2
$\begingroup$

Let $\mathbf G$ be a connected, reductive group over a local field $F$. Assume there is a maximal torus $\mathbf T$ which is split over $F$. Let $\mathbf P$ be a parabolic subgroup of $\mathbf G$ containing a Borel subgroup $\mathbf B \supseteq \mathbf T$, and let $\mathbf P = \mathbf M \mathbf N$ be a Levi decomposition. Let $G = \mathbf G(F), P = \mathbf P(F)$ etc.

Here's a nice way to calculate the Haar measure on $P$. Let $\mu$ be a Haar measure on $N$ which, being nilpotent, is unimodular. Since $P$ is the semidirect product of $M$ and $N$, it suffices to compute how the measure on $N$ changes under conjugation by $M$. That is, there exists a homomorphism $\delta: M \rightarrow (0,\infty)$ such that $\mu(E) = \delta(m) \mu(mEm^{-1})$ for all $E \subseteq N$ Borel and all $m \in M$.

As topological spaces,

$$N = \prod\limits_{\alpha \in \Phi(T,N)} U_{\alpha}$$

where $U_{\alpha}$ is isomorphic to $F$ via an isomorphism $x_{\alpha}: F \rightarrow U_{\alpha}$ such that $$x_{\alpha}(\alpha(t)a) = t x_{\alpha}(a)t^{-1}$$ for all $t \in T$, $a \in F$. Choose an absolute value $| \cdot |$ on $F$ and a Haar measure $\mu_F$ on $F$ such that $\mu_F(aC) = |a| \mu_F(C)$ for all $C \subseteq F$ measurable and $a \in F^{\ast}$.

Now $\mathbf M$ is generated by its radical $\mathbf A$ and its derived group. Since $\delta$ is trivial on the derived group, to calculate $\delta(m)$, it suffices to calculate $\delta(a)$ for $a \in A$. Choose various$E_{\alpha} \subseteq F$ with nonzero finite measure, and let $$E = \prod\limits_{\alpha \in \Phi(T,N)} x_{\alpha}(E_{\alpha})$$

Using the fact that the Haar measure on $N$ is the product of the Haar measures on $U_{\alpha}$, we have

$$\mu(aEa^{-1}) = \prod\limits_{\alpha} \mu_F( \alpha(a)E_{\alpha})= \prod\limits_{\alpha}|\alpha(a)| \mu_F(E_{\alpha})= |\rho(a)| \mu(E)$$

where $\rho = \sum\limits_{\alpha \in \Phi(T,N)} \alpha$.

This principle definitely has a generalization to nonsplit groups, but I have no idea how to get started with it.

$\endgroup$
5
  • $\begingroup$ Can't you just run the above argument in some finite extension $F' / F$ splitting $G$, obtain a (say, left) Haar measure for $P(F')$, and give $P(F)$ the induced measure as a subgroup? $\endgroup$
    – user94041
    Commented Feb 14, 2017 at 0:22
  • $\begingroup$ I'm not sure. Is $P(F)$ open in $P(F')$? In general, if you take a Haar measure on a group $G$, and you restrict the measure to the Borel sets of a closed subgroup $H$, you don't get a Haar measure on $H$. $\endgroup$
    – D_S
    Commented Feb 14, 2017 at 0:28
  • $\begingroup$ Uhm I see. No, $P(F)$ is not open in $P(F')$. $\endgroup$
    – user94041
    Commented Feb 14, 2017 at 4:59
  • $\begingroup$ In order to compute the Haar measure on $P$, these computations are not necessary. If $dm$ is a Haar measure on $M$ and $dn$ one on $N$, you get a Haar integral on $P$ by $\int_M\int_N f(mn)\,dn\,dm$. You only need an extra factor in the reverse order $nm$. $\endgroup$
    – user1688
    Commented Feb 14, 2017 at 7:43
  • $\begingroup$ What you really do compute here is the modular function. This computation carries over to any parabolic $P=LN$, since it has a finite index subgroup of the form $MAN$, where $A$ is a split torus. $\endgroup$
    – user1688
    Commented Feb 14, 2017 at 8:47

0

You must log in to answer this question.