# Start a new discussion

## Not signed in

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

## Site Tag Cloud

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

• CommentRowNumber1.
• CommentTimeOct 20th 2014

Started the page cd-structure.

• CommentRowNumber2.
• CommentAuthorMarc Hoyois
• CommentTimeOct 21st 2014

A minor thing, but it’s not true in general that the (∞,1)-topos of sheaves for a cd-structure is hypercomplete. The (∞,1)-topos of a coherent topological space is always given by a cd-structure but it may not be hypercomplete (see counterexample 6.5.4.5 in HTT).

• CommentRowNumber3.
• CommentTimeOct 21st 2014
• (edited Oct 21st 2014)

Thanks for that correction. I guess I was misled by the fact that, for noetherian schemes, the Nisnevich topology is generated by a cd-structure, and the associated topos is hypercomplete, while both of these facts are false in general (right?).

Do you know any way to characterize the toposes that arise from cd-structures? I guess a necessary condition is generation under contractible colimits by the Yoneda embedding, right?

• CommentRowNumber4.
• CommentAuthorMarc Hoyois
• CommentTimeOct 21st 2014
• (edited Oct 21st 2014)

I don’t know any sufficient conditions. Another fairly strong necessary condition is that representables are compact in the (∞,1)-topos of sheaves.

The Nisnevich topology on non-noetherian schemes is always a confusing topic. Let me try to shed some light on it. Originally, Nisnevich defined the following pretopology: $\{f_i: X_i\to X\}$ is a covering if each $f_i$ is étale and if every point $x\in X$ has a preimage in some $X_i$ with the same residue field. Let’s call this topology $Nis_{bad}$. In DAG XI, Lurie has a slightly different definition of the Nisnevich topology on affine schemes, which can be extended to all schemes by throwing in the Zariski topology; let’s call this topology $Nis$. So we have

$Zar \leq Nis \leq Nis_{bad} \leq et.$

Now let’s fix a scheme $X$ and consider, for each of these topologies $\tau$, the (∞,1)-topos $X_\tau$ of $\tau$-sheaves on the category $Et_X$ of étale $X$-schemes (that’s a large category but nevertheless you get a topos). Then:

• $X_{et}\subset X_{Nis_{bad}} \subset X_{Nis} \subset X_{Zar}$
• If $X$ is locally noetherian, $X_{Nis_{bad}}=X_{Nis}$. In general they are different.
• If $X$ is qcqs, then $X_{Zar}$, $X_{Nis}$, and $X_{et}$ are coherent. In general, $X_{Nis_{bad}}$ isn’t.
• If $X=\lim_\alpha X^\alpha$ where the $X^\alpha$’s are qcqs, then $X_\tau=\lim_\alpha X^\alpha_\tau$ for $\tau=Zar,Nis,et$. Not so for $\tau=Nis_{bad}$.
• If $X$ is qcqs, then $X_{Zar}$ and $X_{Nis}$ can be described as presheaves on finitely presented étale $X$-schemes satisfying a Mayer-Vietoris condition. Not so for $X_{Nis_{bad}}$ or $X_{et}$.
• (Corollary) If $X$ is qcqs, then $X_{Zar}$ and $X_{Nis}$ are compactly generated (by finitely presented étale $X$-schemes). In general, $X_{Nis_{bad}}$ and $X_{et}$ aren’t.
• If $X$ is noetherian and of finite Krull dimension, then $X_{Zar}$ and $X_{Nis}=X_{Nis_{bad}}$ are hypercomplete.

For these reasons, $Nis$ is often more appropriate than $Nis_{bad}$. But $Nis_{bad}$ isn’t all that bad. For example, for any $X$, the (∞,1)-topos $X_{Nis_{bad}}^\wedge$ has a conservative family of points given by the henselizations of the local rings of étale $X$-schemes. For $X$ qcqs, $X_{Nis}^{\wedge}$ also has enough points by Lurie’s generalization of Deligne’s theorem, but I don’t know a nice description of the points.

• CommentRowNumber5.
• CommentAuthorMarc Hoyois
• CommentTimeOct 21st 2014

I changed a bit the definition of the associated topology, because what was defined before was not really a pretopology.

• CommentRowNumber6.
• CommentTimeOct 22nd 2014

• CommentRowNumber7.
• CommentAuthorUrs
• CommentTimeJun 26th 2022

have expanded the Idea-section a bit,

boosted the references a little,

cross-linked with the new entry Brown-Gersten property

made completely decomposable Grothendieck topology and variants redirect here

• CommentRowNumber8.
• CommentAuthorDavidRoberts
• CommentTimeJun 27th 2022
• (edited Jun 27th 2022)

The Zariski topology is given as an example, but not particularly explained. I’ll add a little bit

• CommentRowNumber9.
• CommentAuthorDavidRoberts
• CommentTimeJun 27th 2022

Added a description of the squares that define the Zariski topology.

I guess one should get other Grothendieck topologies defined by classes of finite open covers in the same fashion, but I haven’t added these.