We say $G$ is the Zappa-Szep product of two subgroups $K$ and $P$ if $K\cap P = \{e\}$ and the function $K\times P \to G$, $(k,p)\mapsto kp$, is bijective.

The Iwasawa decomposition shows that we can have amenable $K$ and $P$ such that $G=KP$ is non-amenable. However in such examples $K$ is usually not amenable when considered as a discrete group.

Where is a good place for me to look for examples of "naturally occurring" Zappa-Szep products where the two "factors" are amenable as discrete groups, yet the larger group itself is non-amenable? Can this ever happen for e.g. lattices in semisimple Lie groups?

  • 4
    $\begingroup$ The first example is $\text{SL}_2(\mathbb{R})$, where $K=\text{SO}(2)$ is abelian and $P$, the upper-right matrices group, is solvable. $\endgroup$ – Uri Bader Jan 9 '18 at 7:59
  • 1
    $\begingroup$ Other simple Lie groups will not have this property (unless I am forgetting a very special case). This is true also over the $p$-adics. Lattices in such (including the $\text{SL}_2$ case) will not be either - they have very few amenable subgroups. $\endgroup$ – Uri Bader Jan 9 '18 at 8:08
  • $\begingroup$ @UriBader Thanks - I was going to include SL(2,R) in the question but was typing in haste. I someow think of SL(2,R) with discrete topology as not very natural, so what I was hoping for was something countable and finitely presented. But I admit that this is just wishful thinking $\endgroup$ – Yemon Choi Jan 9 '18 at 8:35

Here is an observation of Yair Glasner that I previously missed.

Let $k$ be the field of real algebraic numbers and consider $G=\text{PSL}_2(k)$. Let $P=B(\mathcal{k})$ be the standard Borel subgroup and $K=\text{PSO}_2(k)$. Then $K$ and $P$ are solvable, $K\cap P=\{e\}$ and $G=KP$, but $G$ is not amenable (as it contains free groups). All countable examples I am aware of are variants of this example. In particular, I am not aware of any finitely generated example.

  • $\begingroup$ Thanks a lot for this answer. I'll leave it for a while in case any other answers turn up. $\endgroup$ – Yemon Choi Jan 12 '18 at 20:47
  • 1
    $\begingroup$ You need $k$ to consist of real algebraic integers, or at least to have no square root of $-1$. Indeed otherwise the matrix $\begin{pmatrix}1 & 0\\ i & 1\end{pmatrix}$ is not in $KP$. Actually, if you consider $PGL_2$ (to avoid determinant issues), any field $k$ in which $-1$ is not a square works. $\endgroup$ – YCor Jan 14 '18 at 7:39

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.