6
$\begingroup$

This question arose from a discussion in the comment section of another MathOverflow question with Mike Shulman and Alec Rhea, which raised the following point:

Vague Question: Are adjunctions an appropriate notion of "sameness" for categories?

While viewing adjunctions as such might seem strange, there seem to be a few reasons to adopt this point of view:

  1. They are in a sense a weakening of equivalences: any equivalence of categories $F,G\colon\mathcal{C}\rightleftarrows\mathcal{D}$ can be made into an adjoint equivalence $(F,G,\eta,\epsilon)\colon\mathcal{C}\rightleftarrows\mathcal{D}$, and these are in turn precisely those adjunctions whose co/units are natural isomorphisms.
  2. Secondly, having an adjunction between categories $\mathcal{C}$ and $\mathcal{D}$ implies that $\mathrm{N}_{\bullet}(\mathcal{C})$ and $\mathrm{N}_{\bullet}(\mathcal{D})$ are homotopy equivalent as simplicial sets.

So, given a functor $F\colon\mathcal{C}\longrightarrow\mathcal{D}$, we have the following chain of strict implications $(1)\Rightarrow(2)\Rightarrow(3)$:

  1. $F$ admits a left or a right adjoint $G\colon\mathcal{D}\longrightarrow\mathcal{C}$.
  2. Taking nerves induces a simplicial homotopy equivalence $\mathrm{N}_{\bullet}(F)\colon\mathrm{N}_{\bullet}(\mathcal{C})\longrightarrow\mathrm{N}_{\bullet}(\mathcal{D})$.
  3. Taking nerves induces a simplicial weak homotopy equivalence $\mathrm{N}_{\bullet}(F)\colon\mathrm{N}_{\bullet}(\mathcal{C})\longrightarrow\mathrm{N}_{\bullet}(\mathcal{D})$. That is, $F$ is a Thomason equivalence.

In particular, the difference between conditions $(1)$ and $(2)$ is that adjoint functors require not only the existence of natural transformations $\eta\colon\mathrm{id}_{\mathcal{C}}\Rightarrow G\circ F$ and $\epsilon\colon F\circ G\Rightarrow\mathrm{id}_{\mathcal{D}}$, corresponding to simplicial homotopies $\mathrm{N}_{\bullet}(\mathrm{id}_{\mathcal{C}})\Rightarrow\mathrm{N}_{\bullet}(G\circ F)$ and $\mathrm{N}_{\bullet}(F\circ G)\Rightarrow\mathrm{N}_{\bullet}(\mathrm{id}_{\mathcal{D}})$, but also that these be coherent natural transformations/homotopies in the sense that they satisfy the triangle equalities. So we could perhaps summarise these conditions as follows:

  1. $F$ is a "coherent homotopy" equivalence.
  2. $F$ is a homotopy equivalence.
  3. $F$ is a weak homotopy equivalence.

In trying to assess the differences between these, one may consider the following question:

Question. Which properties of categories are preserved under the above conditions?

That is, given categories $\mathcal{C}$ and $\mathcal{D}$ such that we have a (coherent, weak) homotopy equivalence of categories $\mathcal{C}\to\mathcal{D}$, what are some properties of $\mathcal{C}$ that $\mathcal{D}$ ends up possessing as well, or vice versa?

$\endgroup$
2
  • 5
    $\begingroup$ It seems relevant to say something about basic localizers. That is, following Grothendieck, the functors inducing weak equivalences of classifying spaces are the smallest class of functors containing the adjunctions, closed under Quillen's theorem A, and maybe a couple of other closure conditions. However, my sense is that very few categorical properties are preserved by weak homotopy equivalence. Note also that many categorical properties imply that a category is weakly contractible. $\endgroup$ Commented Apr 23, 2021 at 22:46
  • 3
    $\begingroup$ The weak homotopy type of $C$ is exactly the homotopy type corresponding to the $\infty$-groupoid $C[C^{-1}]$, so that the answer to 3 is "any property of the $\infty$-groupoid $C[C^{-1}]$ " (in particular properties of its fundamental groupoid, or fundamental group). For 1., there will be stronger things like if $C$ has an initial and a terminal element, $D$ has at least one of them $\endgroup$ Commented Apr 24, 2021 at 8:37

0

You must log in to answer this question.

Browse other questions tagged .