I have this doubt from a proposition in Dimension theory from Atiyah Macdonald. The proposition is as follows.

Let $A$ be a noetherian local ring, $\mathcal{m}$ be its maximal ideal, $\mathcal{q}$ an $\mathcal{m}$-primary ideal, $M$ a finitely generated $A$-module, $(M_n)$ a stable $\mathcal{q}$-filtration of $M$. Then, $M/M_n$ has finite length for each $n\geq 0$.

The proof goes as follows. Let $G(A)=\oplus_n\mathcal{q}/\mathcal{q}^{n+1}$, $G(M)=\oplus_m M_n/M_{n+1}$. Then $G_0(A)=A/\mathcal{q}$ is an artin local ring, $G(A)$ is noetherian and $G(M)$ is a finitely generated graded $G(A)$ module. Each $G_n(M)=M_n/M_{n+1}$ is a Notherian $A$-module annihilated by $\mathcal{q}$, hence a noetherian $A/\mathcal{q}$ module and therefore is of finite length since $A/\mathcal{q}$ is Artin. So far it is okay, from this they conclude saying

Hence $M/M_n$ is of finite length and $l(M/M_n)=\Sigma_{r=1}^n l(M_r/M_{r+1})$. How do we get this. I am not able to find any exact sequence where I can use the additivity of length. For example, I thought for $n=2$, this would work, $0\longrightarrow M/M_1\longrightarrow M/M_2\longrightarrow M_1/M_2\longrightarrow 0$. But I am not even able to define maps to make this an SES. ANy help will be appreciated.


Use the short exact sequence $0\rightarrow M_1/M_2 \rightarrow M/M_2 \rightarrow M/M_1 \rightarrow 0$, where all homomorphisms are naturally defined.

  • $\begingroup$ Oh why did it not occur to me! My brain must have shut down! $\endgroup$ – gradstudent Apr 23 '14 at 9:52

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.