5
$\begingroup$

$\DeclareMathOperator\ch{ch}$Let $F$ be a non-archimedean local field, $\mathcal{O}$ its ring of integers, $\mathfrak{p}$ its maximal ideal and $\pi$ a uniformizer. I shall use $\kappa(F)$ to denote the residue field of $F$.

Let further $G = \mathrm{GL}_n$ denote the standard algebraic group scheme of the general linear group. Further I shall denote by

  • $T$ : The standard split torus of $G$ consisting of diagonal matrices,
  • $B$ : the standard Borel subgroup of $G$ of upper-triangular matrices with non-zero elements on the diagonal,
  • $U$ : the standard unipotent subgroup of $G$ of upper-triangular matrices with $1$'s on the diagonal principal. Hence $B = TU$. I will also use $U^{-}$ for the subgroup of lower-diagonal matrices with $1$'s on its diagonal principal.

The roots and the Bruhat order are chosen in a standard way, i.e. all roots $\alpha_{ij}$ shall be positive iff $i > j$ and negative otherwise. For my needs, I realize the Weyl-Group $W = W(G(F),T(F))$ inside $G(\mathcal{O})$ by dropping the modulo $T(\mathcal{O})$-condition (this won't have any effect in my setting).

The Iwahori subgroup J of $G(F)$ is defined as the preimage of $B(\kappa(F))$ under the canonical projection $G(\mathcal{O}) \to G(\kappa(F))$. In other words $$ J = \begin{pmatrix} \mathcal{O}^{\times} & \mathcal{O} & \ldots & \mathcal{O} \\ \mathfrak{p} & \mathcal{O}^{\times} & \ddots & \vdots \\ \vdots & \ddots & \ddots & \mathcal{O} \\ \mathfrak{p} & \ldots & \mathfrak{p} & \mathcal{O}^{\times} \end{pmatrix}. $$ The Iwahori subgroup may be decomposed as the direct product $$ J = U^{-}(\mathfrak{p})T(\mathcal{O})U(\mathcal{O}). $$ The (Iwahori-)Hecke algebra $\mathcal{H}$ is the space of $\mathbb{C}$-valued functions on $G(F)$, that are

  1. Iwahori-bi-invariant, i.e. $f(jgj') = f(g)$ for $j,j' \in J$ and $g \in G(F)$.
  2. compactly supported
  3. smooth, i.e. locally constant.

It is a $\mathbb{C}$-algebra under the convolution product given by $$ (f \ast g)(x) = \int_{G(F)} f(xy^{-1})g(y)dy, $$ where the Haar measure on $G(F)$ is chosen such that $vol(J,dg)=1$.

It is known, that as complex vector spaces, $$ \mathcal{H} = \mathcal{H}_W \otimes X_{*}(T(F)), $$ where $\mathcal{H}_W$ is the Hecke-algebra of $W$ (not necessary here) and $X_{*}(T(F))$ is the (commutative) group of rational cocharacters of $F$ (these are rational homomorphisms $F^{\times} \to T(F)/T(\mathcal{O})$. The cocharacter group $X_{*}(T(F))$ is known to be a free abelian group of rank n, and since I can drop the 'modulo $T(\mathcal{O})$'-condition, I will simply write $$ \pi^{\lambda} = \begin{pmatrix} \pi^{\lambda_1} & & \\ & \ddots & \\ & & \pi^{\lambda_n} \end{pmatrix} $$ for the image of $\pi$ under some $\lambda \in X_{*}(T(F))$.

The group of cocharacters can be realized inside $\mathcal{H}$ as follows: for a dominant $\lambda\in X_{*}(T(F))$ one defines the Hecke operator $$ T_{\lambda} := \ch(J\pi^{\lambda}J), $$ where by $\ch(\cdot)$ I mean the characteristic function of $\cdot$. As far as I understand it, if $\lambda\in X_{*}(T(F))$ is any (!) cocharacter, there are dominant $\lambda', \lambda'' \in X_{*}(T(F))$ s.t. $\lambda = \lambda' - \lambda''$ and one defines $$ \widetilde{T}_{\lambda} := T_{\lambda'} T^{-1}_{\lambda''}. $$ The definition should be independent on the choice of $\lambda'$ and $\lambda''$.

Question: I was wondering if there is any nice description of these $T^{-1}_{\lambda''}$, maybe any reference?

Example: I computed it in the simplest case when $n=2$ and $\lambda = (1,0)$ is the fundamental coweight. I got something like $$ T^{-1}_{(1,0)} = \frac{1}{q} \ch(J\pi^{-(1,0)}J) - \frac{1}{q(q-1)} \ch(J w_2J), $$ where $w_2 = \begin{pmatrix} & 1\\ 1 & \end{pmatrix}$ is the (only) simple root.

$\endgroup$
18
  • $\begingroup$ In general, $T_g$ is defined to be $\textrm{ch}(JgJ)$; the expression you've written (with $T_{\lambda'}T_{\lambda''}^{-1}$) is how one defines the Bernstein basis of the Bernstein subalgebra. In any event, one way to write down $T_\lambda^{-1}$ is using the element $T_{-\lambda}^*$ and a reduced expression for $\pi^\lambda$ in the extended affine Weyl group of $G$. For a reference, see Proposition 4.13(1) in this article of Vigneras: mathscinet.ams.org/mathscinet-getitem?mr=3484112 $\endgroup$
    – krl
    Commented Jul 6, 2022 at 1:13
  • $\begingroup$ hi @krl, first of all thanks for you reply. Thanks for your hint on the definition of a (general) $T_{\lambda}$, I have now modified it and put a tilde on it, but it is not really relevant now. I have taken a look at the paper.. I don't know if we understand each other. The element $T*_{-\lambda}$ is defined using $T^{-1}_{\lambda}$, so as I see it, this would be just running in a circle. I also didn't understand what exactly you mean with "reduced expression" for $\pi^{\lambda}$? $\endgroup$ Commented Jul 10, 2022 at 18:19
  • $\begingroup$ In the paper of Vigneras, $T_{-\lambda}^*$ is defined as $T_u T_{s_r^{-1}}^*\cdots T_{s_1^{-1}}^*$, where $T_{s_i^{-1}}^* = T_{s_i^{-1}} - c_{s_i^{-1}}$ (notation of that Lemma 4.12), and where $s_1 s_2 \cdots s_r u$ is a reduced expression for $\pi^\lambda$. In particular, you can define $T_{-\lambda}^*$ without inverses. (Since you are working with with the Iwahori-Hecke algebra, some of these expressions will simplify.) Proposition 4.13(1) of the article then shows that $T_\lambda^{-1} = q_{\lambda}^{-1}T_{-\lambda}^*$. $\endgroup$
    – krl
    Commented Jul 11, 2022 at 17:28
  • $\begingroup$ As for defining a reduced expression, I would recommend reading a bit more about (extended) affine Weyl groups, and working out the example of $\textrm{GL}_n(F)$. For example, if $\lambda = (1,0)$, then a reduced expression is $\pi^\lambda = \begin{pmatrix} \pi & 0 \\ 0 & 1\end{pmatrix} = \begin{pmatrix} 0 & 1 \\ 1 & 0\end{pmatrix}\begin{pmatrix} 0 & 1 \\ \pi & 0 \end{pmatrix}$. The first element is a simple reflection ($s_1$ above), and the second element has length 0 and is contained in $\Omega$ ($u$ above). $\endgroup$
    – krl
    Commented Jul 11, 2022 at 17:32
  • $\begingroup$ I would ask you maybe some questions if I got it correct.. 1) Vigneras works with the Iwahori-Matsumoto (IM) presentation, and so, in order to invert $T_{\lambda}$, I need to make the conversion from the Bernstein-Zelevinsky (BZ) presentation to the IM-presentation, and then kind of 'unfold it back'. 2) The $c_{s_i} = q-1$ and $q_s = q$ (mod sign) in the $\mathrm{GL}_n$-case, right? (I looked it up but I still don't really know what the Kottwitz morphism is), and hence I also don't know what $\Omega$ is. I would understand it as the 'defect' from to $\mathrm{GL}_n$ to $\mathrm{SL}_n$ $\endgroup$ Commented Jul 18, 2022 at 20:19

0

You must log in to answer this question.