Let $B$ be a noetherian integral domain and consider the polynomial ring $B[t]$. Let $\mathfrak{p}$ be a prime ideal in $B[t]$ of height $1$ such that $\mathfrak{p} \cap B = (0)$. Show that $B[t]_{\mathfrak{p}}$ is just $K[t]_{\mathfrak{m}}$ where $K$ is the field of fractions of $B$ and $\mathfrak{m}$ is some maximal ideal of $K[t]$.

Note this isn't a homework problem, it's actually a step in a proof in Hartshorne that I got stuck on, showing that the preimage of the generic point under the fibered product is a regular codimension $1$ point. For reference it's Proposition II.6.6.

It seems like this should be straight forward, but I'm having trouble making it rigourous. We have $B[t]_{\mathfrak{p}}$, and since $\mathfrak{p}$ doesn't contain any non-zero element in $B$, then $B[t]_{\mathfrak{p}}$ must at least contain $K[t]$. If $B$ was a UFD then it would be easy, since we could say that $\mathfrak{p}$ is principal. But say $\mathfrak{p}$ is generated by $\{ f_{1}, f_{2}, \ldots , f_{r} \}$. I can "see" intuitively that the result should be $K[t]_{\langle f_{1}, \ldots, f_{r}\rangle}$, and of course $\langle f_{1}, \ldots, f_{r} \rangle$ would be principal in $K[t]$. But I can't seem to be able to make this rigorous, which is worrying to me because I feel like this should be extremely straight forward. Any advice?

  • $\begingroup$ Since it's just a step in a proof, are you sure all the hypotheses are necessary ? I seem to have a proof that doesn't use the height of $\mathfrak{p}$ $\endgroup$ – Maxime Ramzi May 5 '18 at 12:30
  • $\begingroup$ No I'm not sure that hypothesis is necessary for this particular claim. $\endgroup$ – Luke May 5 '18 at 13:19

Let me give it a try: let $\mathfrak{m}$ be a maximal ideal of $K[t]$ containing $\mathfrak{p}$ ($\mathfrak{p}K[t]$ contains non nonzero constant element because $\mathfrak{p}$ doesn't, so it's a strict ideal: such an $\mathfrak{m}$ exists).

The idea of my proof is to show that $K[t]_\mathfrak{m}$ has the universal property of $B[t]_\mathfrak{p}$, which will allow us to prove the isomorphism between the two (and it's easy to check that it's "the right isomorphism")

Consider a morphism $f: B[t]\to C$, $C$ a ring such that for $x\notin \mathfrak{p}$, $f(x) \in C^\times$.

Compose with the inclusion $i: B\to B[t]$ this yields a morphism $f\circ i: B\to C$, and your hypothesis concerning $\mathfrak{p}\cap B$ + the hypothesis concerning $f$ yields that this extends to $g: K\to C$.

By the universal property of $K[t]$, sending $K$ to $C$ with $g$ and $t$ to $f(t)$ yields a morphism $s: K[t] \to C$, which corresponds to the following commutative diagram :

$\require{AMScd} \begin{CD} B @>>> K;\\ @VV^iV @VVV \\ B[t] @>>> K[t];\\ @VV^fV @VV^sV \\ C@>>^{id_C}>C \end{CD}$

Now let $A \in K[t], A\notin \mathfrak{m}$. For some $q\in B$, $qA\in B[t]$. $s(qA) = f(qA)$. Now if $qA \in \mathfrak{p}$ then $qA \in \mathfrak{m}$ hence (by maximality) $q\in \mathfrak{m}$ or $A\in \mathfrak{m}$. The second option is impossible by hypothesis, but so is the first since $q\in K^\times$. Hence $qA\notin \mathfrak{p}$ and so $f(qA)\in C^\times$. Hence $s(q)s(A)\in C^\times$, and so $s(A) \in C^\times$.

By the universal property of the localization at $\mathfrak{m}$, we get a morphism $e:K[t]_{\mathfrak{m}}\to C$ making the whole thing commute:

$\require{AMScd} \begin{CD} B @>>> K;\\ @VV^iV @VVV \\ B[t] @>>> K[t] @>>> K[t]_{\mathfrak{m}};\\ @VV^fV @VV^sV @VV^eV \\ C@>>^{id_C}>C @>>^{id_C}> C \end{CD}$

So apart from uniqueness which is left to prove, $K[t]_{\mathfrak{m}}$ satisfies the universal property of $B[t]_{\mathfrak{p}}$, so it is isomorphic to it.

Let's now prove the uniqueness part: assume the morphism $l: K[t]_{\mathfrak{m}}$ makes the traingle commute:

$\require{AMScd} \begin{CD} B[t] @>>> K[t]_{\mathfrak{m}};\\ @VV^fV @VV^lV \\ C@>>^{id_C}>C \end{CD}$

Then the induced map $K\to C$ is necessarily obtained by the universal property of $K$ with respect to $B$, because of the uniqueness there, similarly for the map $K[t]\to C$, and so the map $l:K[t]_\mathfrak{m}\to C$ is the one that is induced by the universal property with respect to $K[t]$ , that is $l=e$.

So uniqueness follows, and we have our proof.

(If anyone knows how to make diagonal arrows instead of putting $id_C$'s everywhere, please feel free to edit)


Here's a proof with a more straightforward flavor (and no universal property reasoning). These very basic ideas come up again and again when working with polynomial domains and so-called uppers to zero, and once they're digested i feel the statement in question does become something of the one-liner-no-brainer that you hoped it would.

The only assumptions we need are that $B$ is an integral domain and $\mathfrak{p}$ is a nonzero prime ideal of $B[t]$ such that $\mathfrak{p} \cap B = 0$. (We could deduce that $\mathfrak{p}$ is nonzero from the much stronger assumption that it has height $1$).

We will just use two very elementary facts about polynomial localizations, which you can easily verify:

Given a domain $B$ and prime $\mathfrak{p} \subset B[t]$ such that $\mathfrak{p} \cap B = \mathfrak{p_0}$, it's always the case that

(1) $\mathfrak{p} B_{\mathfrak{p_0}}[t]$ is a prime ideal of $B_{\mathfrak{p_0}}[t]$.
(2) $B[t] \cap \mathfrak{p}B_\mathfrak{p_0}[t] = \mathfrak{p}$

In particular if $\mathfrak{p_0} = 0$ we get that

(1') $\mathfrak{p}K[t]$ is a maximal ideal of $K[t]$ (since primes are maximal in $K[t]$, it being a $PID$) and
(2') $B[t] \cap \mathfrak{p}K[t] = \mathfrak{p}$

By (1') we can take $\mathfrak{m} = \mathfrak{p}K[t]$.
By (2') we see that every element of $B[t] - \mathfrak{p}$ is also an element of $K[t] - \mathfrak{m}$, and thus $B[t]_\mathfrak{p}$ canonically embeds in $K[t]_\mathfrak{m}$: any fraction in $B[t]_\mathfrak{p}$ is a perfectly good fraction in $K[t]_\mathfrak{m}$.

The opposite inclusion is even simpler. Consider any element of $K[t]_\mathfrak{m}$. Write it as $\frac{g}{h}$ with $g \in K[t]$ and $h \in K[t] - \mathfrak{p}K[t]$. Multiply $g$ and $h$ by a common denominator $d \in B$ to get an equivalent fraction $\frac{g'}{h'}$ with $g',h' \in B[t]$. Since $h'$ is also not in $\mathfrak{p}K[t]$ (else $h = \frac{h'}{d}$ would have been too) we have that $h' \in B[t] - \mathfrak{p}$, and $\frac{g}{h} \sim \frac{g'}{h'}$ is a fraction in $B[t]_\mathfrak{p}$.


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.