MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let G be a group and g and h be elements in G. If g commutes with all conjugates of g and h commutes with all conjugates of h, can one conclude that gh commutes with all conjugates of gh?


share|cite|improve this question
up vote 9 down vote accepted

No; the smallest counterexamples are given by the groups SmallGroup(54,5) and SmallGroup(54,6) (in GAP's SmallGroups library); these are groups of the form $$G_1 = ((3 \times 3) : 3) : 2 \text{ and } G_2 = (9 : 3) : 2.$$ In both cases, there are 15 elements with the given property (i.e. that they commute with all their conjugates), so they can't possibly form a subgroup.

share|cite|improve this answer
These are similar to Mark Sapir's example, except the extra element of order 2 makes the normal closure of gh a little bigger, and so non-abelian. In his example, too many of the normal subgroups are abelian, and your extra automorphism removes some normal subgroups (of the Fitting). – Jack Schmidt Jun 14 '11 at 16:29
Thank you so much for your answer! May I know how $G_1$ and $G_2$ are defined and how do you know that there are exactly 15 elements in each group with the given property? – Zuriel Jun 15 '11 at 10:54
@Zuriel: I wrote a tiny computer program in Sage to find these examples. But Jack's comment indicates how to construct $G_1$ explicitly as a subgroup of $PSL(3,3)$: it is the subgroup generated by the Sylow $3$-subgroup consisting of upper-triangular matrices with $1$'s on the diagonal, together with the involution $\operatorname{diag}(1, -1, -1)$. You can now explicitly compute the conjugacy classes by hand, if you wish, and you will see that $5$ of these classes are commutative, with sizes $1 + 2 + 3 + 3 + 6 = 15$. – Tom De Medts Jun 16 '11 at 8:38

You have that the normal subgroups $U, V$ generated by $g$ and by $h$ are Abelian. This only implies that $UV$ is normal and solvable of class 2. Take the group of unitriangular 3 by 3 matrices $H_3$ over ${\mathbb Z}_p$.

It has two normal Abelian subgroups $U,V$ containing the center (= the derived subgroup) generated by the elementary matrix $E_{1,3}(1)$: $U$ is generated as a subgroup by the center and $E_{1,2}(1)$, $V$ is generated by the center and $E_{2,3}(1)$. Both subgroups are normal because they contain the derived subgroup. The product of these two subgroups is the whole $H_3$ (hence non-abelian). So you can take $g=E_{1,2}(1), h=E_{2,3}(1)$.

share|cite|improve this answer
This answers a slightly different question. For your g,h, the normal subgroup generated by gh is also abelian, though the subgroup generated by {g,h} is not abelian. – Jack Schmidt Jun 14 '11 at 16:27
@Jack: You are right. I did not understand the question. An element $g$ such that $g^G$ is commutative is called Engel of degree 2: it satisfies $[[x,g],g]=1$ for every $x\in G$. So the question asks whether the product of two Engel elements of degree 2 is an Engel element of degree 2. There are classical papers by Baer and others about it. – Mark Sapir Jun 14 '11 at 19:32
Thank you for your answer! May I know more details (title, etc.) of those classical papers by Baer and others? I would like to read them. – Zuriel Jun 15 '11 at 10:57

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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