5
$\begingroup$

We get a "nice" Haar measure on $G=SL(2,R)$ in Iwasawa coordinates $G=NAK$ as follows: $dg=dx {dy\over y^2} dk$. Here $N=\{ n_x\}$, $A=\{a_y\}$ and $K=SO(2)$. Note that $dg=dn\, da\, dk$ is a product of Haar measures on the three subgroups.

Question: In what generality does this happen? If I have some nice Lie group $G$, and I have, say, a local bijection $H\times K\to G$ (in a neighborhood of $e$), is $dg=dh\, dk$? (No in general; e.g. $\bar N\times A\times N\to G=SL(2,R)$. Haar measure is a mess involving all three variables.)

Thanks!

$\endgroup$
3
  • $\begingroup$ I'm not sure what you mean by writing $\mathrm dg = \mathrm dx(\mathrm dy/y^2)\mathrm dk$ and $\mathrm dg = \mathrm dn\,\mathrm da\,\mathrm dk$; $\mathrm da$ is $\mathrm dy/y$, not $\mathrm dy/y^2$ (assuming $a = \begin{pmatrix} y \\ & y^{-1} \end{pmatrix}$). $\endgroup$
    – LSpice
    Apr 12, 2021 at 20:40
  • $\begingroup$ But, to the general question, I suspect that, absent the obvious case where conjugation by $H$ acts by measure-preserving isomorphisms on $K$, or vice versa, there is no very good general answer for when this happens. $\endgroup$
    – LSpice
    Apr 12, 2021 at 20:42
  • $\begingroup$ Without thinking at all deeply about this, my suspicion is that in the first example one is really getting a product of Haar measures on P and K where P=NA; and then it so happens that the parabolic P has a nice semidirect product decomposition. I also suspect that for this particular example the compactness of K is important, but this is just a tentative hunch and I could easily be shown wrong here $\endgroup$
    – Yemon Choi
    Apr 12, 2021 at 23:23

1 Answer 1

2
$\begingroup$

FYI, the answer to my question is this Stack Exchange post (h/t Stephen D Miller):

https://math.stackexchange.com/a/2325533

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.