I am not an expert at all in the subject of Lie groups, lattices, arithmetic groups and rigidity. But, lately I am interested in Margulis superrigidity theorem, which in most versions can be stated as follows:

Theorem. Let $G$ and $G'$ be semisimple connected real center-free Lie groups without compact factors with $\mathrm{rk}(G)\geq 2$, $\Gamma < G$ be an irreducible lattice, and $\pi: \Gamma \to G'$ a homomorphism with $\pi(\Gamma)$ being Zariski dense in $G'$. Then $\pi$ extends to a rational epimorphism $\pi':G\to G'$.

Here "$H$ is without compact factors" means, for $H$ center-free, that if we write $H=\prod_{i=1}^kS_i$ with $S_i$ simple, then each $S_i$ is non-compact, or equivalently has positive (real) rank; the (real) rank $\mathrm{rk}(H)$ of $H$ is the sum of the ranks of all $S_i$. Irreducibility of a lattice $\Gamma$ means that its projection in $H/S_i$ has a dense image for all $i$.


Do you have examples of Lie groups and lattices for which Margulis theorem applies, and also groups for which the theorem does not holds. In particular, do this theorem apply for $G=G'=\mathrm{PSL}(n,\mathbb{R})$ and $\Gamma=\mathrm{PSL}(n,\mathbb{Z})$. Does this implies that $\mathrm{Out}(\mathrm{PSL}(n,\mathbb{Z}))$ is finite?

Thank you all.

  • $\begingroup$ If $G$ semisimple is written as $S_1\dots S_n$ where $S_i$ are its simple normal subgroups, "no compact (simple) factor" means that none of the $S_i$ is compact. Probably MathStackExchange seems more appropriate. $\endgroup$ – YCor Dec 8 '17 at 13:10
  • 2
    $\begingroup$ For $n\ge 3$, $Out(PSL_n(\mathbf{Z}))$ is finite. More generally $Out(\Gamma)$ is finite for every irreducible. lattice $\Gamma\subset G$, $G$ semisimple not locally isomorphic to $SL_2(\mathbf{R})$. This follows from Mostow rigidity (which is anterior to the stronger Margulis' superrigidity): every automorphism of $\Gamma$ extends to an automorphism $G$. Since $Out(G)$ is finite, a finite index subgroup of $Aut(\Gamma)$ consists of the normalizer $N$ of $\Gamma$ in $G$. Since $N$ contains $\Gamma$ with finite index, $Out(\Gamma)$ finite follows. $\endgroup$ – YCor Dec 8 '17 at 16:10
  • 1
    $\begingroup$ (For $n\ge 2$ $Out(PSL_2(\mathbf{Z}))$ is also finite, as mentioned by Igor, but for reasons essentially unrelated to (super)rigidity: but this does not hold for its torsion-free finite index subgroups) $\endgroup$ – YCor Dec 8 '17 at 16:12
  • $\begingroup$ @YCor: Why does $N$ contain $\Gamma$ as a finite index subgroup? $\endgroup$ – user113290 Dec 8 '17 at 16:26
  • 1
    $\begingroup$ PS: there are several questions in your question; some but not all rather belong on MathSE. The question why (super)rigidity implies finiteness of Out is reasonable here. Possibly you could reedit accordingly. $\endgroup$ – YCor Dec 8 '17 at 16:33

Actually, the way it is stated, the "Theorem" is false. Take $G=PSL_3(\mathbb R)$, $G'=SL_3(\mathbb R)$, and $\Gamma ' \subset G'$ a torsion free lattice. Then the map $G'\rightarrow G$ sends $\Gamma '$ isomorphically to a lattice $\Gamma \subset G$. But the map $\Gamma \simeq \Gamma '$ does not extend to a rational epimorphism $PSL_3(\mathbb R) =G \rightarrow G'=SL_3(\mathbb R)$. It is not just that $G,G'$ are center freee but are of adjoint type (over $\mathbb C$, they should be centre free).

(the modified question has been answered by many people in the answers and comments).


Compact factors:

$$SO(3) \times SO(3).$$

No compact factors $SL(3, \mathbb{R}).$

$SL(3, \mathbb{R})$ - theorem applies. $SL(2, \mathbb{R})$ -- does not (rank 1). $Out(PSL(2, \mathbb{Z})$ is finite, however - it consists of the Dyer automorphism (which is an involution) and the identity, see my answer to this question.

  • $\begingroup$ Than you for your answer. I was tinking on finiteness of $PSL(n,\mathbb{Z})$ for all $n\geq 2$, I already edited my question. Do you know if all that outer automorphisms groups are finite? do youknow any good reference? $\endgroup$ – user113290 Dec 8 '17 at 15:58
  • $\begingroup$ It is known that $SL(n,\mathbb{R})$ is diffeomorphic to $SO(n)\times \mathbb{R}^m$ with $m=n(n-1)/2$. That was actually the reason I got confused in the first place, I mean, it seems like $SL(n,\mathbb{R})$ does have a compact factor $\endgroup$ – user113290 Dec 8 '17 at 16:33
  • $\begingroup$ Yes "compact factor" is somewhat ambiguous. It should be "compact simple factor in this case", but this remains ambiguous. I explained the intended meaning in my first comment to your question. $\endgroup$ – YCor Dec 8 '17 at 17:29
  • $\begingroup$ I see, so, the "word" factor should be understood as group theoretic factors, not as direct factors. I guess the decomposition of $SL(n,\mathbb{R})=SO(n)\times \mathbb{R}$ is not a contradiction siince $\mathbb{R}$ is not considered simple as a Lie group. $\endgroup$ – user113290 Dec 8 '17 at 18:01
  • 1
    $\begingroup$ It's not the point. First, $SL_n(\mathbf{R})=SO(n)\times \mathbf{R}$ is senseless since they are not even homeomorphic (different dimension). It's true that $SL_n(\mathbf{R})$ is homeomorphic to $SO(n)\times \mathbf{R}^{n(n+1)2-1}$. But it's not a group isomorphism. $\endgroup$ – YCor Dec 8 '17 at 22:04

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.