4
$\begingroup$

Let $H=H_1\ast H_2$ be a free product of non-trivial groups $H_1$ and $H_2$. We call an element $h\in H$ hyperbolic if $h\not\in H_i^g\overset{\textrm{def}}{=}\left \{ g^{-1}fg\ |\ f\in H_i \right \}$ for $i=1,2$ and $g\in H$.

Let $h_1,h_2\in H$ be hyperbolic. It is not very difficult to prove that if $h_1$ and $h_2$ are hyperbolic and $[h_1,h_2]\overset{\textrm{def}}{=}h_1^{-1}h_2^{-1}h_1h_2\ne 1,$ then $[h_1,h_2]$ is hyperbolic.

Is it possible that the element $[h_1,h_2]$ is conjugate (in $H$) to its inverse? That is, can the element $[h_1,h_2]$ be of the form $[h_1,h_2]=g_1g_2$, where $g_1^2=g_2^2=1$?

$\endgroup$
3
  • $\begingroup$ What does $H_i^g$ mean? $\endgroup$
    – user1688
    Dec 18, 2017 at 8:38
  • 2
    $\begingroup$ You seem to be claiming that if $[h_1,h_2]$ is conjugate to its inverse then it is necessarily a product of two involutions. That's certainly not true for arbitrary elements $h_1,h_2$ in an arbitrary group. If it is true under your special assumptions, then you should explain why. $\endgroup$ Dec 18, 2017 at 10:28
  • $\begingroup$ @NeilStrickland I provided a justification in my answer $\endgroup$
    – YCor
    Dec 18, 2017 at 21:00

1 Answer 1

4
$\begingroup$

Yes:

Assuming that $H_1$ has an element $u$ of order 2 and $\max(|H_1|,|H_2|)\ge 3$, there exist hyperbolic $h_1,h_2$ such that $[h_1,h_2]$ is conjugate to its inverse and nontrivial (and hyperbolic).

First case: $|H_2|\ge 3$. Pick two distinct nontrivial commuting elements $b,c$ in $H_2$ (exercise: these exist). Then writing $b^{-1}=B$, $c^{-1}=C$ $$b[ub,uc]B = b(BuCuubuc)B = uCbucB = uCbuBc,$$ which is hyperbolic and conjugate to its inverse $CbuBcu$ (it is indeed product of two elements of order 2: $u$ and $CbuBc$).

Second case: $|H_1|\ge 3$. Pick $a\in H_1\smallsetminus\{1,u\}$ and $b\in H_2\smallsetminus\{1\}$. Then, writing $a^{-1}=A$, $b^{-1}=B$ $$b[aub,ab]B=b(BuABAaubab)B=uABuba,$$ which is hyperbolic and conjugate to its inverse $ABubau$ (it is indeed the product of two elements of order 2: $u$ and $ABuba$).

If your claim (that the only possibility for a nontrivial commutator of hyperbolic to be conjugate to its inverse is to be product of two elements of order 2) is correct ($\ast$), then it is a necessary condition that $H_1$ or $H_2$ has an element of order 2.

($\ast$) I finally can check it (this thus addresses Neil Strickland's comment): suppose that some element of a free product $H_1 \ast H_2$ is hyperbolic and conjugate to its inverse. Let $P$ be the stabilizer of its axis in the Bass-Serre tree. Then $P$ acts on this axis and does not preserve an orientation. Let $Q$ be the fixator (=pointwise stabilizer) of this axis. Then $Q$ fixes a directed edge, which in the case of the Bass-Serre tree of a free product has a trivial stabilizer; Hence $Q$ is trivial. So $P$ acts faithfully on the axis, and hence is an infinite dihedral group. This proves that there exist elements of order 2, and more precisely shows that hyperbolic elements in $P$ are products of two elements of order 2.

$\endgroup$

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.