Consider a category $\mathsf C$ admitting a quadruple adjunction as below.

$$(\Pi_0 \dashv \text{disc} \dashv \Gamma \dashv \text{codisc}) : \mathsf{C} \stackrel{\stackrel{\longrightarrow}{\longleftarrow}}{\stackrel{{\longrightarrow}}{{\longleftarrow}}} \mathsf{B} \;$$ A good example is the category of locally connected spaces with $\mathsf{B}=\mathsf{Set}$. In this case the connected-components functor $\Pi_0$ implies $\mathsf C$ is extensive.

Suppose now $\mathsf C$ is extensive and admits the exponentials below. In Axiomatic Cohesion, Lawvere defines an adjoint quadruple as above to be cohesive if:

  1. $\Pi_0$ preserves finite products and $\text{codisc}$ is fully faithful.
  2. $\Pi_0$ preserves $\mathsf{B}$-parametrized powers in the sense of a natural isomorphism $$\Pi_0(X^{\text{disc}I})\cong \Pi_0(X)^{I}.$$
  3. The canonical arrow $\Gamma\Rightarrow \Pi_0$ is epimorphic.

Lawvere later defines such an adjunction to be sufficiently cohesive if each object has a "contractible envelope":

  1. For every $X$ there exists a monic map $X\to Y$ with $Y$ contractible in the sense that $\Pi_0(Y^A)=\mathbf 1$ for all $A$.

Then it is proved that a topos of cohesion ($\mathsf C$ is a topos) is sufficiently cohesive iff the truth-value object is connected, and also iff all injective objects are connected. Here, connectedness of $X$ means $\Pi_0(X)\cong \mathbf 1$.

For locally connected spaces it seems the injective objects are the inhabited indiscrete spaces, whose connectedness reflects their cohesion. This is opposed to case of sets where $\text{codisc}$ is the identity and sets are seldom singletons.

Can a sufficiently cohesive category (quadruple adjunction) have a conservative $\Gamma$?

I ask for the following reason: For locally connected spaces $\Gamma$ is the "points functor" and I think of the fact it is faithful but not conservative as reflecting that "discontinuous" arrows (i.e breaking cohesion) may be underlain by "isomorphisms on stuff"; I want to know if this is "correct" intuition or not.

  • $\begingroup$ Just a self-note to start thinking about the answer: $\Gamma$ is conservative iff the counit of $d\dashv \Gamma$ is a strong epi. There are many categories on the $n$Lab page that exhibit cohesion over (s)Set or another topos, you can start checking how often $\Gamma$ is conservative with this criterion. $\endgroup$
    – fosco
    Nov 13 '17 at 0:44

Let ${\phi : \mathrm{disc} \rightarrow \mathrm{codisc}}$ be the canonical map.

The map ${\Gamma \phi }$ is an iso in the base so, if $\Gamma$ is conservative, $\phi$ is an iso.

In other words, if $\Gamma$ is conservative, the original string of adjoints is a quality type. So, if the original string is sufficiently cohesive then the whole thing is degenerate.

  • $\begingroup$ Dear @Mendieta, why do you call such a string a "quality type", and where can I read more about such terms (preferably with intuition)? $\endgroup$
    – Arrow
    Oct 4 '20 at 10:23
  • $\begingroup$ Dear @Arrow, the terminology and intuition comes from the paper by Lawvere that you cite in your question. They are also studied in the earlier tac.mta.ca/tac/volumes/1996/n8/n8.pdf by Johnstone, with different terminology, naturally. $\endgroup$
    – Mendieta
    Oct 4 '20 at 14:44
  • $\begingroup$ @Arrow, there is also a talk by F. Marmolejo in CT2017, but I don't know if that was published. mat.uc.pt/~ct2017/slides/marmolejo_f.pdf $\endgroup$
    – Mendieta
    Oct 4 '20 at 15:35
  • $\begingroup$ Thank you very much! $\endgroup$
    – Arrow
    Oct 5 '20 at 16:09

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.