1
$\begingroup$

Let $R$ be a Henselian discrete valuation ring with algebraically closed residue field $k$ ($R$ is not necessarily complete), $X$ a regular surface over $\mathrm{Spec}(R)$ and a sequence of locally free sheaves $\mathcal{F}_n$ on $X_n:=X \times_R \mathrm{Spec}(R/m^n)$, where $m$ is the maximal ideal of $R$. Does there exists a locally free sheaf $\mathcal{F}$ on $X$ such that its pull-back to $X_n$ for any $n$ is isomorphic to $\mathcal{F}_n$? We know that this is true if $R$ is complete.

$\endgroup$

1 Answer 1

5
$\begingroup$

No, that is not true, for essentially the same reason that I explained a couple of days ago in a different question. For simplicity, assume that $R$ contains its residue field $k$. Let $(X,x)$ be a smooth, projective curve over $k$ of genus $g>0$ together with a $k$-point, so that $\text{Pic}^0_{X/k}$ is a smooth, projective variety of positive dimension $g$, and there is a Poincaré sheaf $\mathcal{L}$ on $X\times_k \text{Pic}^0_{X/k}$ whose restriction to $\{x\}\times \text{Pic}^0_{X/k}$ is isomorphic to the structure sheaf.

Let $v:\text{Spec}(\widehat{R})\to \text{Pic}^0(X)$ be a $k$-morphism that does not factor through $\text{Spec}(\widehat{R}) \to \text{Spec}(R)$. For instance, let $k$ be characteristic $0$, let $u:\text{Pic}^0_{X/k} \to \mathbb{A}^g_k$ be a morphism that is unramified at some point $[L]$ mapping to $0$, let $R$ be the Henselization of $k[t]_{\langle t \rangle}$, and let $v$ be the morphism sending the closed point to $[L]$ and with $u\circ v$ sending $t$ to $(e^t-1, 0, \dots, 0)$. For such a morphism $v$, the pullback of $\mathcal{L}$ to $X\times_k \text{Spec}(\widehat{R})$ restricts on every $X\times_k \text{Spec}(R/\mathfrak{m}^n)$ to an invertible sheaf $\mathcal{L}_n$, and there is no algebraization over $\text{Spec}(R)\times_k X$.

$\endgroup$

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.