9
$\begingroup$

Has the Euler characteristic with compact support of $\mathrm{SL}_n(\mathbb R)/\mathrm{SL}_n(\mathbb Z)$ been computed ? References? Thanks.

$\endgroup$
7
  • 1
    $\begingroup$ For $n$ even, this will be odd dimensional, and hence have euler characteristic 0. In general, this will be a bundle over the symmetric space $X$ of symmetric positive definite matrices of determinant 1 (unimodular positive definite quadratic forms or metrics on $R^n$), with fiber $SO(n)$. Hence if the base or fiber has odd dimension, then the Euler characteristic will be 0. For the other dimensions, the volume has been computed and should be proportional to the euler characteristic by applying Chern-Gauss Bonnet to the base. arxiv.org/abs/math/0402085 $\endgroup$
    – Ian Agol
    Commented Jul 21, 2020 at 22:55
  • 1
    $\begingroup$ Many thanks Ian. My question is about $\chi_c$ (compact support). So odd dimensional (non-compact) it is non zero. However your answer and reference are useful. My more general interest is actually finding a stratification of this space into a finite number of -interesting-strata, possibly open cells. In case you also have a thought on this, thank you in advance for sharing it. $\endgroup$ Commented Jul 21, 2020 at 23:12
  • 1
    $\begingroup$ Ah, I see. Somehow I was thinking you wanted the euler characteristic of the compactification (which I think was shown to exist by Borel-Serre). $\endgroup$
    – Ian Agol
    Commented Jul 21, 2020 at 23:28
  • $\begingroup$ Do you know the value for small $n$? at least $n=2$? $\endgroup$
    – YCor
    Commented Jul 22, 2020 at 6:30
  • 1
    $\begingroup$ Hi, Sadok. By Poincaré duality, this would be the same as the ordinary Euler characteristic, computed from homology, right? $\endgroup$ Commented Jul 22, 2020 at 22:03

1 Answer 1

2
$\begingroup$

I think that the euler characteristic is 0 for the following reasons.

Firstly, the space $SL_N(\mathbb{R})$ is a bundle over the symmetric space $SO(N,\mathbb{R})\backslash SL_N(\mathbb{R}) = SP(n,\mathbb{R})=X$, the space of symmetric positive-definite real matrices of determinant 1. For a discussion of this symmetric space, see e.g. Bridson-Haefliger II.10. Then $SL_N(\mathbb{R})/SL_N(\mathbb{Z})$ is a bundle over $X/SL_N(\mathbb{Z})$ with fiber $SO(N,\mathbb{R})$. Note that this is an orbifold bundle, but that by passing to a torsion-free subgroup, one can assume that it is a manifold (and since you're interested in euler characteristic, this just multiplies by the index).

Now the space $X/SL_N(\mathbb{Z})$ admits a bordification by Borel-Serre. Hence $SL_N(\mathbb{R})/SL_N(\mathbb{Z})$ has a bordification by an $SO(N,\mathbb{R})$-bundle over the Borel-Serre bordification. Hence it is the interior of a manifold with boundary $M$. In this case, $H^*_c(SL_N(\mathbb{R})/SL_N(\mathbb{Z}))\cong H^*(M,\partial M)$. Then by Lefschetz duality, $\chi(H^*_c(M,\partial M))=\chi(M)$.

But since $M$ is a bundle with fiber $SO(N,\mathbb{R})$, and $\chi(SO(N,\mathbb{R}))=0$ (any Lie group has a nowhere vanishing vector field), we have $\chi(M)=\chi(SO(N,\mathbb{R}))\times \chi(X/SL_N(\mathbb{Z})) =0$, since the euler characteristic of bundles is the product of the euler characteristic of the base and the fiber.

$\endgroup$
2
  • $\begingroup$ This seems right. I think we can avoid the compactification by restricting the orbi-bundle over strata (locally closed). Over some strata of the base, the fiber is SO(n) and over other strata it is a quotient of SO(n) by a finite group I believe. In all cases, the fiber is compact with zero Euler characteristic. We then get a stratification of $SL_n(\mathbb R)/SL_n(\mathbb Z))$ by bundles whose $\chi_c=0$ since $\chi_c$ is multiplicative on bundles. Since $\chi_c$ is additive over strata, this gives the answer $0$ as well. $\endgroup$ Commented Jul 24, 2020 at 9:32
  • $\begingroup$ Well, you need to know also that the cohomology is finite so that the Euler characteristic is well-defined. This follows from the compactification, or from a finite spine. But as I said, thinking about strata can be avoided by passing to a manifold cover (a level 3 principal congruence subgroup should suffice). $\endgroup$
    – Ian Agol
    Commented Jul 24, 2020 at 14:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .