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 11 of 11
For an (∞,1)-topos to be cohesive, we require that Δ have a left adjoint ʃ that preserves products. However, even if no such adjoint exists, we can still define the “shape” of an object X∈H to be the hom-functor H(X,Δ(−)):∞Gpd→∞Gpd, which is left-exact and accessible. (This is just the shape of the over-topos H/X.) Is there a condition on H under which this non-representable shape preserves products? That is, such that H(X,Δ(−))×H(Y,Δ(−))≃H(X×Y,Δ(−))? Here the product on the left should probably be taken in the category LexAcc(∞Gpd,∞Gpd), since that’s what’s true if Δ does have a left adjoint preserving products.
I don’t think that’s formulated right. Even if ∫ exists, I don’t think we expect that H(∫X,−)×H(∫Y,−)=H(∫(X×Y),−), do we?
This is probably not quite what you’re looking for, but ∫:∞Top→∞Grpd preserves finite products of proper ∞-topoi. More generally, if X is proper then ∫(X×Y)≃∫Y∘∫X for any Y, by proper base change, and ∫X preserves filtered colimits. If F,G∈LexAcc, there’s always a canonical map F×G→F∘G, and it’s an equivalence if F preserves filtered colimits.
@Charles: Oh, oops, I meant the product should be taken in LexAcc(∞Gpd,∞Gpd)op. That’s necessary to get the variance right, and I believe the dual Yoneda embedding landing in there does preserve products, just like the ordinary Yoneda embedding landing in a category of colimit-preserving functions preserves those colimits.
@Marc: What do you mean by ʃ:∞Top→∞Gpd? Only locally contractible ∞-topoi have a representable shape, right?
My bad, I meant ∫:∞Top→LexAcc(∞Gpd,∞Gpd)op.
Ah, okay. That seems at least related.
Perhaps asking whether ʃ:H→LexAccop preserves products isn’t quite the right question. When H is cohesive, we have ʃH=1, so that the product ʃX×ʃY is equivalently the pullback ʃX×ʃ1ʃY, i.e. the pullback ʃ(H/X)×ʃHʃ(H/Y). So maybe in general the place to start is by asking what pullbacks ʃ:Top→LexAccop preserves, and then in what cases (H/X)×H(H/Y) is such, and then perhaps add the condition that H is connected (so that ʃH=1).
By the way, let me remind everyone that ʃ is not an integral sign but an esh, ʃ
I realized there is another condition that make sense for arbitrary (connected) H and is equivalent to ʃ preserving products if H is locally connected: namely, that the discrete objects are an exponential ideal. It seems natural to conjecture that this is equivalent to the non-representable shape preserving products, and I can see part of one direction of a proof, but I’m stuck at
If F,G∈LexAcc, there’s always a canonical map F×G→F∘G, and it’s an equivalence if F preserves filtered colimits.
What is this canonical map, and why is it an equivalence if F preserves filtered colimits?
Oh, oops – if the discrete objects are an exponential ideal, then in particular they are closed under exponentials, and hence under arbitrary products; thus Δ is continuous, so by the adjoint functor theorem it has a left adjoint. So that condition is actually no weaker.
@Mike Suppose F is co-represented by the cofiltered diagram (Xi), and similarly G by (Yj). Then F×G is co-represented by (Xi×Yj). So we have a map
(F×G)(U)=colimi,jMap(Xi×Yj,U)=colimi,jMap(Xi,Map(Yj,U))→colimiMap(Xi,colimjMap(Yj,U))=(F∘G)(U).The arrow can also be written as colimjF(Map(Yj,U))→F(colimjMap(Yj,U)).
Ah, but of course the map goes the other way in Pro(∞Gpd).
Ok, thanks.
1 to 11 of 11