Do there exist non-isomorphic finitely generated groups, $G$ and $H$, along with epimorphisms $\phi:G\rightarrow H$ and $\psi:H\rightarrow G$, such that every generating set of these groups is an image of a generating set, that's mean, for every generating sets $\{g_1,\dotsc,g_m\}$ and $\{h_1,\dotsc,h_n\}$ of $G$ and $H$ respectively, there are generating sets $\{y_1,\dotsc,y_m\}$ and $\{x_1,\dotsc,x_n\}$ such that $$g_i=\psi(y_i),\quad\text{and}\quad h_j=\phi(x_j)$$
-
5$\begingroup$ The condition "every generating subset is an image of an generating subset", for a group epimorphism, is empty, since the inverse image of any generating subset is generating. So it is enough to answer the first part of the question. Take Abels's group as in my answer here mathoverflow.net/questions/59209/hopfian-property, but with $F_p$ replaced by $Z=\mathbf{Z}$. Killing the central $Z[t]$ yields an isomorphic group. The intermediate quotient $2t^{-1}Z\oplus Z[t]$ yields an intermediate quotient which is not isomorphic to the original group with epimorphisms in both directions. $\endgroup$– YCorCommented Dec 30, 2015 at 11:03
-
$\begingroup$ Your "that's mean" is not correct: "every generating set of these group is an image of a generating set" is always true (and remains true if you add "finite"). In words, your (new) assumption is "every finite generating subset of $H$ is a one-to-one image of a generating subset of $G$ and vice versa". Anyway my example fulfills this assumption [and even the stronger assumption that every lift of a generating subset of $H$ is a generating subset of $G$, and vice versa]. $\endgroup$– YCorCommented Dec 30, 2015 at 22:12
-
$\begingroup$ @YCor: Could you please kindly write down your groups $G$ and $H$ in details. It is very interesting for me too. But I could not follow your quick point in the above. $\endgroup$– Alireza AbdollahiCommented Jan 6, 2016 at 5:03
-
$\begingroup$ @YCor: Could you please make me known that if $G$ or $H$ is amenable, finitely presented and residualy finite? $\endgroup$– MSMalekanCommented Jan 6, 2016 at 15:20
-
$\begingroup$ In my example $G$ and $H$ are solvable (hence amenable). They are not finitely presented but I can believe this can be arranged. On the other hand residually finite is hopeless since any groups answering your question are f.g. and non-Hopfian, hence not residually finite. $\endgroup$– YCorCommented Jan 6, 2016 at 16:37
2 Answers
In the paper,
S. Thomas, On the concept of "largeness'' in group theory J. Algebra 322 (2009), no. 12, 4181–4197,
it is shown that your "bi-surjectabilty relation" between finitely generated groups is strictly more complex (in the sense of Borel reducibility) than the isomorphism relation. The proof necessarily gives uncountably many examples of bi-surjective groups which are not isomorphic.
I only want to explain Yves's example in details.
Put $\mathbf Z:=\mathbb Z[t,t^{-1}]$, the ring of Laurent polynomials. Let $B$ be a group of matrices [\begin{pmatrix} 1&P&R\\ 0&D&Q\\ 0&0&1 \end{pmatrix} ] where $P$, $Q$ and $R$ belongs to $\mathbf{Z}$, and $D\in\langle t\rangle$. The group $B$ is easily checked to be finitely generated, and its center consists of unipotent matrices with a single possibly non-trivial element in the upper right corner. It is clearly isomorphic to $\mathbf{Z}$. Also, the map $$ \Phi:B\rightarrow B,\quad \begin{pmatrix} 1&P&R\\ 0&D&Q\\ 0&0&1 \end{pmatrix}\mapsto \begin{pmatrix} 1&P&tR\\ 0&D&tQ\\ 0&0&1 \end{pmatrix}$$ introduces an automorphism of $B$.
Set $N_k:=\bigoplus\{\mathbb Zt^m:m=k,k+1,\dots\}$, $k\in\mathbb Z$. The automorphism $\Phi$, implies that $B/N_k\cong B/\Phi(N_k)=B/N_{k+1}$, $k\in\mathbb Z$. Now, let $G:=B/N_0$ and $H:=B/(2\mathbb Z\oplus N_1)$. Then non-isomorphic groups $G$ and $H$ satisfy the conditions.
-
$\begingroup$ Oops, $\mathbf{Z}$ usually denotes $\mathbb{Z}$, I would have avoided this notation. $\endgroup$– YCorCommented Feb 13, 2016 at 18:58
-
$\begingroup$ Yes, it works this way. $H$ is not isomorphic to $G$ because the center of $G$ is torsion-free but the center of $H$ is not. $\endgroup$– YCorCommented Feb 13, 2016 at 19:00
-
2$\begingroup$ @YCor I guess you could say $\mathbf{Z} = \mathbb{Z}[\mathbb{Z}]$ (the group ring of the group $\mathbb{Z}$). :-) $\endgroup$ Commented Mar 15, 2016 at 1:48