8
$\begingroup$

This is something which I'm sure is well known to experts which I would appreciate some information about. In his paper [1], Scholze proves (e.g. Theorem 8.4, Theorem 8.8) that on a proper adic space $X$ over $\textrm{Spa}(k,\mathcal{O}_{k})$, with $k$ an appropriate non-archimedean field, the natural map $H^{i}(X_{\textrm{pet}}, \hat{\mathbb{Z}}_{p}) \otimes B_{\textrm{dR}} \to H^{i}(X_{\textrm{pet}}, \mathbb{B}^{+}_{\textrm{dR}})$ is an isomorphism, where I use the notation "pet" for the pro-etale site. I think a similar statement should be true for $X$ the adic space associated to a smooth algebraic variety over $k$, but I cannot find the appropriate reference.

I am aware of work [2] of Diao, Lan, Liu and Zhu which works with a general smooth variety, but I am really interested in the "intermediate" comparison between cohomology wich coefficients in $\hat{\mathbb{Z}}_{p}$ and $\mathbb{B}^{+}_{\textrm{dR}}$ rather than the actual comparison isomorphism of $p$-adic Hodge theory, and it seems to me they only handle this intermediate fact in the proper case. The related work [3] seems to do something like what I want in Proposition 3.3.4, but in a compactly supported setting which isn't really what I'm interested in.

If it helps I am really just looking to understand the maps $H^{i}(X_{\textrm{pet}}, \hat{\mathbb{Z}}_{p}) \otimes B_{\textrm{dR}} \to H^{i}(X_{\textrm{pet}}, \mathbb{B}^{+}_{\textrm{dR}})$ when $X = \mathbb{G}_{m}^{w}$ is the $w$-fold product of the multiplicative group, so an explicit description of this map would also be appreciated.

Edit: I may have found what I want in Theorem 7.14 of [4]; I will take a look at the paper and report back.

  1. https://www.math.uni-bonn.de/people/scholze/pAdicHodgeTheory.pdf
  2. https://arxiv.org/pdf/1803.05786.pdf
  3. https://arxiv.org/pdf/1912.13030.pdf
  4. https://arxiv.org/pdf/1801.01779.pdf
$\endgroup$

1 Answer 1

5
$\begingroup$

The result is false in the open case.

If true, a long exact sequence would show that also

$$H^i(X_{\mathrm{proet}},\hat{\mathbb Z}_p)\otimes k\to H^i(X_{\mathrm{proet}},\hat{\mathcal O}_X)$$

is an isomorphism, where I assume $k$ is algebraically closed (which I think is also implicit in the question, or otherwise one should base change $X$ to an algebraic closure before taking cohomology). But this can't even be true for $i=0$ in general, as the right-hand side receives a map from $H^0(X,\mathcal O_X)$ -- if $X$ is affine, this is infinite-dimensional.

One key issue is that if $X$ is open, it is not quasicompact as a rigid space, and hence taking cohomology does not commute with filtered colimits. Thus, it is not even clear that the map

$$ H^i(X_{\mathrm{proet}},\hat{\mathbb Z}_p)\otimes \mathbb Q_p\to H^i(X_{\mathrm{proet}},\hat{\mathbb Q}_p)$$

is an isomorphism. In fact, it is not an isomorphism even for $X=\mathbb A^1$ and $i=1$, where the left-hand side vanishes but right-hand side is infinite-dimensional, see Corollary 3.10 of Le Bras' Thesis.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.