3
$\begingroup$

Let $f: X \to Y$ a finite morphism of schemes (therefore $f$ affine and the quasi koherent $\mathcal{O}_Y$-algebra $\mathcal{A}= f_*(\mathcal{O}_X)$ such that $X = Spec(\mathcal{A})$ is a finite $\mathcal{O}_Y$-algebra)

Therefore practically that means that for every affine open set $V = Spec(S) \subset Y$ the $S=\Gamma(V,\mathcal{O}_Y)$-algebra $A= \Gamma(V,\mathcal{A})$ is finite

I want to show that then $f$ is proper.

Here my attempts:

I want to use following valuative criterion for properness:

$f$ is proper (so separated, of finite type, and universally closed) if and only if in every diagram

$$ \require{AMScd} \begin{CD} Spec(F) @>{g} >> X \\ @VViV @VVfV \\ Spec(R) @>{t}>> Y \end{CD} $$

for every discrete valuation ring $R$ and $F = Frac(R)$ there exist $l:Spec(R) \to X$ such that $a = l \circ i$ and $t = f \circ l$ holds.

The spectrum of DVR $R$ has the structure $Spec(R) = \{\sigma, \eta\} = \{\sigma\} \cup Spec(F)$ with generic point $ \eta$ and unique maximal ideal $\sigma$.

Since $Spec(F)$ is a sigleton, we can reduce the problem to affine case $X= Spec(A), Y = Spec(S)$ with $A$ finite $S$-algebra.

For underlying topological spaces it's obviously how to define $l: Spec(R) \to X = Spec(A)$:

Let $s_0:= t(\eta), s_1 := t(\sigma)$. Obviously $s_0 \subset s_1$ as prime ideals. Because $= s_0= t \circ i (\eta) = f \circ g(\eta)$, we conclude that $a_0:= g(\eta)$ is lying over $s_0$. Since $f$ is finite morphism the going up theorem holds, therefore there exist a $a_1$ lying over $s_1$.

Define $l$ set theoretically by setting $l(\eta)= a_0, l(\sigma) := a_1$.

Now to my problem: How to see that $l$ defined in this way provides a morphism of schemes, not only a map which commutates with the commutative square?

$\endgroup$
  • 3
    $\begingroup$ You should show that a finite map is separated (every affine map trivially is), of finite type (trivial) and universally closed (finite maps are closed by going-up and they are universally closed since finite is preservered under base change). $\endgroup$ – MooS Mar 4 '18 at 20:38
  • $\begingroup$ To put it more concretely: Showing that a finite morphism is proper by using the valuative criterion will basically just reproof the direction 'proper --> valuative criterion holds' of the valuative criterion. $\endgroup$ – MooS Mar 4 '18 at 21:01
2
$\begingroup$

As you have remarked, it is sufficient to look at the case where $X = \operatorname{Spec} A$ and $Y = \operatorname{Spec} S$ are both affine. In general, when defining morphisms between affine schemes, it is easier to define the corresponding map between the rings. So instead of trying to define construct directly a scheme map from $\operatorname{Spec} R$ to $X$, it is easier to construct a ring map from $A$ to $R$. This is the same thing, since we have a canonical isomorphism $\operatorname{Hom}_{Schemes}(\operatorname{Spec} R, X) = \operatorname{Hom}_{Rings}(A, R)$.

Turning the valuative criterion that you quoted into commutative algebra, we see that what you want to show is equivalent to the following proposition.

Proposition Let $A$ be a finite $S$-algebra. Then for any discrete valuation ring $R$ with fraction field $F$ and any a commutative diagram $$ \require{AMScd} \begin{CD} S @>{t} >> R \\ @VVfV @VViV \\ A @>{g}>> F \end{CD} $$ there is a ring map $A \to R$ that fits in the diagram.

Of course, we usually think of $R$ as a subring of $F$, so saying that there is a morphism $A \to R$ as in the diagram really just means that the map $g$ has image contained in $R \subset F$.

To prove the above proposition, we just note that the image $g(A)$ of $A$ will be finite and therefore integral over $R$. So $g(A)$ is contained in the integral closure of $R$ in $F$. But $R$ is a discrete valuation ring, and such rings are integrally closed in their fraction field. This gives us $g(A) \subset R$ as needed.

$\endgroup$

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.