Let $N$ be a type ${\rm II}$ factor, with trace $\tau$. Consider its fundamental group$$ \mathcal{F}(N)= \{ \tau(p)/\tau(q) \ | \ p,q \text{ non-zero finite projections in } N \text{ and } pNp \simeq qNq \}. $$

Let $\alpha$ be a free ergodic measure preserving action of a countable ICC group $\Gamma$ on a $\sigma$-finite standard Borel measure space $(X,\mu)$. Then $L^{\infty}(X,\mu) \rtimes_{\alpha} \Gamma$ and $L(\Gamma)$ are type ${\rm II}$ factors.

Question: Is $\mathcal{F}(L^{\infty}(X,\mu) \rtimes_{\alpha} \Gamma)$ a subgroup of $\mathcal{F}(L(\Gamma))$?

I did not find a counterexample in the following reference: On the fundamental group of ${\rm II}_1$ factors and equivalence relations arising from group actions, by Sorin Popa and Stefaan Vaes.

Application: A positive answer would solve the free group factor isomorphism problem.
Proof: The group measure space construction $\mathcal{M} = L^{\infty}(\mathbb{S}^{1}, Leb) \rtimes_{\alpha} \mathbb{F}_{2}$ in this post is a ${\rm III}_1$ factor, so that its core $\widetilde{\mathcal{M}} = \mathcal{M} \rtimes_{\sigma} \mathbb{R} = L^{\infty}(\mathbb{S}^1 \times \mathbb{R}_{+}^*, Leb) \rtimes_{\widetilde{\alpha}} \mathbb{F}_2$ (see this answer) is a ${\rm II}_{\infty}$ factor of fundamental group $\mathbb{R}_{+}^*$ (moreover $\widetilde{\alpha}$ is free and ergodic). But by assumption $\mathcal{F}(\widetilde{\mathcal{M}})$ would be a subgroup of $\mathcal{F}(L(\mathbb{F}_{2}))$, so that $\mathcal{F}(L(\mathbb{F}_{2})) = \mathbb{R}_{+}^*$ also, implying that for all $n \ge 2$, $L(\mathbb{F}_{n}) \simeq L(\mathbb{F}_{2})$, by the works of Voiculescu and Radulescu.

Bonus question: Can every subgroup of $\mathcal{F}(L(\Gamma))$ be realized as $\mathcal{F}(L^{\infty}(X,\mu) \rtimes_{\alpha} \Gamma)$?

For this bonus question, I expect at most a counter-example (because a proof could be very hard).

Naive approach for a positive answer to the main question:

Let $N$ be $L^{\infty}(X,\mu) \rtimes_{\alpha} \Gamma$, as specified above. First of all, $L(\Gamma)$ can be taken as a subfactor of $N$. Take $t \in \mathcal{F}(N)$, then there are projections $p,q \in L(\Gamma) \subset N$ such that $\tau(p)/\tau(q) = t$. Then, by definition of the fundamental group, $pNp$ is isomorphic to $qNq$ (because the isom. class of such compression depends only on the trace of the projection). Let $\Phi: pNp \to qNq$ be an isomorphism. Then $\Phi(pL(\Gamma)p)$ = $qKq$, for some $K \subset N$.

Can we choose $\Phi$ such that we can take $K = L(\Gamma)$?

If so, $t$ is in $\mathcal{F}(L(\Gamma))$, and the result follows.


I asked Stefaan Vaes by email, below is his answer (reproduced with his authorization):

For instance, take $G$ to be the semidirect product of $\mathbb{Z}^2$ and ${\rm SL}(2,\mathbb{Z})$. The group von Neumann algebra $L(G)$ has trivial fundamental group and this was even the very first ${\rm II}_1$ factor that was known to have trivial fundamental group (paper of Popa in Annals of Math). Now fix a prime number $p$ and take the essentially free, ergodic, pmp action of $G$ on $\mathbb{Z}_p^2$ (where $\mathbb{Z}_p$ are the $p$-adic integers). The action of $G$ on $X=\mathbb{Z}_p^2$ is given as follows: $\mathbb{Z}^2$ acts by translation and ${\rm SL}(2,\mathbb{Z})$ acts in the "obvious" way.

Then consider the subset $Y$ of $X$ given by $Y = (p \mathbb{Z}_p)^2$. The restriction of the orbit equivalence relation to $Y$ is precisely the orbit equivalence relation of the semidirect product of $(p \mathbb{Z})^2$ and ${\rm SL}(2,\mathbb{Z})$. So, multiplication by $p$ will make sure that $p^2$ belongs to the fundamental group of the orbit equivalence relation, and thus to the fundamental group of the crossed product ${\rm II}_1$ factor.

Conclusion: the crossed product has a nontrivial fundamental group, but $L(G)$ has trivial fundamental group.

Now, the topology of $\mathbb{Z}_p$ is that of a Cantor set. Then:

Can we still hope that the question has a positive answer if we restrict to a more "regular" case like a continuous action on a topological manifold (with or without boundary), or even (if necessary) a smooth action on a smooth manifold (with or without boundary)?

Here is an other answer of Stefaan:

When it comes to von Neumann algebras or ergodic equivalence relations, there is no way to "see" such topological or geometric properties of group actions. So I am sure that there are also counterexamples that have natural, nice topological or geometric models. It is just a bit harder to find such counterexamples.


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.