3
$\begingroup$

$(R,m)$ and $(S,n)$ are local rings (commutative Noetherian with 1).
Let $f:R\to S$ be a local homomorphism/monomorphism ($f(m)\subseteq n$), such that the natural induced homomorphism $R/m\to S/n$ is an isomorphism, and such that $S$ is finite $R$-module via $f$ (So it is integral extension). Let $M$ is an $S$-module (not necessarily finite). So $M$ is also an $R$-module via $f$.

If $M$ is an Artinian $S$-module, is it an Artinian $R$-module?
If $M$ is an Artinian $R$-module, is it an Artinian $S$-module?

The first is true for $M=S$, since in this case S is of finite length.

Thank you.

$\endgroup$
0

1 Answer 1

3
+50
$\begingroup$

Since Artinian modules for $S$ and for the completion of $S$ are the same we may assume that $S$ and $R$ are complete.

Let $E_S$ be the injective hull of $S/n$. Then every Artinian $S$-module is a submodule of $E_S^N$ for some $N\ge0$. Thus it suffices to prove that $E_S$ is Artinian as an $R$-module.

Let $I:=f(m)S$. Then the fiber $S/I$ is finite length as an $S$-module, hence a finite dimensional $k:=R/m$-vector space. Now consider the annihilator $A$ of $f(m)$ in $E_S$. It is also the annihilator of $I$ in $E_S$ which is Matlis dual to $S/I$. This implies that $A$ is also a finite dimensional $k$-vector space. Let $d$ be its dimension.

By the universal property of the injective hull, each $R$-homomorphism $k\to E_S$ extends to a homomorphism $E_S\to E_R$. This gives an $R$-homomorphism $\phi:E_S\to E_R^d$.

Claim: $\phi$ is injective. Let $K=\ker\phi$. Then by construction $K\cap A=0$. But every element of $E_S$ is killed by some power of $I$. Hence every element of $K$ is killed by some power of $m$. Thus $K=0$.

Now we are done since $E_R^d$ is clearly $R$-Artinian.

The opposite direction is obvious.

$\endgroup$
2
  • $\begingroup$ thanks. where you used this: " A is also a finite dimensional k-vector space. Let d be its dimension."? and why $K\cap A=0$ $\endgroup$
    – user 1
    Jun 13, 2016 at 14:08
  • $\begingroup$ Let $v_1,\ldots,v_d$ be a basis of $A$. Then $Rv_i$ is an $R$-submodule of $E_S$ isomorphic to $k=R/m$. Since $E_R$ is injective the map $Rv_i=k\hookrightarrow E_R$ extends to an $R$-linear map $\Phi_i:E_S\to E_R$. Combined this gives a map $\phi:E_S\to E_R^d$. Moreover the restriction of $\phi$ to $A$ is $Rv_1+\ldots Rv_d=k^d\to E_R^d$ and therefore injective. This means $K\cap A=0$. $\endgroup$ Jun 13, 2016 at 15:18

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.