6
$\begingroup$

An FC-group is a group in which every element has a finite conjugacy class. A group G is said to be maximally almost periodic if there is an injective homomorphism from G into a compact Hausdorff group.

My question is whether every discrete countable FC-group G is maximally almost periodic.

In general an FC-group G is isomorphic to a subgroup of the direct product of a locally finite FC-group and a torsionfree abelian group. Since discrete abelian groups are maximally almost periodic, this reduces the general question to the case where G is additionally assumed to be locally finite.

$\endgroup$
  • 1
    $\begingroup$ Obviously $G/Z(G)$ is residually finite, hence MAP, so it works when $Z(G)$ is trivial. In general it sounds tricky. $\endgroup$ – YCor Aug 7 '14 at 9:52
  • $\begingroup$ Why is $G/Z(G)$ residually finite when $G$ is FC? $\endgroup$ – Andy Oct 8 '18 at 9:42
  • $\begingroup$ Ah ok, just figured this out. If $g\in G$ is not in the center, there is some $h\in G$ with which it does not commute. Then $C_h$ is a finite index normal subgroup avoiding $g$. $\endgroup$ – Andy Oct 8 '18 at 9:48
6
$\begingroup$

Here's a counterexample. Consider a non-abelian finite, 2-step nilpotent group $F$. Let $Z$ be its derived subgroup. Define $G$ as the quotient of the direct sum $F^{(\mathbf{N})}=\bigoplus F_n$ of copies of $F$ indexed by $\mathbf{N}$ by identification of all copies of $Z$. Thus, writing $H=F/Z$, the group $G$ lies in a central extension, with central kernel $Z$ and quotient the infinite direct sum $H^{(\mathbf{N})}$. I claim that $G$ is not maximally almost periodic, and more precisely that any homomorphism of $G$ into a compact Hausdorff group kills $Z$.

0) $G$ is an FC-group: this is because it's quotient of a direct sum of finite groups.

1) I claim that for every finite index subgroup $M$ of $G$ we have $Z\subset [M,M]$. Indeed, let $M_n$ be the image of $M$ in the $n$-th copy $H_n$ of $H$ in $H^{(\mathbf{N})}$. Then the image of $M$ in $H$ is contained in $\bigoplus M_n$. Since it has finite index, this implies that for all but finitely many $n$, we have $M_n=H_n$. Let $M'$ be the inverse image of $M$ in $F^{(\mathbf{N})}$ and $M'_n$ its projection on $F_n$. Then for all but finitely many $n$, we have $M'_nZ_n=F_n$. Since $F_n$ is nilpotent and $Z_n=[F_n,F_n]$, this implies that for those $n$, we have $M'_n=F_n$, and hence for those $n$, we have $Z_n\subset [M'_n,M'_n]$. By fixing such an $n$ and projecting into $G$, we deduce that $[M,M]$ contains $Z$.

2) To conclude, it's enough to show that every homomorphism $f$ from $G$ to a compact Lie group $L$, with dense image, kills $Z$. Fix $f$. Since $G$ is nilpotent, $L^0$ being a compact nilpotent connected Lie group, it is abelian. Define $M=f^{-1}(L^0)$. Then $M$ has finite index, hence $Z\subset [M,M]$. Since $f$ maps $M$ into the abelian group $L^0$, we deduce that $[M,M]\subset ker(f)$, hence $Z\subset ker(f)$.


Besides, as I mentioned, the result is true for FC-groups $G$ with trivial center: indeed these groups act faithfully on themselves by conjugation, and since the orbits are finite, the closure of the image of $G\to Sym(G)$ is compact. This is more generally true for residually finite FC-groups. However since as you mention, it is also true for abelian groups (which may fail to be residually finite), it's unclear what it the greatest generality.

$\endgroup$
  • $\begingroup$ By the way, there is a nice substitute property for maximal almost periodicity which does hold for all countable FC-groups: namely every countable FC-group has a free ergodic probability measure preserving action with generalized discrete spectrum. (The MAP groups are precisely the groups which admit a free ergodic p.m.p. action with discrete spectrum.) $\endgroup$ – Robin Tucker-Drob Aug 9 '14 at 16:07

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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