Let $G$ be a connected, simply connected, semi-simple algebraic group defined and split over a local non-arch field $k$ with integer ring $R$. Let $B$ be the corresponding reduced building. Fix an apartment $A$ and corresponding root system $\Phi$ and split torus $T$; then for each $x \in A$ and $r>0$, we have a corresponding Moy-Prasad filtration subgroup $G_{x,r}$. Our hypotheses give $G_x=G_{x,0}$.

Define $\Omega_{A,r}$ as the set $$ \{ y \in A : \forall \alpha \in \Phi, \vert \alpha(x-y) \vert \leq r \}.$$

Is it true that $T(R)G_{x,r} = \cap_{y\in \Omega_{A,r}} G_y$ ?

(With $r>0$ the product gives a group which is contained in the given intersection. Equality seems to be about whether all hyperplanes $H_{\alpha,r}$ meet $\Omega_{A,r}$.)

Let $A(x)$ be the set of all apartments of $B$ containing $x$. Let $Z$ be the (finite) center of $G$.

Is it true that $ZG_{x,r} = \cap_{A \in A(x)}\cap_{y\in \Omega_{A,r}}G_y$ ?
Equiv by (1): Is $ZG_{x,r} =\cap_{g \in G_x} (gT(R)g^{-1})G_{x,r}$?

This is part of the larger question: I would like to understand the stabilizers of certain subsets of $B$ (which are not contained within any apartment $A$). Pointers to any literature much appreciated.


Regretfully the answer to the first question is no.

For example consider a group of type $G_2$. Normalize your roots so the short roots have length 1, for example, and let $x=0$. Write $\Phi = \Phi^l \cup \Phi^s$ as a union of long and short roots, respectively, and define the corresponding sets $\Omega^l_{A,r}$ and $\Omega^s_{A,r}$. Write $B_r$ for the disk of radius $r$.

Then $\Omega^l_{A,r} \subseteq B_{2r/3}$ whereas $B_r \subseteq \Omega^s_{A,r}$. Consequently the parenthetical comment fails.

There was also a typo in the parenthetical comment, so let's be explicit. (Given a root $\alpha$, write $U_\alpha$ for the corresponding root subgroup and normalize so $U_\alpha(R)= U_\alpha \cap G_0$.)

To see that $T(R)G_{x,r} \subseteq \cap_{y\in \Omega_{A,r}} G_y$, it suffices to verify this on a set of generators. We have $T(R) \subseteq G_y$ for all $y\in A$. Then the rest of $G_{x,r}$ is generated by the groups $U_\alpha(P^{\lceil r-\alpha(x)\rceil})$.
For each $y \in \Omega_{A,r}$ we have $\alpha(x)-\alpha(y)\leq r$ so $-\alpha(y) \leq r-\alpha(x)$ which implies $U_\alpha(P^{\lceil r-\alpha(x)\rceil}) \subseteq U_\alpha(P^{\lceil -\alpha(y)\rceil})$. Thus $T(R)G_{x,r}$ is contained in each of these $G_y$.

Conversely, in the case of $G_2$ described above, the inequalities imply that for sufficiently large $r$, there is no $y \in \Omega_{A,r}$ for which the equality $\lceil -\alpha(y) \rceil = \lceil r-\alpha(x) \rceil$ ($=\lceil r \rceil$) holds; therefore the intersection will be strictly larger than $T(r)G_{x,r}$.

It follows that the answer to the first version of the second question is also no. The second version is true for $SL(2)$, for example, so perhaps there is an argument via Chevalley generators and relations.


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.