Let $(X,\mathcal{O}_X)$ be a rigid $K$-space with a finite affinoid covering $(U_i)_{i\in I}$ such that any intersection of the $U_i$ is affinoid too.

Equipping the direct products with the maximum norm, the Čech complex associated to the covering is a complex of Banach $K$-vector spaces:

$0 \rightarrow \prod_{i\in I} \mathcal{O}_X(U_i) \rightarrow \prod_{i,j\in I}\mathcal{O}_X(U_i \cap U_j) \rightarrow \ldots$

Is it then in general true that the image of each boundary map is closed? If not are there necessary/sufficent conditions to get this property?

  • 1
    $\begingroup$ I am also very interested in this question. I asked a more general version here: mathoverflow.net/questions/338149/… $\endgroup$ Oct 8, 2019 at 13:33
  • $\begingroup$ I'd like to add that this true for $X$ beeing proper, but I'm interested in the case where proper is not needed. $\endgroup$
    – KKD
    Oct 9, 2019 at 9:01
  • $\begingroup$ Yes, in fact it suffices that $X$ is proper over some affinoid space, not necessarily $\mathrm{Sp}(K)$. $\endgroup$ Oct 9, 2019 at 10:51

1 Answer 1


This is false in general. Let $X$ be a smooth separated quasicompact rigid space over $\mathbf{Q}_p$. Let $\mathcal{O}^+$ be the sheaf of power-bounded functions. Then the following three statements are actually equivalent:

1) The torsion subgroup of $H^1(X,\mathcal{O}^+)$ is killed by a power of p.

2) For any covering $\mathcal{U}$ of $X$ by finitely many affinoids, the torsion subgroup of $\check{H}^{1}_{\mathcal{U}}(X,\mathcal{O}^+)$ is killed by a power of $p$.

3) For some covering $\mathcal{U}$ of $X$ by finitely many affinoids, the torsion subgroup of $\check{H}^{1}_{\mathcal{U}}(X,\mathcal{O}^+)$ is killed by a power of $p$.

This is a straightforward exercise in playing with the Cech-to-derived functor spectral sequence, using critically a theorem of Bartenwerfer that $H^1(U,\mathcal{O}^+)$ is killed by a power of $p$ for any smooth affinoid $U$.

On the other hand, for a fixed covering $\mathcal{U}$ of $X$ by finitely many affinoids and some fixed $n$, consider the following statements.

4) The torsion subgroup of $\check{H}^{n}_{\mathcal{U}}(X,\mathcal{O}^+)$ is killed by a power of $p$.

5) The image of the differential $d^{n-1}_{\mathcal{U}}$ is closed.

Then 5) implies 4), essentially by the open mapping theorem. (It might actually be true that 4) and 5) are equivalent, but I didn't think it through carefully.)

So to find an example where 5) fails (for $n=1$), we just need to find some smooth separated quasicompact $X$ where 1) fails, i.e. where $H^1(X,\mathcal{O}^+)$ contains torsion of arbitrarily high $p$-power order. Such spaces do exist, although not obviously so: de Jong found an example of such an $X$ last year (at my request).

Here is a sketch of de Jong's construction. Take an elliptic curve $E / \mathbf{Z}_p$ with good reduction. Choose a line bundle $L$ on $E$ which is non-torsion in the Picard group but which is trivial modulo $p$. Let $Y \to E$ be the total space of $L$, and let $\mathfrak{Y}$ be the formal completion of $Y$ along $(p)$. Finally, let $X$ be the rigid generic fiber of $\mathfrak{Y}$. Then $H^1(X,\mathcal{O}^+)$ contains torsion of arbitrarily high $p$-power order. (Idea: Reduce to the fact that $H^1(Y,\mathcal{O})=\oplus_{n \geq 0} H^1(E, L^{\otimes n})$ contains torsion of arbitrary $p$-power. Details omitted.)

  • $\begingroup$ Recenlty somebody told me something similar. Thats why I am back to your answer. So I got the impression that if the space is maybe not to special it could be true. Are there not to strong conditions on $X$ which guarantee the strictness resp. finite p-torsion? I think of $X$ being an admissible open subset in the rigid analytification of a partial flag variety with the property above. $\endgroup$
    – KKD
    Nov 26, 2020 at 11:38
  • $\begingroup$ After thinking again about your answer, I got unsure about your notation and then about your statements. First of all to which Čech complex complex belongs $d^{n-1}_{\mathcal{U}}$ to? How is the open mapping theorem used to say something about the torsion part? But in general you argue with the sheaf of power bounded functions, whereas I consider the structure sheaf. What is the link here? $\endgroup$
    – KKD
    Jun 1, 2021 at 10:44

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.