The Stacks project

10.74 An application of Ext groups

Here it is.

Lemma 10.74.1. Let $R$ be a Noetherian ring. Let $I \subset R$ be an ideal contained in the Jacobson radical of $R$. Let $N \to M$ be a homomorphism of finite $R$-modules. Suppose that there exists arbitrarily large $n$ such that $N/I^ nN \to M/I^ nM$ is a split injection. Then $N \to M$ is a split injection.

Proof. Assume $\varphi : N \to M$ satisfies the assumptions of the lemma. Note that this implies that $\mathop{\mathrm{Ker}}(\varphi ) \subset I^ nN$ for arbitrarily large $n$. Hence by Lemma 10.51.5 we see that $\varphi $ is injection. Let $Q = M/N$ so that we have a short exact sequence

\[ 0 \to N \to M \to Q \to 0. \]

Let

\[ F_2 \xrightarrow {d_2} F_1 \xrightarrow {d_1} F_0 \to Q \to 0 \]

be a finite free resolution of $Q$. We can choose a map $\alpha : F_0 \to M$ lifting the map $F_0 \to Q$. This induces a map $\beta : F_1 \to N$ such that $\beta \circ d_2 = 0$. The extension above is split if and only if there exists a map $\gamma : F_0 \to N$ such that $\beta = \gamma \circ d_1$. In other words, the class of $\beta $ in $\mathop{\mathrm{Ext}}\nolimits ^1_ R(Q, N)$ is the obstruction to splitting the short exact sequence above.

Suppose $n$ is a large integer such that $N/I^ nN \to M/I^ nM$ is a split injection. This implies

\[ 0 \to N/I^ nN \to M/I^ nM \to Q/I^ nQ \to 0. \]

is still short exact. Also, the sequence

\[ F_1/I^ nF_1 \xrightarrow {d_1} F_0/I^ nF_0 \to Q/I^ nQ \to 0 \]

is still exact. Arguing as above we see that the map $\overline{\beta } : F_1/I^ nF_1 \to N/I^ nN$ induced by $\beta $ is equal to $\overline{\gamma _ n} \circ d_1$ for some map $\overline{\gamma _ n} : F_0/I^ nF_0 \to N/I^ nN$. Since $F_0$ is free we can lift $\overline{\gamma _ n}$ to a map $\gamma _ n : F_0 \to N$ and then we see that $\beta - \gamma _ n \circ d_1$ is a map from $F_1$ into $I^ nN$. In other words we conclude that

\[ \beta \in \mathop{\mathrm{Im}}\Big(\mathop{\mathrm{Hom}}\nolimits _ R(F_0, N) \to \mathop{\mathrm{Hom}}\nolimits _ R(F_1, N)\Big) + I^ n\mathop{\mathrm{Hom}}\nolimits _ R(F_1, N). \]

for this $n$.

Since we have this property for arbitrarily large $n$ by assumption we conclude that the image of $\beta $ in the cokernel of $\mathop{\mathrm{Hom}}\nolimits _ R(F_0, N) \to \mathop{\mathrm{Hom}}\nolimits _ R(F_1, N)$ is zero by Lemma 10.51.5. Hence $\beta $ is in the image of the map $\mathop{\mathrm{Hom}}\nolimits _ R(F_0, N) \to \mathop{\mathrm{Hom}}\nolimits _ R(F_1, N)$ as desired. $\square$


Comments (3)

Comment #6547 by wqx on

I know it can be used to prove the Direct Summand Conjecture in charateristic p.So is there another application?

Comment #6548 by wqx on

And there is a typo in the proof.In the last seven lines,'induced by \beta is equal to \gamma_n d_1',it should be \bar{ \gamma_n},and 'N/I^n' should be N/I^nN


Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.




In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 02HN. Beware of the difference between the letter 'O' and the digit '0'.