Questions on the construction of a Noetherian domain that is not an N-1 ring

114 Views Asked by At

In this wiki article there is an example I'm trying to understand. The same example seems to appear in two articles, actually.

The relevant definition is

An $N$-$1$ ring is an integral domain $A$ whose integral closure in its quotient field is a finitely generated $A$ module

The example is

If $R$ is the subring of the polynomial ring $k[x_{1},x_{2},...]$ in infinitely many generators generated by the squares and cubes of all generators, and $S$ is obtained from $R$ by adjoining inverses to all elements not in any of the ideals generated by some $x_{n}$, then $S$ is a 1-dimensional Noetherian domain that is not an $N$-$1$ ring.

First off, I think it makes sense that $R=k[\{x_i^2\mid i\in\mathbb N\}\cup\{x_i^3\mid i\in\mathbb N\}]$. Already it seems like its integral closure should be $R_0=k[\{x_i\mid i\in\mathbb N\}]$ which would seem to be not a finitely generated $R$ module. So this is already a domain that isn't N-1, right?

My second question is what is meant by the rest of the construction. Let $M_1$ be the multiplicative subset of $R_0$ that is the intersection of complements of ideals of the form $(x_i)$, and let $M_2=M_1\cap R$. Am I interpreting this correctly that $S=RM_2^{-1}$? If I'm reading that right, then how do I see that it makes $S$ $1$-dimensional and Noetherian?

$R$ itself does not look Noetherian, since you can generate an infinite ascending chain whose terms are sums of the first $n$ ideals of the form $(x_i^2, x_i^3)$. That also seems to show $R$'s Krull dimension is infinite. I'm not very clear on how the localization step fixes this in $S$.

1

There are 1 best solutions below

17
On BEST ANSWER

I agree that $R$ is already not N-1: the integral closure is $k[x_1,\cdots]$ which is not finitely generated over $R$, as any finite collection of generating elements would only be able to express finitely many of the $x_i$.

I agree with your interpretation that $S=M_2^{-1}R$. To see that $S$ is one-dimensional, the intuition is that localizing at $M_2$ removes any subvariety of $\operatorname{Spec} R$ which is contained in two or more of the subvarieties $V(x_i^2,x_i^3)$, leaving you with only these sets as proper subvarieties, which gives dimension one. Indeed, if $I\subset R$ is an ideal which contains both $I_i=(x_i^2,x_i^3)$ and $I_j$ for some $i\neq j$, then it intersects $M_2$ and $M_2^{-1}I=S$. So every prime ideal of $S$ comes from a prime ideal of $R$ which contains at most one $I_i$, and it is not so hard to show that such an ideal must actually be $I_i$ or $0$. Therefore the only prime ideals of $S$ are $0$ and $M_2^{-1}I_i$, and as the only nontrivial inclusions among these are $0\subset M_2^{-1}I_i$, $S$ is of dimension one. Further, as any ring where any prime ideal is finitely generated is noetherian and all of these ideals are finitely generated, we have that $S$ is noetherian.

(For what it's worth, I agree with your arguments in the final paragraph too.)