I have some issue to solve following exercise. The exercise is from a French book on Algebra (cours d'Algèbre) from Jean Querré. The book is from the 1970's.

If the center $Z(G)$ of a group $G$ is cyclic and $G/Z(G)$ is commutative then there exists a group $H$ such that $H \times H$ is isomorphic to $G/Z(G)$.

  • $\begingroup$ Is $G$ finite or possibly infinite? $\endgroup$ Jun 28, 2015 at 2:31
  • 1
    $\begingroup$ In particular case , If $G$ be abelian then $Z(G)=G$ and $H=\{1\}$.(in this case $Z(G)$ is not nessecery cyclic) $\endgroup$
    – 49-49
    Jun 28, 2015 at 2:52
  • $\begingroup$ @Matt Samuel Nothing special is mentionned in the exercise regarding the group cardinal. $\endgroup$ Jun 28, 2015 at 13:01
  • $\begingroup$ The paper Groups with preassigned central and central quotient group by Baer should be useful, see corollary 7.2. $\endgroup$ Jun 29, 2015 at 11:50

1 Answer 1


This answer is almost complete and I think (read hope) that it's correct up to the point where I get stuck. (I've posted this as an answer because it's too long for a comment and I should go to bed.)

It suffices by the first isomorphism theorem to find a group $H$ and a surjective homomorphism $\theta : G \to H \times H$ for which $\ker \theta = Z(G)$.

By spelling out the definitions it is easy to see that $G/Z(G)$ is commutative if and only if $g^{-1}h^{-1}gh \in Z(G)$ for all $g,h \in G$, which occurs if and only if $[G,G] \le Z(G)$, where $[G,G]$ is the commutator subgroup of $G$.

This implies that the map $G \to [G,G]$ defined by $x \mapsto g^{-1}x^{-1}gx$ is a homomorphism for fixed $g \in G$, since if $y \in G$ then $$\begin{align} (g^{-1}x^{-1}gx)(g^{-1}y^{-1}gy) &= g^{-1}(x^{-1}gxg^{-1})y^{-1}gy \\ &= g^{-1}y^{-1}(x^{-1}gxg^{-1})gy \\ &= g^{-1}y^{-1}x^{-1}gxy \\ &= g^{-1}(xy)^{-1}g(xy) \end{align}$$ Likewise the map $x \mapsto x^{-1}h^{-1}xh$ is a homomorphism for fixed $h \in G$.

Since $Z(G)$ is cyclic, so is $[G,G]$, and hence it is generated by a single element $g^{-1}h^{-1}gh$ for some $g,h \in G$.

Fix $g,h \in G$ such that $[G,G] = \langle g^{-1}h^{-1}gh \rangle$. Let $H = [G,G]$ and define $\theta : G \to [G,G] \times [G,G]$ by $$\theta(x) = (g^{-1}x^{-1}gx, x^{-1}h^{-1}xh)$$ Now

  • We know $\theta$ is a homomorphism by above remarks.
  • $\theta$ is surjective. Indeed, if $a \in [G, G] \times [G, G]$ then there exist $m,n$ such that $$a = ( (g^{-1}h^{-1}gh)^m, (g^{-1}h^{-1}gh)^n )$$ But $\theta(h) = ( g^{-1}h^{-1}gh, 1 )$ and $\theta(g) = ( 1, g^{-1}h^{-1}gh )$, so $$a = \theta(h)^m\theta(g)^n = \theta(h^mg^n)$$ hence $a \in \mathrm{im}\, \theta$.
  • (This is where I get stuck.) We need to show that $\ker \theta = Z(G)$. The $\supseteq$ direction is obvious. For $\subseteq$, suppose $x \in \ker \theta$ and let $y \in G$. We need to show $xy=yx$. Well $x \in \ker \theta$ implies $xg=gx$ and $xh=hx$, so $x$ commutes with $g$ and $h$ (and their inverses). I'm sure a bit more hacking would yield $xy=yx$ but I haven't quite managed to do it. The fact that $x^{-1}y^{-1}xy \in [G,G]$ and hence $x^{-1}y^{-1}xy = (g^{-1}h^{-1}gh)^n$ for some $n$ should be useful.
  • 3
    $\begingroup$ A $p$-group $P$ is called special if its center $Z(P)$, derived subgroup $P'$ and Frattini subgroup $\Phi(P)$ are all equal. If this subgroup is additionally cyclic, then $P$ is called extraspecial. As there are extraspecial groups of order $p^n$ for all odd n and primes $p$, you won't be able to finish your proof with $H = G'$. Try looking at a maximal abelian subgroups $A$ of $G$ instead, and see if $H = A/Z(G)$ works. $\endgroup$
    – j.p.
    Jun 28, 2015 at 14:16
  • 1
    $\begingroup$ On second thought, taking a maximal abelian subgroup won't work. Instead assuming $G$ finite, I'd reduce to the case $G$ being a $p$-group. $(g, h)\mapsto [g, h]$ induces a non-degenerate skew-symmetric bilinear map $G/Z\times G/Z \to G'$. I'd induct on the order of $G/Z$ by finding a matching element for an element of $G/Z$ of maximal order. $\endgroup$
    – j.p.
    Jun 29, 2015 at 15:29
  • 1
    $\begingroup$ I think the finite case, and even the finitely generated case, is relatively easy. For the general case you probably need to take a maximal subgroup that does decompose in the required way and then show it has to be the whole group. I tried to write it down but I didn't get the details right. $\endgroup$
    – Derek Holt
    Jun 30, 2015 at 7:33
  • $\begingroup$ @DerekHolt: Any chance that you'll give it a second try? The infinite case would interest me. Thanks! $\endgroup$
    – j.p.
    Jul 10, 2015 at 16:24

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.