)/C_G(A)(F)$. Let $\pi$ be the natural projection $W_\mathrm{aff}\rightarrow W$. It's not hard to show that if $w\in W$ and $s\in W_\mathrm{aff}$ is a simple reflection (relative to $I$), then \begin{equation} \tag{1}\label{parabolic_mult} P_0(F) w\pi(s) I\subseteq P_0(F) w I s I \subseteq P_0(F) w\pi(s) I \sqcup P_0(F) wI, \end{equation} and if furthermore $s\in W$, then \begin{equation} \tag{2}\label{iwahori_mult} P_0(F) sw I\subseteq P_0(F) s P_0(F) w I \subseteq P_0(F) sw I\sqcup P_0(F) wI. \end{equation}

Now let $P$ be a parabolic subgroup of $G$ containing $P_0$, and let $K$ be a parahoric subgroup of $G(F)$ containing $I$. Then \begin{equation} \tag{3}\label{bruhat} P(F) = \bigsqcup_{u\in W_{M_P}}P_0(F)uP_0(F), \end{equation} where $W_{M_P}$ is the Weyl group of the standard Levi component of $P$, and \begin{equation} \tag{4}\label{iwahori} K = \bigsqcup_{v\in W_K}IvI, \end{equation} where $W_K$ is the Coxeter subgroup of $W_\mathrm{aff}$ consisting of those elements with representatives in $K$.

Given $w\in W$, consider the double coset $P(F)wK$. From (\ref{bruhat}) and (\ref{iwahori}), $$P(F)wK = \bigcup_{u\in W_{M_P},v\in W_K}P_0(F)uP_0(F)wIvI.$$ Applying (\ref{parabolic_mult}) one simple generator of $W_K$ at a time, we get that this union is equal to $$\bigcup_{u\in W_{M_P},v\in W_K}P_0(F)uP_0(F)w\pi(v)I.$$ Applying (\ref{iwahori_mult}) one simple generator of $W_{M_P}$ at a time, we find that this is equal to $$\bigcup_{u\in W_{M_P},v\in W_K}P_0(F)uw\pi(v)I.$$ It follows from the Iwasawa decomposition that $w,w'\in W$ represent the same double coset in $P(F)\backslash G(F)/K$ if and only if $w,w'$ lie in the same double coset in $W_{M_P}\backslash W/\pi(W_K)$.

$\endgroup$
2
  • $\begingroup$ Welcome to MO! (And extra characters.) $\endgroup$
    – LSpice
    Commented Jul 13, 2023 at 6:37
  • $\begingroup$ (sorry, I posted a comment but it was nonsense). This looks great, thanks! $\endgroup$ Commented Jul 13, 2023 at 8:12

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .