MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

In this question, all representations are finite-dimensional representations over $\mathbb{C}$.

Fix some $n \geq 3$. Assume that $V$ is a representation of $\text{SL}(n,\mathbb{Z})$. Also, assume that $W$ is a subrepresentation of $V$ and set $V' = V/W$, so we have a short exact sequence

$$0 \longrightarrow W \longrightarrow V \longrightarrow V' \longrightarrow 0$$

of $\text{SL}(n,\mathbb{Z})$ representations. Assume that the actions of $\text{SL}(n,\mathbb{Z})$ on $W$ and $V'$ extend to actions of $\text{SL}(n,\mathbb{C})$. Question : Must the action of $\text{SL}(n,\mathbb{Z})$ on $V$ extend to an action of $\text{SL}(n,\mathbb{C})$?

Margulis superrigidity says that the action virtually extends, but I can't construct representations where it doesn't extend on the nose. Of course, if it extends then $V = V' \oplus W$ as $\text{SL}(n,\mathbb{C})$-representations.

share|cite|improve this question
What does "virtually extends" mean? If it means "can be extended to a subgroup of finite index of $SL(n,\mathbf C)$", the question becomes trivial as $SL(n,\mathbf C)$ does not have any proper subgroups of finite index as it is connected in the Zariski topology. (Note that the map $SL(n,\mathbf C) \to PSL(n,\mathbf C)$ has finite kernel, but $PSL(n,\mathbf C)$ doesn't lift to a subgroup.) – Guntram Mar 21 '11 at 7:37
"Virtually extends" means that there is a representation of $SL(n,\mathbf{C})$ that coincides with the given representation of $SL(n,\mathbb{Z})$ on a subgroup of finite index of $SL(n,\mathbb{Z}).$ – Victor Protsak Mar 21 '11 at 12:30
This is intriguing, but maybe the assumption that actions of the discrete group extend to "actions" of the Lie/algebraic group could be clarified: are the latter actions taken in the sense of an abstract group or as a Lie/algebraic group? (Borel-Tits work on abstract homomorphisms of algebraic groups might or might not be relevant.) Also, the set-up seems meaningful for more arbitrary simple Lie/algebraic groups of rank at least 2. – Jim Humphreys Mar 21 '11 at 14:10
@Jim : Extends to a representation of the Lie/algebraic group. And it is definitely interesting for general lattices in higher rank Lie groups! I just chose the first nontrivial case to make the statement accessible to the widest range of people. – Tim O Mar 21 '11 at 14:54
You probably mean $SL(n,R)$ in place of $SL(n,C)$, if you want $SL(n,Z)$ to be a lattice. – Ben Wieland Mar 21 '11 at 22:51
up vote 9 down vote accepted

Consider the surjective map of $SL(n,\Bbb Z)$-modules $Hom_{\Bbb C}(V',V)\to Hom_{\Bbb C}(V',V')$. Tim tells us that the identity map from $V'$ to $V'$ lifts to an $f:V'\to V$ which is invariant under a finite index subgroup $\Gamma $ of $SL(n,\Bbb Z)$. Then by averageing one can make it invariant under $SL(n,\Bbb Z)$.

share|cite|improve this answer
+1: But you might want to make this more explicit. I had to dig through the comments and then think for a while to decode your answer. – Daniel Litt Mar 21 '11 at 18:16
Yes, this is a case where the rule "more is more" works better than "more is less". – Jim Humphreys Mar 21 '11 at 22:06

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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