Is there a finitely presented sofic group which is not residually finite, but all of its finitely generated subgroups are Hopf groups?

It seems like the Baumslag Solitar groups $BS(m,n)$ don't work (i.e. for $|m|=1$ or $|n|=1$ or $|m|=|n|$ they are residually finite, and otherwise they contain a non-Hopf finitely generated subgroup).

Note: Thanks to YCor for pointing out that in the initial formulation I said "subgroups" instead of "finitely generated subgroups". The latter is my intention.

  • 1
    $\begingroup$ What's a non-Hopfian subgroup in $BS(2,4)$? recall that for $|m|,|n|>1$, $BS(m,n)$ is Hopfian iff $m,n$ have the same prime divisors, but is residually finite only when $|m|=|n|$. In particular $BS(2,4)$ is Hopfian and non-RF, but I'm not sure whether all its subgroups are Hopfian. $\endgroup$
    – YCor
    Oct 1, 2016 at 14:51
  • 3
    $\begingroup$ Also you should clarify if you really mean "all its subgroups are Hopf groups". Because whenever a group admits a free subgroup, then it admits a free subgroup of infinite rank, which is clearly not Hopfian. Maybe you mean, "all finitely generated subgroups are Hopfian"? $\endgroup$
    – YCor
    Oct 1, 2016 at 14:57
  • $\begingroup$ @YCor: Yes. I meant finitely generated subgroups. Thanks for the correction! As for non-Hopfian subgroup of $BS(2,4)$, the subgroup $\langle a,b^{-1}ab,b^2\rangle$ works by Meskin's "Nonresidually finite one-relator groups" / TAMS 1972, Vol. 164. (see Lemma 2.2 and the reference within). $\endgroup$ Oct 1, 2016 at 15:58

1 Answer 1


Houghton's group $H_3$ (see Section 5.3 here for a definition) is (locally finite)-by-$\mathbf{Z}^2$, which easily implies that all its finitely generated subgroups are Hopfian. (Not all its subgroups are Hopfian: it admits an isomorphic copy of $F^{(\mathbf{N})}$ as a subgroup for every finite group $F$.) It is finitely presented (K. Brown 1987, reference at the above link). It is not residually finite because it has the finitary symmetric group $S_\infty$ as a subgroup. It is amenable hence sofic.

Edit (Jan. 12 '20): the claim that finitely generated (locally finite)-by-$\mathbf{Z}^d$ groups are Hopfian is false, even for $d=1$ (one can construct a counterexample as central extension of a lamplighter group). Therefore I retract my claim that all finitely generated subgroups of $H_3$ are Hopfian; I don't know if it's true. Yet $H_3$ is Hopfian as well as its finite index subgroups, using that the locally finite kernel is virtually simple).

(As long as this is not fixed the answer should probably be unaccepted.)


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.