Let’s take another look at the ArXiv, shall we? (Originally the plan was for this to be weekly, oops.)
One of the great triumphs of mid-20th-century combinatorial group theory was the result (due to Higman, Neumann, and Neumann) that any countable group can embed into a 2-generator group. This paper starts by mentioning this result and several other embedding results in the same vein.
[As an aside, in the introduction the authors state of the above theorem, “This was a major breakthrough, providing some of the first evidence that finitely generated groups are not structurally simpler than countable groups and thus are far from tame or classifiable.” Descriptive set theory actually gives us a framework to discuss exactly these issues, and in this framework the above sentence is half-right. One speaks of the complexity of the equivalence relation associated to a classification problem, and two different equivalence relations are compared using Borel reductions. These slides (pdf) give an overview.
Presumably here one would like to talk about classifying groups up to isomorphism. It’s true that classifying finitely generated groups up to isomorphism is far from tame; indeed isomorphism of finitely generated groups is what we call a universal countable Borel equivalence relation, meaning it’s as complicated as it could possibly be. However, isomorphism of countable groups is a universal among all equivalence relations arising from a (Polish, i.e. continuous) -action. This is far more complicated than universal countable. So it would seem safe to say that finitely generated groups are in fact structurally simpler than countable groups, although they are still not structurally simple in absolute terms.
If instead we were to look at the biembeddability relation (the equivalence relation arising naturally from the embeddability structure), then the gap is even larger. Biembeddability of finitely generated groups is a universal countable Borel equivalence relation, but biembeddability of countable groups is a universal analytic equivalence relation. This (largely expository) article from the Notices has a picture of how these are related.]
All of these results are in the context of countable discrete groups. As of late, there has been a renewed interest in topological groups (previously, much of the focus here was on Lie groups), and so it makes sense to ask if there are results of a similar flavor to be found here. Natural analogs of “countable” and “finitely generated” are “-compact” and “compactly generated”. One might hope for a theorem like that of Higman, Neumann, and Neumann, with these terms in place of the originals. An old result of Pestov shows that this holds for topological Hausdorff groups. The authors show that this is not possible when looking at locally compact groups:
Theorem [Caprace, Cornulier]: There exists a second countable (hence -compact), topologically simple totally disconnected locally compact group S, such that every continuous (or even abstract) homomorphism of S to any compactly generated locally compact group is trivial.
There are some sharper results in the paper, and a positive result regarding embedding abelian groups. I’m particularly interested in ruling out some group embedding into entire classes of other groups, as I’m currently interested in the question of whether every group embeds into a co-Hopfian group. (See here.) In this paper they manage to relate the fact that a group H has no non-trivial continuous actions on a graph of bounded degree to the fact that all of its continuous homomorphisms to compactly generated groups are trivial. (This is not an equivalence, by the way.) Could there be something similar when talking about embeddings of countable groups into co-Hopfian groups?
Volodymyr Nekrashevych (Texas A&M)
The title of this paper brings to mind a previous paper of Nekrashevych, “A minimal Cantor set in the space of 3-generated groups”. In that paper, he defines a family of branch, just-infinite finitely generated groups for which isomorphism is not smooth (in the sense of Borel equivalence relations). This is notable, because there are very few constructions of this sort. Generally speaking, when group theorists wish to show a given class of groups is large, they perform a smooth construction of continuum-many distinct-up-to-isomorphism groups. But as mentioned earlier, isomorphism of finitely generated groups is rather complicated, and so such constructions could in theory be much more complex. Constructions using free products with amalgamation are currently the only ones we know of which achieve the full complexity. Free products with amalgamation let you do all sorts of things, so the question is if there are any other natural classes of groups for which isomorphism is this complex, for example amenable groups. The difficulty is that one is more restricted in the sorts of constructions one can do while staying in such classes. A first step is to show that isomorphism for these classes is not smooth, as Nekrashevych did for just-infinite branch groups.
This paper uses the same family of groups. The main result is that the members of this family are residually finite and have the same profinite completion. A few other properties of the family are also mentioned. This paper stands alone, and so there is significant overlap with the one I was discussing last paragraph. I imagine that the exposition of the construction might be more polished, simply due to the additional time Nekrashevych has spent thinking about it, but I’ll admit to not having read either paper in depth at this point.