7
$\begingroup$

It is well known that the codomain functor $$cod:\mathcal{C}^\to\to\mathcal{C}$$ from the arrow category of a category $\mathcal{C}$ to itself is a fibration iff $\mathcal{C}$ has binary pullbacks.

Is there another functor that always exists for which promotion to a fibration is equivalent to $\mathcal{C}$ having a terminal object?

I naively tried $!:\mathcal{C}\to{\bf 1}$ and $\{X\}:{\bf 1}\to\mathcal{C}$ without any luck.

This would be of interest for the obvious reason -- having binary pullbacks and a terminal object is equivalent to being finitely complete, so a functor answering the above question positively would allow for a 'fully fibrational' characterization of finite completeness. Further, the construction below yields a functor that always exists such that promotion to a fibration is equivalent to having all pullbacks -- this together with Alexanders answer below offers a fully fibrational characterization of completeness, as originally desired.


Thanks to Alexander and Andrej for pointing out in the comments that $cod$ as above being a fibration only yields binary pullbacks, not arbitrary ones.

EDIT: I was able to fix the following construction to actually yield a functor that always exists such that promoting this functor to a fibration is equivalent to $\mathcal{C}$ having all pullbacks. (The first correction worked; we want them all to be the same commutative square universally, so that all paths from the pullback object into the object we're pulling back over are equal.)

For a category $\mathcal{C}$, consider the category $Sink(\mathcal{C})$ whose objects are sinks in $\mathcal{C}$ indexed over ordinals $\alpha\in{\bf O_n}$ $$\{f_i:X_i\to X\}_{i<\alpha}$$ and whose arrows are only defined from sinks indexed over $1$ to arbitrary sinks $$\hat h:\{f:A\to B\}\to\{g_i:C_i\to D\}_{i<\alpha}$$ given by ordered pairs $$\hat h=\big(\{h_i:A\to C_i\}_{i<\alpha},h:B\to D\big)$$ such that all the obvious squares commute, so $$h\circ f=g_i\circ h_i$$ for all $i<\alpha$. We have an obvious functor $$cod:Sink(\mathcal{C})\to\mathcal{C}$$ $$\{f_i:X_i\to X\}_{i<\alpha}\mapsto X,$$ $$\hat h\mapsto h$$ and this functor is a fibration iff $\mathcal{C}$ has all pullbacks.

We can identify $\mathcal{C}^\to$ with the subcategory of $Sink(\mathcal{C})$ indexed over $1$.

$\endgroup$
6
  • 1
    $\begingroup$ N.B. A category has pullbacks and a terminal object iff it is finitely complete. $\endgroup$ Feb 1, 2022 at 7:46
  • $\begingroup$ @AlexanderCampbell A category is complete iff it has all products and equalizers iff it has all pullbacks and a terminal object. For finite completeness we would only require finite pullbacks and a terminal object. $\endgroup$
    – Alec Rhea
    Feb 1, 2022 at 8:10
  • 4
    $\begingroup$ @AlecRhea: You should have an opinion poll that finds out how many people think that "pullback" means the limit of $\bullet \to \bullet \leftarrow \bullet$. $\endgroup$ Feb 1, 2022 at 9:05
  • $\begingroup$ @AlexanderCampbell Ah, I see the confusion thanks to Andrej -- when I say that a category 'has pullbacks', I mean arbitrary ones not binary ones. (limits of diagrams shaped like arbitrarily many points each with a unique morphism to some specific point) $\endgroup$
    – Alec Rhea
    Feb 1, 2022 at 9:20
  • $\begingroup$ But now I'm wondering if the fibrational characterization of pullbacks only holds for binary ones; I think I need to be more careful here. $\endgroup$
    – Alec Rhea
    Feb 1, 2022 at 9:21

1 Answer 1

10
$\begingroup$

A category $C$ has a terminal object if and only if the canonical functor $C \star \mathbf{1} \to \mathbf{1} \star \mathbf{1} = \mathbf{2}$ is a fibration.

$\endgroup$
3
  • 3
    $\begingroup$ It's funny how $C \mapsto C \star \mathbf 1$ is more of a "colimit-y" construction whereas $C \mapsto C^{\mathbf 2}$ from the motivating example is more of a "limit-y" construction. $\endgroup$
    – Tim Campion
    Feb 1, 2022 at 23:55
  • $\begingroup$ @TimCampion Seeing as $\star$ here is a coproduct (colimit) and $C^2\cong C\times C$ is a product (limit), that doesn't strike me as odd. $\endgroup$
    – Alec Rhea
    Feb 4, 2022 at 5:25
  • $\begingroup$ @AlecRhea By $C^{\mathbf 2}$ I meant the morphism category of $C$, not the product of $C$ with itself. This is indeed a $Cat$-weighted limit, though. I'm not sure how $\star$ is a colimit. The more precise thing to say is that $C \mapsto C \star 1$ commutes with connected colimits and few limits, whereas $C \mapsto C^{\mathbf 2}$ commutes with limits and few colimits. I find it odd because these two constructions are playing analogous roles in the story here -- $C \star 1$ is to terminal objects as $C^{\mathbf 2}$ is to pullbacks. $\endgroup$
    – Tim Campion
    Feb 4, 2022 at 11:17

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.