4
$\begingroup$

Let $X$ be a subset of a group $G$. We say that $X$ is left amenable with respect to $G$ if there is a function $\mu:\mathcal P(G)\to [0,\infty]$ with the following three properties.

  1. $\mu(A\cup B)=\mu(A)+\mu(B)$ for every pair of disjoint subsets $A, B$ of $G$.

  2. $\mu(gA)=\mu(A)$ for all $g\in G$ and $A\in \mathcal{P}(G)$.

  3. $\mu(X)=1$.

The subset $X$ is right amenable with respect to $G$ if there is a function $\mu:\mathcal P(G)\to [0,\infty]$ that has properties (1) and (3) above as well as property (2') below.

2') $\mu(Ag)=\mu(A)$ for all $g\in G$ and $A\in \mathcal{P}(G)$.

It is easy to see that $X$ is right amenable with respect to $G$ if and only if $X^{-1}$ is left amenable with respect to $G$. Moreover, $G$ being either left or right amenable with respect to $G$ is equivalent to $G$ being an amenable group in the usual sense.

Question. What is an example of a subset $X$ of an amenable group $G$ such that $X$ is left amenable with respect to $G$ but not right amenable with respect to $G$?

Tarski proved that $X$ is left (right) amenable with respect to $G$ if and only if $X$ fails to admit a left (right) $G$-paradoxical decomposition. One current reference for this result is the book "The Banach-Tarski Paradox" by G. Tomkowicz and S. Wagon (Corollary 11.2, page 197 in the second edition).

Hence the question posed is asking for an example of a subset $X$ of an amenable group $G$ that admits a right $G$-paradoxical decomposition but no left $G$-paradoxical decomposition.

$\endgroup$
9
  • $\begingroup$ Well we know $X$ can't equal $X^{-1}$. $\endgroup$ Oct 19, 2022 at 1:53
  • $\begingroup$ So $X$ can't be a subgroup. $\endgroup$ Oct 19, 2022 at 1:55
  • 1
    $\begingroup$ Just for context, the way I'm used to show that a group is not supramenable (i.e. admits a nonempty subset that is not amenable relative to $G$) is to find a rank-2 free subsemigroup $S$ in $G$. But indeed $S$ is neither left nor right amenable in $G$ (if $a,b$ are the free generators of $S$ then $aS$ and $bS$ are disjoint subsets of $S$, but this also holds for $Sa$ and $Sb$). So one is looking at different types of subsets. $\endgroup$
    – YCor
    Oct 19, 2022 at 12:33
  • 1
    $\begingroup$ Do you have a reference for this fact (in a general setting: for $G$ acting on $X$ and $E\subset X$, there's a $G$-invariant mean normalized on $E$ iff $E$ is not $G$-paradoxical? $\endgroup$
    – YCor
    Oct 19, 2022 at 14:41
  • 1
    $\begingroup$ The "only if" direction is trivial, and the "if" part was proved by Tarski. One reference is the book "The Banach-Tarski Paradox" by G. Tomkowicz and S. Wagon (Corollary 11.2, page 197 in the second edition of the book). $\endgroup$ Oct 19, 2022 at 15:06

1 Answer 1

4
$\begingroup$

Here is a slightly modified version of an example that was communicated to the author of the question by Nicolas Monod.

Let $k>1$ be an integer and $G$ the Baumslag-Solitar group $\mathrm{BS}(1,k)$; that is, $$G:=\langle a, b\ :\ bab^{-1}=a^k\rangle.$$ Let $A$ and $B$ be the infinite cyclic subgroups generated by $a$ and $b$, respectively. Notice that $G=A^G\rtimes B$, where $A^G\cong \mathbb Z[1/k]$ is the normal closure of $A$ in $G$. Since $G$ is metabelian, it is an amenable group.

Put $X:=AB$. We claim that, relative to $G$, the subset $X$ is right amenable but not left amenable. To prove the latter assertion, write $X_0:=\langle a^k\rangle B$. Observe that $X_0$ and $aX_0$ are disjoint subsets of $X$, and that $bX=X_0$. Hence the existence of a measure $\mu:\mathcal{P}(G)\to [0,\infty]$ satisfying properties (1), (2), and (3) above would imply $$2=\mu(X)+\mu(X)=\mu(X_0)+\mu(X_0)=\mu(X_0)+\mu(aX_0)\leq \mu(X)=1.$$ Therefore $X$ is not left amenable with respect to $G$.

To establish the right amenability of $X$, we invoke the following proposition, due to J.M. Rosenblatt and appearing as Corollary 1.5 in his paper "A generalization of Følner's condition (Math. Scand. 33 (1973), 153-170)." Rosenblatt states the proposition in terms of left amenability; we have repurposed it for right amenability. Following Rosenblatt, we employ the following notation: for any $n$-tuple $(u_1,\dots,u_n)$ and any set $X$, $$\|X\cap(u_1,\dots,u_n)\|:=|\{ i=1,\dots, n:u_i\in X\}|,$$ where the '$|\ \ |$' notation on the right is for cardinality.

Proposition (Rosenblatt). Let $G$ be an amenable group. A subset $X$ of $G$ is right amenable with respect to $G$ if and only if, for every $m$-tuple $(u_1,\dots, u_m)$ and $n$-tuple $(v_1,\dots,v_n)$ from $G$ such that $m<n$, there exists $g\in G$ such that $$\|gX\cap (u_1,\dots,u_m)\|<\|gX\cap (v_1,\dots, v_n)\|.$$

Let $\{t_{\alpha}:\alpha\in I\}$ be a complete set of coset representatives of $A$ in $A^G$. Then $\{t_{\alpha}X:\alpha \in I\}$ is a partition of $G$. In order to apply Rosenblatt's proposition, we let $(u_1,\dots,u_m)$ and $(v_1,\dots, v_n)$ be finite sequences from $G$ with $m<n$. Then $$m=\sum_{\alpha\in I} \|t_{\alpha}X\cap (u_1,\dots u_m)\|\ \ \mbox{and}\ \ n=\sum_{\alpha\in I} \|t_{\alpha}X\cap (v_1,\dots, v_n)\|.$$ Thus it must be the case that, for some $\alpha\in I$, $$\|t_{\alpha}X\cap (u_1,\dots,u_m)\|<\|t_{\alpha}X\cap (v_1,\dots, v_n)\|.$$ Hence $X$ is right amenable with respect to $G$.

The set $X^{-1}$ is, then, left amenable but not right amenable, answering the question as it was formulated.

$\endgroup$
1
  • $\begingroup$ I have edited the argument to reflect the fact that Rosenblatt's proposition refers to sequences rather than sets. I am grateful to Nicolas Monod for pointing out this oversight. $\endgroup$ Nov 1, 2022 at 13:00

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.