Let $R$ be a commutative, Noetherian ring.

Let $d$ be a dimension function: for each finitely generated $R$-module $M$, we assign a natural number, or zero, such that for every $N \leq M$ a submodule we have $d(M) = \max\{d(N), d(M/N)\}$.

Let $J \triangleleft R$ be an ideal and let $m \in \mathbb{N}$. How does it follow that $d(A/J) = d(A/{J^{m}})$?

So far, I think that if we assume that $d$ is invariant under isomorphisms (if $M_1 \cong M_2$ then $d(M_1) = d(M_2)$, which isn't technically given), we get that $d(A/J^m) = \max\{d(A/J),d(A/J^m/A/J)\} \geq d(A/J)$, since for any $m \geq 1$ we have $J^m \leq J$ and so $A/J$ is isomorphic to a submodule of $A/J^m$. How do you get the other direction?

Some useful facts that may be relevant:

  1. The set of minimal prime ideals over $J$ is finite and non-empty.

  2. There are primes $I \subset P_i$ s.t $P_1P_2\cdots P_n \subset I$.

  3. If $M$ is a finitely generated $R$-module then setting $I = \mathrm{Ann}_R(M)$ we get that $R/I$ is isomorphic to an $R$-submodule of $M^n$ for $n$ the number of generators of $M$.

The first two facts follow from $R$ being Noetherian. The third from commutativity as well.


Partial answer:

I'll assume that more precisely the condition is that if $$0\to M'\to M\to M''\to 0$$ is a short exact sequence of modules over $R$ (commutative, Noetherian) with $M$ finitely generated, then $d(M)=\max\{d(M'),d(M'')\}$. This addresses the isomorphism vs set theoretic equality problem that the original statement had.

I suspect the claim in the question is false, but I can't come up with a counterexample, so I'm leaving this partial answer here. Mostly the problem is that I can't think of a function $d$ which has the desired property, so if the OP has one, adding that context to the question would be really helpful.

This is a weird property. First consider a direct sum. $M\oplus N$. $d(M\oplus N)=\max\{d(M),d(N)\}$, which implies that $d(M^n)=d(M)$, so in particular, $d(R^n)=d(R)$. However, if $M$ is finitely generated, then there is a surjection $R^n\to M$ for some $n$, which implies that $d(M)\le d(R^n)=d(R)$. Thus $d(R)$ is an upper bound on all the "dimensions" of modules over $R$. (I put dimensions in quotation marks, because this behaves very differently from the notions of dimension that I am familiar with.)

Now if $I$ is an ideal of $R$, then $d(R)=\max\{d(I),d(R/I)\}$, which means that one of $d(I)$ or $d(R/I)$ equals $d(R)$. If $d(R/I)=d(R)$, then since $R/I$ is a quotient of $R/I^m$ for all $m\ge 1$, we have $d(R/I)=d(R)\le d(R/I^m)\le d(R)$, so $d(R/I)=d(R/I^m)=d(R)$ in this case.

In the other case, $d(R/I)<d(R)$, and $d(I)=d(R)$. This is the case in which we could falsify the claim in the question, because if $d(I^m) < d(R)$, $d(R/I^m)=d(R)$, then there is no clear contradiction.

However, I can't think of any examples of functions $d$ which have the desired property, so I don't have a counterexample at hand.

  • $\begingroup$ thanks for the partial answer - I agree this is weird. Unfortunately I don't have an example on hand, I'll look further into my notes. Indeed the direct sum property you mentioned is strange, I came upon it too.. $\endgroup$ – Mariah Feb 16 '19 at 1:23

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.