7
$\begingroup$

Let $f : X \rightarrow Y$ be a finite type morphism of Noetherian schemes. The valuative criterion for properness runs as follows. Suppose that for any DVR $R$ with fraction field $K$ that any $K$-valued point of $X$ lying above an $R$-valued point of $Y$ extends uniquely to an $R$-valued point of $X$. Then $f$ is proper.

Does it suffice to check instead that any $K$-valued point of $X$ lying above an $R$-valued point of $Y$ extends to an $R'$-valued point of $X$, where $R'$ is the integral closure of $R$ in a finite extension $K'$ of $K$?

$\endgroup$

1 Answer 1

7
$\begingroup$

Yes.

I claim that, for any $K'$ point of $X$, if this point extends both to an $R'$ point and an $K$ point, then it extends to a $R$ point. This obviously proves the result.

Let $x'$ be the closed point of $\mathrm{Spec}(R')$. Let $\mathrm{Spec}(A)$ be an affine neighborhood of the image of $x'$. Then the $R'$ point must factor through $\mathrm{Spec}(A)$. The $K$ point also factors through $\mathrm{Spec}(A)$, as it has the same image as the $K'$ point.

We can now transform the question into algebra. The algebraic statement is that we have a map $A \to K'$ which factors through both $K$ and $R'$. But then the image of this map must lie in $K \cap R' = R$. So the map factors through $\mathrm{Spec}(R)$, and our $K$ point extends to an $R$ point.

$\endgroup$
2
  • 1
    $\begingroup$ Should be "extends to an R point" in the 2nd paragraph, I think. $\endgroup$
    – TSG
    Nov 8, 2009 at 22:52
  • $\begingroup$ A problem with this argument is that R' is not local in general. Of course this will hold if R is henselian. In general, to make the argument work one probably ought to make an extra hypothesis to ensure that any finite set of points of a fibre of f is contained in an open subset (e.g. that f is quasiprojective). Or one can reduce to the henselian case by the same argument, and make no hypothesis. $\endgroup$
    – Bruno Kahn
    Jun 28, 2022 at 15:46

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.