Want to take part in these discussions? Sign in if you have an account, or apply for one below
Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.
1 to 6 of 6
Given a locally connected topos Γ:E⇆Set:Δ, we have the extra left adjoint π0⊣Δ. π0 preserves epis, as any left adjoint does. If E is boolean I’m sure that π0 also reflects epis, at least assuming classical logic in the background. And I can’t think of a non-boolean example where π0 doesn’t reflect epis (just my own ignorance, rather than any compelling mathematical reason).
Alternatively, I’m suspicious of π0 reflecting epis in the case of boolean E without classical logic.
To be concrete, say I’ve got an epi T→V in boolean locally connected E where V is connected (throw in enough colimits if you like). T has a component (as π0(T) is inhabited assuming Set is constructively well-pointed), say T0⊂T. Then f:T0→V induces an iso on applying π0, but there’s no immediately compelling reason I can see for it to be epi. Given that E is boolean, I can see that f is not not epi, since if it were not epi, V would not be connected since V would split as the image of T0 and its complement.
Presumably the topos of directed graphs would be good to consider. Hmmm…
In the last comment that should be reflexive directed graphs (see cohesive topos#Graphs, and indeed, π0 (there called Π0) does not reflect epis.
I’ve probably implicitly used the atomicity in the specific topos I have at hand :-S
The topos of (reflexive directed) graphs is not boolean, though?
Yeah, but that addresses my first question as to a counterexample I was seeking.
Returning the penultimate paragraph of #1, it seems you have assumed that π0 reflects initial objects. That is a weak form of Nullstellensatz, and I’m not sure it holds non-classically.
Here’s the classical proof: take a subcanonical site for the topos consisting of connected objects, which we may do because it is locally connected; then constant presheaves are sheaves, so π0 can be computed as the colimit of the underlying presheaf and we are done.
Of course, if you assume π0 reflects epimorphisms, then it also reflects initial objects: indeed, in a topos, any epimorphism whose codomain is an initial object must be an isomorphism.
Hmm, thanks. I’ve just taken π0 reflecting epis as a hypothesis now. It will hold in the final construction I do, so even if it is restrictive it makes the hypotheses cleaner.
1 to 6 of 6