14
$\begingroup$

Let $\Gamma$ be a discrete subgroup of a connected finite dimensional Lie group $G$. Let $K$ be a maximal compact subgroup of $G$ and denote $X=G/K$. It is well-known that $\Gamma$ acts properly on $X$ via multiplication on the left.

Now suppose that a discrete group $\tilde{\Gamma}$ contains $\Gamma$ as a finite index subgroup (I'm not assuming that $\tilde{\Gamma}$ is a subgroup of $G$.)

When can the action of $\Gamma$ on $X$ be extended to a proper action of $\tilde{\Gamma}$ on $X$? Can you give an example where this is not the case?

Edit: Let my emphasize that I only require the action of $\tilde{\Gamma}$ on $X$ to be proper (i.e. any continuous action with discrete orbits and finite point stablizers). I am not asking whether $\tilde{\Gamma}$ is a subgroup of $G$. The action of $\tilde{\Gamma}$ on $X$ is allowed to have a finite kernel.

Thanks.

$\endgroup$
3
  • 1
    $\begingroup$ Do you want $\tilde\Gamma$ act effectively? Say, if $\tilde\Gamma$ is the product of $\Gamma$ and a finite group, then $\tilde\Gamma$ acts properly via the projection onto $\Gamma$. On the other hand, if you want the action to be effective, there is an easy counterexample: take $X=\mathbb R$, $\Gamma=\mathbb Z$, and $\tilde\Gamma=\mathbb Z\times\mathbb Z_2$. The factors commute and $\mathbb Z_2$ has a fixed point from which one can show that $\mathbb Z_2$ must act trivially. $\endgroup$ Mar 2, 2013 at 0:14
  • $\begingroup$ @Igor: what's the point in adding extra unnatural requirements just to give trivial counterexamples? $\endgroup$
    – YCor
    Mar 2, 2013 at 4:21
  • $\begingroup$ Igor does have a point. One instance of the original version of OP's question, before the edit, is whether every finite group acts on $S^2$. The answer is yes, trivially. But the question becomes interesting if you restrict to faithful actions. $\endgroup$
    – Lee Mosher
    Mar 2, 2013 at 13:31

2 Answers 2

9
$\begingroup$

Here is a simple counter-example. Let $p : S_3 \to S_2$ be a degree 2 covering map from the closed genus 3 surface to the closed genus 2 surface, inducing an index 2 injection $p_* : \pi_1(S_3) \to \pi_1(S_2)$. Pulling back any hyperbolic structure from $S_2$ to $S_3$ via $p$ defines an embedding of Teichmuller spaces $p^* : T(S_2) \to T(S_3)$, the domain having dimension 6 and the range having dimension 12. Pick a point in $T(S_3)$ which is not in the image of this embedding nor in any translate of the image under the action of the mapping class group $MCG(S_3)$ on $T(S_3)$. This point represents a discrete subgroup $\Gamma \approx \pi_1(S_3)$ of $Isom(H^2) = PSL(2,R)$ such that, under the index 2 inclusion of $\Gamma \hookrightarrow \tilde\Gamma \approx \pi_1(S_2)$, the action of $\Gamma$ does not extend to an action of $\tilde\Gamma$.

ADDED AFTER INITIAL COMMENTS: The OP has clarified that he intends only to ask about the existence of topological actions of $\tilde\Gamma$, in which setting this is not a counter-example as pointed out by Yves.

However, in the other direction, rigidity theorems apply to give many interesting positive examples. For instance, suppose $\Gamma$ is a uniform lattice in $Isom(H^n) = SO(n,1)$, $n \ge 3$, meaning a discrete cocompact subgroup. If $\tilde\Gamma$ is a finite index supergroup of $\Gamma$, and if $\tilde\Gamma$ has no finite normal subgroups, then the given isometric action of $\Gamma$ extends to a faithful isometric action of $\tilde\Gamma$.

To prove it, since $\tilde\Gamma$ is quasi-isometric to $\Gamma$ which is quasi-isometric to $H^n$, applying the Sullivan-Tukia quasi-isometric rigidity theorem for $H^n$ it follows that there is a cocompact discrete action of $\tilde\Gamma$ on $H^n$ with finite kernel, but we have hypothesized the kernel away so this action is faithful. We now have two discrete cocompact actions of $\Gamma$: the given one, and the restriction of the one on $\tilde\Gamma$ that we got by applying Sullivan-Tukia. We can therefore apply Mostow rigidity and conjugate the $\tilde\Gamma$ action by an isometry of $H^n$, so that its restriction to $\Gamma$ agrees with the given $\Gamma$ action, and we are done.

Basically the same argument will work whenever the symmetric space satisfies Mostow rigidity and some reasonably strong form of quasi-isometric rigidity, which is true of just about every symmetric space associated to an irreducible semi-simple Lie group (except for $SL(2,R)$ of course).

And if it is OK for the action to have a finite kernel, then just drop the hypothesis that $\Gamma$ has no finite index normal subgroup, and the same argument still works.

ADDED AFTER FURTHER COMMENTS: See Misha's comment for an improvement on this argument that avoids QI-rigidity altogether, using solely Mostow rigidity.

$\endgroup$
11
  • $\begingroup$ @Lee: your argument shows that the action does not extend to an action of $\pi_1(S_2)$ by isometries (i.e. thru the ambient Lie group). But it certainly extends to an action by homeomorphisms. I understand the question as whether the action extends to an action by homeomorphisms. $\endgroup$
    – YCor
    Mar 2, 2013 at 4:28
  • $\begingroup$ I understood the OP as Lee (nice answer, BTW!), and Yves'comment made me realize that the question is ambiguous... $\endgroup$ Mar 2, 2013 at 11:47
  • $\begingroup$ Sorry for the ambiguity. I have edited the question to make it more clear. $\endgroup$
    – Dieter
    Mar 2, 2013 at 12:05
  • $\begingroup$ @unknown: So, as far as I understand, the question is still open (at the MathOF level). It's a bit weird to tick an answer as correct (esp. after less than 1 day) because if somebody finally answers your question, what can you do? do you still expect people to think about it? (of course this answer is interesting and provides relevant info!) $\endgroup$
    – YCor
    Mar 2, 2013 at 14:34
  • 2
    $\begingroup$ @Lee: In fact, you do not have to invoke qi rigidity in the lattice case: Just note that you can always assume that $\Gamma$ is normal in $\tilde\Gamma$ (by replacing $\Gamma$ with its finite index subgroup). Now, it follows from Mostow rigidity that (apart from the $PSL(2,R)$-case) $Aut(\Gamma)$ embeds isometrically in the Lie group $G$. Therefore, modulo the (irrelevant) finite kernel of the homomorphism $\tilde\Gamma\to Aut(\Gamma)$, you have the required embedding. $\endgroup$
    – Misha
    Mar 2, 2013 at 22:55
10
$\begingroup$

Take $\tilde{\Gamma}$ to be a higher genus Baumslag-Solitar group, see section 9 of M. Kapovich, B. Kleiner, Coarse Alexander duality and duality groups, Journal of Differential Geometry, Vol. 69, (2005) Number 2, p. 279-352. Group $\tilde\Gamma$ cannot act properly discontinuously on any contractible 3-manifold, but it contains a finite index subgroup $\Gamma$ which is is Gromov-hyperbolic and isomorphic to the fundamental group of a compact 3-manifold with boundary. Hence, by a corollary of Thurston's geometrization theorem, $\Gamma$ acts isometrically and discretely on hyperbolic 3-space, i.e., is a discrete subgroup of $G=PO(3,1)$.

Edit: Here is an easier example. Note first that every finite group admitting an effective topological action on $R^2$ is either cyclic or dihedral. (B. von Kérékjártó, Über die endlichen topologischen Gruppen der Kugelfläche, Nederl. Akad. Wetensch. Proc. Ser. A vol. 22 (1919) pp. 568-569: There is probably a more accessible reference for this result, but I do not have one.) Now, take, say, a simple noncyclic finite group $F$ and set $\tilde\Gamma= F * F$. Then $\tilde\Gamma$ is virtually free, so it contains a free finite index subgroup $\Gamma$, which embeds in $PSL(2, R)$. On the other hand, every topological $\tilde\Gamma$-action on the plane has to be trivial since each free factor has to act trivially.

A side note: If you are willing to replace the symmetric space $X$ with the $n$-fold product $X^n=X\times ... \times X$, then every isometric action of $\Gamma$ extends to an isometric action of $\tilde\Gamma$ on $X^n$ (where $|\tilde\Gamma:\Gamma|=n$): This is the induced representation construction. Incidentally, the following is an open problem:

Let $\Gamma$ be a discrete isometry group of hyperbolic $n$-space $H^n$ and $\Gamma \subset \tilde\Gamma$ is a finite-index extension. Is it true that $\tilde\Gamma$ admits a properly discontinuous isometric action on $H^m$ for some $m$?

$\endgroup$
5
  • $\begingroup$ Misha, where do you prove that "$\tilde\Gamma$ cannot act properly discontinuously on any contractible 3-manifold"? I only found a proof that $\tilde\Gamma$ cannot act on a coarse PD(3) space (such as the universal cover of a compact aspherical 3-manifold). Of course, for your application you just need to show that $\tilde\Gamma$ cannot act properly discontinuously on the hyperbolic $3$-space, so for the example at hand everything works. $\endgroup$ Mar 2, 2013 at 20:18
  • $\begingroup$ In fact, the reference points to "Geometry of quasi-planes". Could you give a page reference? $\endgroup$ Mar 2, 2013 at 20:36
  • $\begingroup$ Igor, pages in pdf file for some reason do not show up, but it is p. 37-38, after Lemma 3.9. The link is correct. If you prefer archive version arxiv.org/pdf/math/9911003.pdf, it is on pages 41-43. Concerning your 1st question: The result is indeed stated for coarse PD(3) spaces, which does not include contractible 3-manifolds which are not uniformly acyclic. However, if you go through the proof(s), you realize that the proof takes place in a certain metric neighborhood of the group orbit, so the uniform acyclicity assumption can be weakened. $\endgroup$
    – Misha
    Mar 2, 2013 at 21:13
  • $\begingroup$ Oh, the reference is to "Hyperbolic groups with low-dimensional boundary" not to "Geometry of quasi-planes"; sorry I got confused $\endgroup$ Mar 2, 2013 at 21:25
  • $\begingroup$ Thanks, this is a very nice answer. Together with Lee Mosher's answer it provides a full answer to my question, i.e. both positive and negative examples. I would like to accept them both, but since this is impossible I will accept Lee Mosher's answer because it was first. $\endgroup$
    – Dieter
    Mar 4, 2013 at 8:41

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.