Processing math: 100%
Not signed in (Sign In)

Not signed in

Want to take part in these discussions? Sign in if you have an account, or apply for one below

  • Sign in using OpenID

Discussion Tag Cloud

Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.

Welcome to nForum
If you want to take part in these discussions either sign in now (if you have an account), apply for one now (if you don't).
    • CommentRowNumber1.
    • CommentAuthoradeelkh
    • CommentTimeNov 19th 2014

    What is the definition of an infinity-stack on an infinity-site with values in an arbitrary infinity-category (bicomplete, say)? The definition I know works only for stacks of infinity-groupoids.

    • CommentRowNumber2.
    • CommentAuthoradeelkh
    • CommentTimeNov 19th 2014
    • (edited Nov 19th 2014)

    Maybe this: F:CopD is a stack/sheaf iff Hom(d,F()):CopSpc is a sheaf for all objects dD?

    • CommentRowNumber3.
    • CommentAuthorZhen Lin
    • CommentTimeNov 19th 2014
    • (edited Nov 19th 2014)

    That’s reasonable, at least for categories of structures “defined by finite (homotopy) limits”.

    • CommentRowNumber4.
    • CommentAuthoradeelkh
    • CommentTimeNov 19th 2014
    • (edited Nov 19th 2014)

    Zhen Lin: what does that mean? (At the moment I’m interested in stacks of infinity-categories.)

    • CommentRowNumber5.
    • CommentAuthorDylan Wilson
    • CommentTimeNov 19th 2014

    DAGV says, for X a topos, a C-valued sheaf is a limit preserving functor X^op —> C, for any arbitrary C.

    • CommentRowNumber6.
    • CommentAuthoradeelkh
    • CommentTimeNov 19th 2014
    • (edited Nov 19th 2014)

    Dylan: that definition doesn’t take into account the topology on the site though. Oh wait, X is already a topos. I’m confused now…

    • CommentRowNumber7.
    • CommentAuthorZhen Lin
    • CommentTimeNov 19th 2014
    • (edited Nov 19th 2014)

    Well, going back to ordinary examples, we know that a “sheaf of local rings” is not actually defined to be a sheaf taking values in local rings in this sense. But otherwise the obvious thing works well for many examples: groups, rings, etc. These are structures “defined by finite limits”. See sketch. I expect the same holds true for “homotopical” structures.

    Dylan: that definition doesn’t take into account the topology on the site though. Oh wait, X is already a topos. I’m confused now…

    It’s secretly the same as your definition. The idea is that every limit preserving functor 𝒳op𝒮 is representable.

    • CommentRowNumber8.
    • CommentAuthoradeelkh
    • CommentTimeNov 19th 2014

    Right, of course, thanks to both of you. Perhaps this should go on the page infinity-stack.

    • CommentRowNumber9.
    • CommentAuthorMarc Hoyois
    • CommentTimeNov 19th 2014

    Each definition has an advantage over the other though. On the one hand, one can speak of D-valued sheaves on (∞,1)-topoi that don’t arise from a site. On the other hand, one can have D-valued sheaves on C that do not lift to Shv(C) (if D doesn’t have enough limits).

    • CommentRowNumber10.
    • CommentAuthorUrs
    • CommentTimeNov 19th 2014

    The definition recalled in #5 is to be thought of as giving sheaves (stacks) on the topos regarded as a big site equipped with the canonical topology.

    • CommentRowNumber11.
    • CommentAuthorMarc Hoyois
    • CommentTimeNov 20th 2014

    @Urs: one might think about it this way, but it doesn’t actually work for a general (∞,1)-topos.

    • CommentRowNumber12.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014

    Sorry, say again: what doesn’t work?

    • CommentRowNumber13.
    • CommentAuthorMarc Hoyois
    • CommentTimeNov 20th 2014

    The Yoneda embedding HShvcan(H) is only an equivalence if H is a sheaf (∞,1)-topos.

    • CommentRowNumber14.
    • CommentAuthorUrs
    • CommentTimeNov 20th 2014

    Where by “sheaf -topos” you mean topological localization of presheaves, I suppose.

    I’d still think that for #6 this HShcan(H) clarifies what’s going on, conceptually. Hypercompletion may probably be brought into the picture nicely somehow?

    • CommentRowNumber15.
    • CommentAuthorMarc Hoyois
    • CommentTimeNov 20th 2014

    I was wondering about this also, and I’m not sure. That’s what the notion of “hypertopology” on a category C would accomplish, I guess.

    • CommentRowNumber16.
    • CommentAuthorZhen Lin
    • CommentTimeDec 11th 2014

    So, what’s the conclusion? Is it true that a functor 𝒳op𝒮 is representable if and only if it is a sheaf for the canonical topology? This MO question remains unresolved.

    • CommentRowNumber17.
    • CommentAuthorMarc Hoyois
    • CommentTimeDec 12th 2014

    Is this a correct argument? An (∞,1)-topos H is the union of its subcategories of κ-compact objects for all cardinals κ such that H is κ-accessible and closed under finite limits:

    H=colimκHκ.

    So

    Fun(H,Grpdop)=limκFun(Hκ,Grpdop).

    A presheaf on H is a sheaf iff its restrictions to all Hκ’s are sheaves, since every Cech nerve in H is in Hκ for some κ. Hence

    Sh(H)=limκSh(Hκ)=H.
    • CommentRowNumber18.
    • CommentAuthorZhen Lin
    • CommentTimeDec 12th 2014

    That seems correct, modulo the question of whether a sheaf for the canonical topology on Hκ necessarily satisfies descent for Čech nerves in Hκ. But that’s just a question of what the definition of canonical topology is, I guess.

    • CommentRowNumber19.
    • CommentAuthorMarc Hoyois
    • CommentTimeDec 12th 2014

    Right, I was thinking of the topology on Hκ induced by epimorphic covers (in the topos H), so the canonical topology in the sense of HTT. Actually, do we even know that Sh(Hκ)=H?

    • CommentRowNumber20.
    • CommentAuthorZhen Lin
    • CommentTimeDec 12th 2014

    Oh, sorry. I think I had some wires crossed. Yes, it would suffice to check that Sh(Hκ)H. In view of the fact that every limit-preserving functor HopGrpd is representable, it suffices to verify that the right Kan extension of a sheaf (Hκ)opGrpd along the inclusion HκH preserves limits. By accessibility, it should be enough to verify that a sheaf (Hκ)opGrpd preserves limits for κ-small diagrams. But is this really any easier than the question we started with?

    • CommentRowNumber21.
    • CommentAuthorMarc Hoyois
    • CommentTimeDec 13th 2014
    • (edited Dec 13th 2014)

    It doesn’t seem any easier, no. Let me list some facts which I think we know for sure:

    • if you factor the inclusion HPSh(Hκ) as a cotopological localization followed by a topological one, then the topological one is exactly Sh(Hκ), by definition of the canonical topology. Hence H is a cotopological localization of Sh(Hκ).
    • By HTT 6.2.4.6, that the inclusion HSh(Hκ) preserves effective epimorphisms.
    • If H=Sh(C) where C has finite limits, the restriction along CHκ is a geometric morphism Sh(Hκ)H, and the composition
    HSh(Hκ)H

    is the identity. So H is even a retract of Sh(Hκ).