I read the following theorem in a paper without a proof, which I don't understand well. Let $F$ be a global function field, and $v$ be a place of $F$, use $G_r$ to denote $GL_r$.

Theorem: For any integer $r\ge 1$, we have

(1) the space of unitary irreducible admissible representations $\pi$ of $G_r(F_v)$ can be decomposed as the disjoint union of real algebraic varieties $Im[\pi_0]/Fixe(\underline{r},\pi_0)$ indexed by the pair $(\underline{r},\pi_0)$ where

$\bullet$ $\underline{r}$ denotes a partition $r=r_1+...+r_k$ of $r$,

$\bullet$ $\pi_0$ is a unitary irreducible repersentation and square integrable on $G_{r_1}(F_v)\times...\times G_{r_k}(F_v)$,

$\bullet$ $[\pi_0]$ is a complex algebraic variety on which the complex torus $(\mathbb{C}^{\times})^k$ acts simply transitively,

$\bullet$ $Im[\pi_0]$ is a real compact subvariety of $[\pi_0]$ on which the subtorus $U(1)^k$ acts simply transitively,

$\bullet$ $Fixe(\underline{r},\pi_0)$ is a finite group which acts on both $[\pi_0]$ and $Im[\pi_0]$.

This allows us to talk about polynomial functions on this space: These are the functions zero outside a finite number of components whose restriction to each component of $Im[\pi_0]/Fixe(\underline{r},\pi_0)$ are polynomials which can be extended to polynomials on the complex variety $[\pi_0]$ invariant by $Fixe(\underline{r},\pi_0)$.

(2) This space is given the Plancherel measure, such that every compactly supported locally constant function

$$h_v(g):G_r(F_v)\to \mathbb{C}$$ can be spectrally decomposed in the form

$$h_v(g)=\int h_{v,\pi}(g)d\pi, \ \ \ \ \ \ g\in G_r(F_v)$$ where

$\bullet$ each $h_{v,\pi}:G_r(F_v)\to \mathbb{C}$ is an element in the eigenspace associated to $\pi$, that is, is the linear combination of matrix coefficients of $\pi$,

$\bullet$ each function $\pi\to h_{v,\pi}(g)$ is a polynomial.

My question is

(1) What is a reference for this theorem?

(2)This theorem is stated for global function field, and is an analog still true for local fields (archimedean or nonarchimedean) of a global number field? If so, is there a reference?

Thank you very much for any explanation or reference suggestion!


1 Answer 1

  1. Every unitary $\infty$-dim'l irreducible representation can be writen as inducing a square-integrable representation from a parabolic subgroup with Levi subgroup $G' =G_{r_1} \times \dots G_{r_2}$. Googling for Bernstein center might help.

  2. I guess $[\pi_0]$ is the family, where you tensor by unramified one-dimensional representations of $G'$, so that is not a fancy action but does not preserve unitarity.

  3. The character formula for a prabolic induced rep looks for parabolic $P_r = G_r N_r$ and Iwasawa decomposition $G = P_r K_r$ like $$ tr\; \pi(\phi) = tr\; \pi(\phi^{N,K}) , \qquad \phi^{N,K}(g) := \int\limits_{N}\int\limits_{K} \phi(k^{-1}gnk) d k\; d n\;(g \in G_r).$$ This is an exercise in functional analysis.

  4. I can not address the algebraic geometry stuff, but it is probably not that difficult. I would need a definition of $Im[\pi_0]$ and the finite group though, but I have no idea what that should be. May be the finite group means permuting $G_k$'s of equal dimension, that is, the relative Weyl group?

All what I am mentioning holds for arbitrary local fields, but the polynomial statement will only be true for non-archimedean fields.

I would check Laumon - Cohomology of Drinfeld modules I and probably II. Most likely, he does the local things without specifying the characteristic of the local field, but of course does not address the archimedean situation at all.

  • $\begingroup$ I think you are correct: $[\pi_0]$ is all the tensor products with $1$-dimensional representations, $Im[\pi_0]$ is all the tensor products with unitary $1$-dimensional representations, and the finite group is the relative Weyl group. $\endgroup$
    – Will Sawin
    Mar 20, 2014 at 14:09
  • $\begingroup$ @WillSawin My issue is that unitary one-dimensional representation live on the imaginary line, not $U(1)$. $\endgroup$
    – Marc Palm
    Mar 20, 2014 at 14:12
  • $\begingroup$ Ah okay, I see imaginary line modulo $x=2 \pi i / log(q)$, $q$ being the residue characteristic, that's isomorphic to $U(1)$, I guess:) I didn't see that before:\ But that's a confusing embedding of $U(1)$ into $\mathbb{C}^\times$ modulo $x$. $\endgroup$
    – Marc Palm
    Mar 20, 2014 at 14:14

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.