6
$\begingroup$

Let $\Gamma$ be a group with a probability measure preserving action on $(X,\mu)$, and $H$ another group. Recall that a cocycle is a map $c:\Gamma\times X\to H$ such that $c(gg',x)=c(g,g'x)c(g',x)$. Two cocycles $c,c'$ are cohomologous if there is $f:X\to H$ such that $c(g,x)=f(gx)^{-1}c'(g,x)f(x)$.

Let $H$ be a discrete group. Say that $H$ is superrigid if for every lattice $\Gamma$ in a simple, higher rank Lie group $G$, and for every ergodic probability measure preserving action of $\Gamma$ on $(X,\mu)$, every measurable cocycle $c:\Gamma\times X\to H$ is cohomologous to a cocycle with values in a finite group.

For example, if $H$ is a hyperbolic group, then a result of Adams proves that $H$ is superrigid.

My question is : Assume that $H$ contains a normal subgroup $H'$ of finite index, and that $H'$ is superrigid. Does it follow that $H$ is superrigid ?

$\endgroup$
2
  • $\begingroup$ "in a finite group": you mean "in a finite subgroup"? $\endgroup$
    – YCor
    May 23, 2018 at 12:21
  • $\begingroup$ Yes, it has values in $H$, so I mean in a finite subgroup of $H$. $\endgroup$
    – user81562
    May 23, 2018 at 12:32

1 Answer 1

5
$\begingroup$

Yes.

Instead of writing down explicitly such a cocycle (which is not hard), let me take this opportunity to explain how to think of such objects in a "cooridnate free" manner. Given a cocycle $c:\Gamma\times X\to H$ you can consider the space $Y=X\times H$ and endow it with the product measure and with the $\Gamma\times H$ measure preserving action given by the right action of $H$ on the second coordinate and the twisted left action of $\Gamma$ $$(*)\quad\gamma\cdot(x,h)=(\gamma x,c(\gamma,x)h).$$ Clearly, $Y$ is $\Gamma\times H$-ergodic iff $X$ is $\Gamma$-ergodic.

Conversely, let $Y$ be an ergodic measure preserving $\Gamma\times H$-space on which $H$ acts freely and with a finite measured fundamental domain. Then we may form the space of orbits $X=Y/H$ and endow it with a $\Gamma$-invariant finite measure, and any choice of a fundamental domain (ie a section $X\to Y$) will give us an identification $Y=X\times H$ such that the $\Gamma$-action on $Y$ will be realized as in $(*)$ for some cocycle $c$. Different choices of fundamental domains will give cohomologous realization cocycles. In this cohomology class we could find a cocycle taking values in a subgroup $F<H$ iff we can find a $\Gamma$-invariant $H$-equivariant map $Y\to H/F$. It follows, in particular, that in this cohomology class we could find a cocycle taking values in a finite group iff $Y$ admits a $\Gamma$-invariant $H$-equivariant map $Y\to Z$, where $Z$ is a discrete space on which $H$ acts with finite stabilizers. This is all standard.

We now answer the question in view of the above. We consider an ergodic measure preserving $\Gamma\times H$-space $Y$ on which $H$ acts freely and with a finite measured fundamental domain. Note that $Y$ need not $\Gamma\times H'$-ergodic, but the finite group $H/H'$ acts transitively on the collection of $\Gamma\times H'$ ergodic components. Let $Y=\cup_{i=1}^n Y_i$ be the corresponding ergodic decomposition. Note that each $Y_i$ is an ergodic measure preserving $\Gamma\times H'$-space on which $H'$ acts freely and with a finite measured fundamental domain. By the assumption on $H'$ (which I object calling "super-rigidity") we can find for each $i$ a $\Gamma$-invariant $H'$-equivariant map $\pi_i:Y_i\to Z_i$, where $Z_i$ is a discrete space on which $H'$ acts with finite stabilizers. Set $Z=\cup_{i=1}^n Z_i$ and consider the $\Gamma$-invariant $H'$-equivariant map $\pi=\cup \pi_i:Y\to Z$. Consider now the induction of $\pi$ from $H'$ to $H$, namely the map $H\times_{H'} Y \to H\times_{H'} Z$. As $Y$ is an $H$ space we may identify $H\times_{H'} Y$ with $H/H'\times Y$ in a $H\times \Gamma$-equivariant way. We obtain a $\Gamma$-invariant $H$-equivariant map $Y \to \text{Maps}(H/H',H\times_{H'} Z)$. We are done, observing that $\text{Maps}(H/H',H\times_{H'} Z)$ is a discrete space on which $H$ acts with finite stabilizers.

$\endgroup$

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.