Let $G$ be a finitely generated group (optionally torsion-free). Let $N$ be a submonoid of $G$ (that is, a subsemigroup with $1$).

A (cancellative) monoid/semigroup $S$ is right reversible if for all $a,b \in S$, it follows that $Sa \cap Sb \neq\emptyset$. This is the Ore condition for semigroups, and allows one to invert the elements of $S$ so that the set of products, $S\cdot S^{-1}$, is a group.

Suppose that $G$ is nilpotent. Does it follow that every submonoid $N$ is right reversible?

What are the right conditions on $G$ to guarantee this property?

Something like nilpotent by finite comes to mind, since this implies the group ring ${\bf Z} G$ is Goldie. Obviously, this is a strong condition on $G$, since it fails for large classes of groups, e.g., those containing a non-abelian free group, or probably, those for which the group ring is not Goldie.

The question arose from a study of random walks on discrete groups, related to the boundary of one of the usual boundaries. It is a little too complicated to explain the connections here.


It follows from Lemma 2 in Grigorchuk's https://page-one.live.cf.public.springer.com/pdf/preview/10.1007/BF01138837 that all subsemigroups of a group $G$ are right reversible iff it contains no free semigroup on 2 generators.

A free subsemigroup is not right reversible. Conversely suppose $S$ is not right reversible (say $aS\cap bS=\emptyset$, note for me right reversible is dual to the OP's definition). Then observe that $a,b$ generate a free subsemigroup. For words beginning with distinct letters are not equal and if they begin with the same letter you can cancel it to get a shorter relation.

  • 1
    $\begingroup$ And from Theorem 1, necessary and sufficient is that the group be almost nilpotent. Thanks, Ben. $\endgroup$ – David Handelman Jul 29 '17 at 18:47
  • $\begingroup$ Yes. Any group of subexponential growth has this property. $\endgroup$ – Benjamin Steinberg Jul 29 '17 at 18:53
  • $\begingroup$ My colleague, Vadim Kaimanovich, just told me that the result also follows from an argument in his 1985 thesis. $\endgroup$ – David Handelman Jul 29 '17 at 23:09
  • $\begingroup$ Could be. I think Grigorchuk also announced this in a 1985 conference proceedings $\endgroup$ – Benjamin Steinberg Jul 29 '17 at 23:34

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.