An inclusion of sites f: D→C is dense if it induces an equivalence between the categories of sheaves on C and D. Likewise, f is ∞-dense is it induces an equivalence between the ∞-categories of ∞-sheaves on C and D. An ∞-dense inclusion is dense but the opposite is false in general, see Counterexample in Lurie's Higher Topos Theory.

A nontrivial example of an ∞-dense inclusion of sites is given by the inclusion Cart→Man of cartesian spaces into all smooth manifolds, as explained in Is the site of (smooth) manifolds hypercomplete?

As it turns out, in the 1-categorical case we have an ample supply of dense inclusions of sites, as explained by Proposition C2.2.16 of Johnstone's Sketches of an Elephant: essentially small sites of definition of a given Grothendieck topos E that are subcanonical can be identified with full subcategories of E whose objects form a separating family, with the Grothedieck topology induced from the canonical topology on E.

Every inclusion of such sites is dense, but not necessarily ∞-dense.

I wonder if there is a (practical) criterion that allows us to check whether a dense inclusion is also ∞-dense. We can assume the source site D to be hypercomplete, if necessary.

The example that I have in mind takes D to be the site of smooth manifolds (or cartesian spaces) and C a subcategory of the category of sheaves of sets on smooth manifolds that contains some infinite-dimensional smooth manifolds (e.g., smooth mapping spaces Map(M,N) between finite-dimensional manifolds).

What additional conditions are needed to ensure that a dense inclusion of sites is ∞-dense?

  • $\begingroup$ Can you be more specific about what kind of subcategory $C$ is allowed to be? Does it contain at least one $n$-manifold for all $n$ for example? $\endgroup$ – David Carchedi Nov 13 '13 at 18:37
  • $\begingroup$ @DavidCarchedi: In the examples I have in mind C is supposed to be some category of infinite-dimensional manifolds, so in particular it contains all finite-dimensional manifolds. $\endgroup$ – Dmitri Pavlov Nov 14 '13 at 12:09
  • $\begingroup$ In that case, as long as your infinite-dimensional manifolds are locally modeled on "convenient vector spaces" (e.g. Frechet spaces), then $f$ is infinity dense. I'm a bit busy this moment, but I will try to type up the proof soon. $\endgroup$ – David Carchedi Nov 14 '13 at 17:12
  • $\begingroup$ @DavidCarchedi: Ok, thanks a lot for letting me know this. I'm eagerly waiting for the details. $\endgroup$ – Dmitri Pavlov Nov 14 '13 at 17:38
  • $\begingroup$ @DmitriPavlov : I don't quite see how the counterexample you quote from HTT is a counter example of this. To me it is the typical counter example of a non-Hypercomplete infinity topos of sheaves. I don't see the connection with what you are talking about. $\endgroup$ – Simon Henry Nov 19 '18 at 9:37

If you content yourself with the hypercomplete case, a sufficient practical criterion is given in my paper joint with Tony Yue Yu, that you can find here: http://arxiv.org/abs/1412.5166. The result we are concerned with is lemma 2.35. It is the analogue of the well known result in algebraic geometry, that can be found here: http://stacks.math.columbia.edu/tag/03A0.


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.