In the article cited below, I. Castellano gives a proof for the following result (Proposition 4.1).

Let $G$ be a compactly generated totally disconnected locally compact group. Suppose that $G$ acts discretely on a tree $\mathcal T$ such that

  1. the group $G$ is acting without edge inversions;
  2. the quotient graph $G\backslash\mathcal T$ is finite;
  3. the edge stabilisers $G_e$ are compact open subgroups of $G$.

Then the vertex stabilizers $G_v$ are compactly generated.

Here acting discretely means that the stabilizers are open subgroups of $G$.

The proof as cited is rather involved and uses cohomology arguments, while the proposition looks rather innocent. Is there a more straightforward, elementary proof for this result? Note that the proposition should also hold for trees that are not locally finite (and in this case, the topology becomes a lot more subtle).

[1] Castellano, I., Rational discrete first degree cohomology for totally disconnected locally compact groups. arXiv:1506.02310 [math.GR].

  • 3
    $\begingroup$ Yes I think it can be proved using plain Bass-Serre theory. I have to think about details... $\endgroup$ – YCor Sep 10 '19 at 20:14

As YCor suggests, proceed by Bass–Serre theory. We can write $G$ in the form

$\frac{G_{v_1} \ast \dots \ast G_{v_m} \ast F(E)}{\langle \langle \overline{e}\alpha_e(g)e\alpha_{\overline{e}}(g)^{-1} \; (g \in G_e), \; e\overline{e}, \; e \; (e \in E') \rangle \rangle}$

where $E$ is a set of representatives for the edges in the quotient graph, $E' \subseteq E$ is a set of representatives for the edges in a spanning tree of the quotient graph, and $\alpha_e: G_e \rightarrow G_{o(e)}$ is the natural embedding of an edge group into the vertex group of the origin of the edge. Note that $E$ is finite and the Bass–Serre relations are gluing together vertex groups along compact open subgroups $G_e$. (This also ensures that everything is fine in terms of the group topology: you can just start with the topology of one of the edge groups, and then as a group topology, it extends uniquely to the whole of $G$.) Write $F$ for the image of $F(E)$ in $G$.

Let's look at $G_{v_1}$. We have a compactly generated open subgroup $H_0$ of $G_{v_1}$ generated by $\alpha_e(G_e)$ for all $e \in E$ such that $o(e)=v_1$. In other words, all the amalgamation of $G_{v_1}$ with the other vertex groups happens inside $H_0$.

Now write $G_{v_1}$ as a directed union $\bigcup_{i \in I}H_i$, where each $H_i$ is compactly generated and $0$ is the least element of $I$. Let $K_i$ be the subgroup of $G$ generated by $H_i \cup G_{v_2} \cup G_{v_3}\cup\dots\cup G_{v_n} \cup F$. Then $G = \bigcup_{i \in I}K_i$; since $G$ is compactly generated, in fact $G = K_i$ for some $i$. In particular, every $g \in G_{v_1}$ can be expressed as a product of elements of $H_i \cup G_{v_2} \cup G_{v_3}\cup\dots\cup G_{v_n} \cup F$. Using the normal form theorem for graphs of groups, we conclude that $g \in H_i$. Thus $G_{v_1}$ is compactly generated.

  • 1
    $\begingroup$ Thank you! This is indeed more natural. I will try to generalise this method to groups acting on polyhedral complexes (using complexes of groups in the sense of Bridson and Haefliger). $\endgroup$ – Jens Bossaert Oct 10 '19 at 12:47

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.