Graded homotopy structures

29th July, 2019

As I mentioned in a previous post, I recently saw a talk by Rachel Hardeman on the A-homotopy theory of graphs, and it really intrigued me. In particular, it seemed to me that there was some nice structure that could be abstractified: that of a “graded homotopy structure”, as I’ve been calling it in my head. Rather than trying to type out everything in #math.CT:matrix.org, I’ve decided to post it here, in the hope that I might be able to get some answers.


16/9/19 EDIT: As one of my supervisors recently pointed out to me, the property of being an `-homotopy’ is not transitive, and so this example is really a non-example. I’ll keep the post here for reference purposes, but the only useful/true bits are those quoted from [RH19].


The main reference is [RH19] Rachel Hardeman. Computing A-homotopy groups using coverings and lifting properties. arXiv: 1904.12065.

Preliminaries

  • Graphs consist of vertices and edges , where we write the edge between vertices and as . All graphs are assumed to be simple (no multiple edges between any two points or loops on a single point) and have a distinguished vertex . We write to mean the graph along with its distinguished vertex.
  • A (weak) graph homomorphism is a map of sets such that, for all , either or . It is said to be based if .
  • The cartesian product of the graphs and is the graph with vertex set , with distinguished vertex , and with an edge between and whenever
    • and ; or
    • and .
  • The path of length , denoted by , is the graph with vertices labelled from to , and edges for . The path of infinite length, denoted by , has vertices labelled by .
  • We say that two graphs homomorphisms are A-homotopic, written , if there exists some and a graph homomorphism such that
    • for all ;
    • for all ; and
    • for all .

    We say that two graphs and are A-homotopic if there exist graph homomorphisms and such that and .

  • N.B. A-homotopy theory is possibly very different from what you might, at a quick first glance, expect. For example, any two cyclic graphs and (for ) are A-homotopic if and only if , and is contractible (i.e. homotopic to the graph with a single point and no edges) for , but not for any .
  • The A-homotopic fundamental group of a graph can be defined, as well as a simplicial structure on the group of cochains, and all this sort of stuff. This (amongst other nice formalisations that we would hope for) is done in [RH19].

“Graded homotopy” structure

Given two A-homotopic graph homomorphisms we can ask for the minimal such in the definition of the A-homotopy. We then say that the A-homotopy is an -homotopy, and we extend this definition slightly to allow for the fact that we can trivially consider an -homotopy as an -homotopy (in various ways, corresponding to the classical idea of simplicial (co)face/(co)degeneracy (depending on your choice of nomenclature) maps). That is, we say -homotopic to mean “-homotopic with ”.

We say that two graphs are -homotopic if there exist graph homomorphisms and such that is -homotopic to and is -homotopic to , with .

Then we can consider the category , which has objects being equivalence classes of -homotopic graphs, and morphisms being equivalence classes of -homotopic graph homomorphisms. This gives us the following structure:

  • ;
  • ;
  • functors that are surjective on objects, where functoriality relies on the fact that -homotopies can be considered as -homotopies.

We can think of the number as some sort of “complexity” of the homotopy: small correspond to “homotopies that can be performed in a few steps” (here it is a good idea to see some of the examples in [RH19] to get an idea of how graph homotopies behave).

Questions

If anybody has any answers to, or comments about, the following questions (or this post in general) then please don’t hesitate to get in touch!

  1. What is this structure? Some sort of enrichment? Does it already have a name?
  2. What other examples exist? For example, it would be nice to get something similar for the category of chain complexes of an abelian category, but I see no way a priori of assigning “complexity” to a homotopy for an arbitrary choice of abelian category. If things are enriched over metric spaces, however, then this is a different story…
  3. It seems believable that we could define something analogous with instead of . Could we do so for arbitrary (bounded-below) posets?
  4. Does this tie in to the idea of “approximate composition” (c.f. Walter Tholen’s talk on metagories at ACT2019).