4
$\begingroup$

Let $K$ be a non-archimedean field complete with respect to a discrete valuation with ring of integers $\mathcal{R}$, uniformizer $\pi$ and residue field $k$. Consider an affinoid analytic $K$-variety $X=Sp(A)$ with an affine formal model $\mathfrak{X}=Spf(A^{\circ})$ where $A^{\circ}\subset A$ is the set of power-bounded elements. Let $f\in A^{\circ}$ be an element such that its clas in $A^{\circ}/\pi$ is non-zero. As the generic fiber functor sends complete localizations on $\mathfrak{X}$ to Laurent subdomains in $X$, it follows that the affinoid subdomain $X(\frac{1}{f})=\{ x\in X \text{ such that } \vert f(x)\vert \geq 1 \}$ has an affine formal model given by the formal spectrum of $A^{\circ}\langle f^{-1}\rangle$. Let $\mathfrak{X}_{k}$ denote the special fiber of $\mathfrak{X}$. By definition, the special fiber associated to the formal scheme $\mathfrak{X}(\frac{1}{f})=Spf(A^{\circ}\langle f^{-1}\rangle)$ is a Zariski open of $\mathfrak{X}_{k}$. Hence, all properties of the special fiber $\mathfrak{X}_{k}$ which are local in the Zariski topology (irreducibility, smoothness etc) will also hold in $\mathfrak{X}(\frac{1}{f})_{k}$. My question is as follows: does this generalize to more general Laurent subdomains of $X$? For example, consider $f\in A^{\circ}$ as before, and consider subdomains of the form $Y=X(\frac{\pi^{n}}{f},\frac{f}{\pi^{n}})=\{ x\in X \text{ such that } \vert f(x)\vert = n \}$. As far as I know, these are not necessarily the generic fiber of an affine open subspace of $\mathfrak{X}$, so we cannot argue as above to construct affine formal models of $Y$ with topological properties similar to those of $\mathfrak{X}$. By Raynaud's theory we know there is an admissible formal blow-up $\mathfrak{X}'\rightarrow \mathfrak{X}$ such that there is an open $Z\subset \mathfrak{X}'$ such that its generic fiber is $Y$. However, admissible formal blow-up seems to have a weird behaviour at the topological level, so I don't know which properties of the topological space assocaited to $\mathfrak{X}$ are preserved under these kinds of maps. I would like to know if the fact that $X$ admits an affine formal model such that its special fiber has some topological property (mainly interested in irreducibility) implies that an affinoid subdomain of the form $X(\frac{\pi^{n}}{f},\frac{f}{\pi^{n}})$ also admits such an affine formal model. What about subspaces of the form $X(\frac{\pi^{n}}{f})$? Would something like this hold for Weierstrass subdomains of $X$?

$\endgroup$
3
  • $\begingroup$ I am confused by your terminology. For me, a formal model for a rigid analytic $K$-variety $X$ is a $p$-adic formal $\mathcal O_K$-scheme $\mathfrak X$ whose adic generic fiber is isomorphic to $X$, and this choice is not a priori necessarily unique. I am not sure how it is related to your concept. $\endgroup$
    – Z. M
    Dec 16, 2022 at 15:17
  • $\begingroup$ Dear @Z.M , yes this choice is not unique. I would like to know if in the case above there exist any admissible formal scheme over $Spf(\mathcal{R})$ such that its special fiber is integral and its generic fiber is $Y$. I do not claim that such model is unique, I just would like to know if there is one. $\endgroup$ Dec 16, 2022 at 16:12
  • $\begingroup$ In fact, if there is a formal model $\mathfrak{Y}$ of $Y$ which has integral special fiber but It is not affine It would also be enough. $\endgroup$ Dec 16, 2022 at 16:20

1 Answer 1

2
+200
$\begingroup$

For an affinoid $X=\mathrm{Sp}A$, the number of Shilov points of $X$ is a lower bound for the number of irreducible components of the special fiber of any formal model of $X$. This follows, for instance, from Proposition 2.2 of this paper.

So then it's easy to make concrete examples: $X=\mathrm{Sp}K \langle T \rangle$ has an obvious formal model with irreducible special fiber, but any formal model of the Laurent domain $U=\mathrm{Sp}K \langle T,\pi/T \rangle \subset X$ will have at least two irreducible components in its special fiber, since $U$ has two Shilov points.

Hope this helps!

$\endgroup$

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.