7
$\begingroup$

Is there a homological criterion for the condition $A(B \cap C) = AB \cap AC$ for ideals in a ring $R$? I mean a statement such as "the given equation holds if and only if (some $\operatorname{Tor}$, $\operatorname{Ext}$, local cohomology, etc) group vanishes/does not vanish".

Note that this is a local question, since we are asking when the inclusion $A(B \cap C) \to AB \cap AC$ is surjective. So I'm happy to assume $R$ local.

This is coming from the similar-looking fact that $AB = A \cap B$ if and only if $\operatorname{Tor}^1(R/A,R/B) = 0$.

(I asked this on StackExchange and did not receive a response.)

$\endgroup$
3
  • $\begingroup$ Related might be equivalent conditions for a domain to be a Pruefer domain: en.wikipedia.org/wiki/Pr%C3%BCfer_domain $\endgroup$
    – Todd Trimble
    Feb 27, 2015 at 2:01
  • $\begingroup$ math.stackexchange.com/questions/1166544/… $\endgroup$
    – user26857
    Feb 27, 2015 at 2:05
  • $\begingroup$ Indeed, it is sufficient for $R$ to be Prüfer. If $R$ is assumed Noetherian, it is sufficient for $R$ to be Dedekind. But I am interested in the geometric setting, so I'd like $R$ to be, say, finitely-generated over a field. Then it's an interesting question about the intersections of $R/A$, $R/B$ and $R/C$. $\endgroup$ Feb 27, 2015 at 4:15

2 Answers 2

5
$\begingroup$

Ah, here it is: we have $A(B\cap C) = AB \cap AC$ if and only if the natural map

$$\operatorname{Tor}_1^R(R/A,B) \oplus \operatorname{Tor}_1^R(R/A,C) \to \operatorname{Tor}_1^R(R/A,B+C)$$

is surjective.

This is not very pretty, but it's still entirely homological (for example, it can be computed in the completion.) For a more 'geometric' formulation, we may use the natural isomorphism $$\operatorname{Tor}_2^R(M,R/I) \to \operatorname{Tor}_1^R(M,I)$$ to express this in terms of $R/A$, $R/B$, $R/C$, and $R/(B+C)$, so this does indeed relate (somehow) to how $A$ intersects with $B$ and $C$.

Proof: Take $0 \to B \cap C \to B \oplus C \to B + C \to 0$ and tensor with $R/A$. Note that

$$\frac{AB \cap AC}{A(B \cap C)} = \ker\big(\frac{B \cap C}{A(B \cap C)} \to \frac{B}{AB} \oplus \frac{C}{AC} \big)$$

and so is the cokernel of the map above. (In particular, we are asking for the first boundary map in the long exact sequence for $\operatorname{Tor}$ to be zero.)

$\endgroup$
0
0
$\begingroup$

An obvious sufficient condition would be $$\operatorname{Tor}_1^R(R/A,R/B)=\operatorname{Tor}_1^R(R/A,R/C)=\operatorname{Tor}_1^R(R/A,R/(B\cap C))=0.$$ Unless for some reason you don't want this condition.

$\endgroup$
2
  • $\begingroup$ Dear Mahdi, I agree that this is sufficient, but I think it is a not necessary condition. $\endgroup$ Feb 27, 2015 at 21:24
  • $\begingroup$ In fact $\operatorname{Tor}_1^R(R/A,R/(B\cap C))=0$ is sufficient by itself, since then the inclusions $A(B\cap C) \subset AB \cap AC \subset A \cap B \cap C$ collapse. $\endgroup$ Feb 27, 2015 at 21:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.