Let $A$ be an integral domain with quotient field $K$. We say that $A$ is a valuation ring if for any $0 \neq x \in K$, $x$ or $\frac{1}{x}$ lies in $A$. Then $A$ is necessarily a local ring. If $A$ is a valuation ring, then any ring $B$ between $A$ and $K$ is also valuation ring. In fact, $B$ is necessarily the localization of $A$ at one of its prime ideals (if $\mathfrak m$ is the maximal ideal of $B$, then $B$ is the localization of $A$ at $A \cap \mathfrak m$). This also implies something cool: since the ideals of a valuation ring are lineary ordered (easy to prove), so are all the intermediate rings between $A$ and $K$.

This recent question (Can a valuation ring properly contains another valuation ring with the same field of fractions?) got me thinking about something. The question I linked to is whether such a $B$ as I've indicated above, may be anything other than $A$ and $K$.

The answer to that question is yes: let $A$ be a valuation ring with $1 < \textrm{Dim } A < \infty$ (they exist), and let $\mathfrak p$ be a prime ideal of $A$ which is not maximal and not zero. Then $A_{\mathfrak p}$ cannot be equal to $A$ or $K$; if it were equal to one of those, then in particular they would have to be isomorphic, and they would then have the same Krull dimension. However, here we have $\textrm{Dim } K = 0 < \textrm{Dim } A_{\mathfrak p} < \textrm{Dim } A$.

This argument relies on the impossibility of these rings being isomorphic, because of their distinct dimensions. So my question is: what about the infinite dimensional case?

Does there exist an infinite dimensional valuation ring $A$, and a nonzero nonmaximal prime ideal $\mathfrak p$ of $A$, such that $A$ and $A_{\mathfrak p}$ are abstractly isomorphic?

First of all, for a commutative ring $A$ with infinite Krull dimension, it can certainly happen that $A_{\mathfrak p} \cong A_{\mathfrak q}$ for prime ideals $\mathfrak p \subsetneq \mathfrak q$. For example, $k$ a field, and $$A = k[X_1, X_2, ...], \mathfrak q = (X_1,X_3,X_5, ...), \mathfrak p = (X_3,X_5, ...)$$ I only know one example of an infinite dimensional valuation ring (https://math.berkeley.edu/~ogus/Math%20_256A--08/bigval.pdf), and tomorrow I'm gonna try to play around with the prime ideals of this ring. Right now, I'm too tired but I wanted to write my thoughts down in case I forgot, and also in the hope that someone who knows more about valuation rings can point me in the right direction.

  • 1
    $\begingroup$ "Abstractly" I mean some isomorphism, rather than a canonical isomorphism. $\endgroup$ – D_S Mar 21 '16 at 13:09

Yes, this can definitely happen. Let $G\mapsto A(G)$ be any operation which takes a totally ordered abelian group $G$ and gives you a valuation ring $A(G)$ with value group $G$ which is functorial with respect to isomorphisms (that is, an (order-preserving) isomorphism $G\to H$ induces an isomorphism $A(G)\to A(H)$ which turns valuations of elements $A(G)$ into valuations of elements of $A(H)$ via $\alpha$, and this respects composition; this will be satisfied by any operation that can reasonably be called a "construction on totally ordered abelian groups"). There are many such operations; for instance, you can use Hahn series with coefficients in any field.

Now suppose $G$ is a totally ordered abelian group, $S\subset G$ is an upward-closed set of positive elements such that $a+b\in S$ implies $a\in S$ or $b\in S$, and $\alpha:G\to G$ is an automorphism such that $\alpha(S)$ is a proper subset of $S$. It is easy to construct such triples $(G,S,\alpha)$. For instance, you can take $G$ to be freely generated as an abelian group by elements $g_n$ for $n\in\mathbb{Z}$, ordered such that each $g_n$ is positive and larger than every integer multiple of $g_{n-1}$. You can then let $S$ be the set of positive elements which have a nonzero coefficient of $g_n$ for some $n\geq 0$, and let $\alpha:G\to G$ be given by $\alpha(g_n)=g_{n+1}$.

Given such a triple $(G,S,\alpha)$, note that the set of elements of $A(G)$ whose valuation is in $S$ form a prime ideal $\mathfrak{p}\subset A(G)$. Furthermore, $\alpha$ induces an automorphism $\alpha_*:A(G)\to A(G)$, and $\mathfrak{q}=\alpha_*(\mathfrak{p})$ is just the set of elements of $A(G)$ whose valuation is contained in $\alpha(S)$. In particular, this means $\mathfrak{q}\subset\mathfrak{p}$. Then $\alpha_*$ induces an isomorphism between the localizations $A(G)_\mathfrak{p}$ and $A(G)_\mathfrak{q}$. But since $\mathfrak{q}\subset\mathfrak{p}$, $A(G)_\mathfrak{p}$ can itself be considered as a (nontrivial) localization of $A(G)_\mathfrak{q}$. So $A(G)_{\mathfrak{q}}$ is a valuation ring which is abstractly isomorphic to a localization of itself at a nonzero prime ideal, namely $\mathfrak{p}A(G)_\mathfrak{q}$.


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.