8
$\begingroup$

An object $X$ of a cartesian closed category $\mathbf C$ is atomic if $({-})^X \colon \mathbf C \to \mathbf C$ has a right adjoint (hence is also internally tiny). Intuitively, atomic objects are "very small" (as the name suggests), and consequently there aren't usually many tiny objects in $\mathbf C$.

However, is this necessarily the case? More precisely, do there exist any non-posetal cartesian closed categories in which every object is atomic?

If there are no non-posetal examples, are there any nontrivial posetal examples? (The terminal category forms a trivial example of a posetal example.)

$\endgroup$
9
  • $\begingroup$ Does it really imply tiny-ness ? It seems like this deduction uses that $*$ is tiny. $\endgroup$ Apr 27, 2022 at 11:15
  • 2
    $\begingroup$ I think we can show that there is no cocomplete category where every object is tiny. What follows is only an idea: when a cocomplete category K has a dense generator of tiny objects G, then K is Psh(G). Because of course, K is dense in itself, we would obtain that K is its own presheaf category. I think one can show that is is never the case. $\endgroup$ Apr 27, 2022 at 11:24
  • $\begingroup$ @MaximeRamzi: thanks, I have clarified that I meant "internally tiny"; I overlooked the distinction between the two notions. $\endgroup$
    – varkor
    Apr 27, 2022 at 12:12
  • $\begingroup$ @IvanDiLiberti : I think there are two issues with your sketch : 1- as I pointed out above, atomic need not imply tiny, 2- The equivalence "K has a dense generator of tiny objects G iff K = Psh(G)" holds only if G is small: you cannot define Psh(G) -> K by left Kan extension otherwise, because you did not assume K had large colimits (in other words, for your full sketch you need K to have K-sized colimits, i.e. K must be a poset) $\endgroup$ Apr 27, 2022 at 12:14
  • $\begingroup$ @MaximeRamzi do not forget the Wittgensteinian difference between to say and to show. $\endgroup$ Apr 27, 2022 at 12:18

2 Answers 2

11
$\begingroup$

Building on Maxime's answer -- if $C$ is cartesian closed and has an initial object $0$, and if $0$ is atomic (or even just tiny), then $C$ is the terminal category. For $1 = 0^0 = 0$ (the former equation holds in any cartesian closed category with an inital object $0$; the latter holds because $(-)^0$ preserves initial objects). That is, $C$ is pointed, and the only pointed cartesian closed category is the terminal one.

$\endgroup$
2
  • 4
    $\begingroup$ Thanks. I had spotted that having an initial object made the category pointed, but completely overlooked that this implied triviality. $\endgroup$
    – varkor
    Apr 27, 2022 at 18:48
  • $\begingroup$ Exactly the same thing happened for me :D that's a nice observation ! $\endgroup$ Apr 28, 2022 at 7:16
6
$\begingroup$

This is a partial answer : if $C$ has finite coproducts, then $C$ must be posetal. In fact, I only need biproducts of the form $X\coprod X$.

Indeed, because $C$ is cartesian closed, $X\times -$ commutes with these coproducts and in particular it is easy to check that $\underline{\hom}(* \coprod *, Y)\cong Y \times Y$, naturally in $Y$ ($\underline\hom$ denotes the internal hom).

In particular, if $*\coprod *$ is atomic, the canonical morphism $(Y_0\times Y_0)\coprod (Y_1\times Y_1)\to (Y_0\times Y_0)\coprod (Y_0\times Y_1)\coprod (Y_1\times Y_0)\coprod (Y_1\times Y_1)$ is an isomorphim (here I'm denoting by $Y_0$ or $Y_1$ the same object $Y$, it's simply to indicate the "position", i.e. what the morphism is).

This morphism is of the form $X\overset{in_1}\to X\coprod X$, and the claim is that this is an isomorphism. This implies that any two morphisms $X\to Z$ must be equal: $\hom(X,Z)\times \hom(X,Z)\cong \hom(X,Z)$.

Here $X$ is $Y\times Y$, for an arbitrary $Y$. Any of the two projections $Y\times Y\to Y$ is split, so it follows that any two morphisms out of $Y$ must be equal. $Y$ was arbitrary, so $C$ is posetal.

$\endgroup$
1
  • 1
    $\begingroup$ Thank you, this is already very strong evidence that atomocity of every object is an extremely restrictive assumption. In other words, every tiny object must be "superinitial" (assuming enough coproducts). $\endgroup$
    – varkor
    Apr 27, 2022 at 13:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.