0
$\begingroup$

Let $R$ be a commutative ring. Suppose $R$-modules $X,A,B,C$ and $Y$ are given such that the outer two rows and the outer two columns in the following diagram are exact.

$\hskip1in$ Diagram

Does it follow that there exists an $R$-module $D$ such that the augmented diagram commutes and the dotted row and column are exact? Or are there counterexamples?

Such a $D$ would provide a simultaneous extension of $C$ by $X$ and of $Y$ by $A$.

This is easily seen to be true if either the left column splits (in which case we may take $D=C\oplus X$) or if the bottom row splits (in which case we may take $D=A\oplus Y$). So the answer is positive if $R$ is a field, for instance. I'm not sure how to prove the general case, neither have I been able to produce a counterexample.

Any idea or reference would be welcome.

Edit: In light of Jason Starr's comment providing a counterexample over $R=\mathbb Z[t]$, I would also be interested in sufficient conditions on $R$ such that the claim is true. For instance:

Does the claim hold if $R$ is a PID? Hereditary?

$\endgroup$
2
  • 2
    $\begingroup$ That is not true. Let $C$ be a projective $R$-module, and let $B\subset C$ be an $R$-submodule that is not projective, e.g., $C=R=\mathbb{Z}[t]$ and $B=\langle 2,t \rangle$. Let $A\to B$ a surjection from a projective $R$-module. If such a diagram exists, then the middle column would split ($C$ is projective), implying the first column splits, implying that $B$ is projective. $\endgroup$ Jun 29, 2016 at 17:36
  • $\begingroup$ @JasonStarr: Thanks, this is certainly useful information. Do you perhaps know of any sufficient conditions on $R$ under which the claim holds? (Aside from $R$ being a field?) $\endgroup$
    – Dejan Govc
    Jun 29, 2016 at 18:03

1 Answer 1

0
$\begingroup$

As pointed out by Jason Starr in his comment, this is not true in general.

However, it is true if $R$ is hereditary. To see this, notice that the exact sequence $$0\to X\to A\to B\to0$$ gives rise to the following long exact sequence of $\operatorname{Ext}$-modules: $$0\to\operatorname{Hom}(Y,X)\to\operatorname{Hom}(Y,A)\to\operatorname{Hom}(Y,B)\to\phantom{0}\\\phantom{0}\to\rlap{\;\operatorname{Ext}(Y,X)}\phantom{\operatorname{Hom}(Y,X)}\to\rlap{\;\operatorname{Ext}(Y,A)}\phantom{\operatorname{Hom}(Y,A)}\to\rlap{\;\operatorname{Ext}(Y,B)}\phantom{\operatorname{Hom}(Y,B)}\to0$$ Note that all higher $\operatorname{Ext}$-modules vanish since $R$ is hereditary.

In particular, $\operatorname{Ext}(Y,A)\to\operatorname{Ext}(Y,B)$ is an epimorphism. By the classical interpretation of elements of $\operatorname{Ext}$-modules as extensions, this means that there is a map of extensions as follows:

$\hskip1in$ enter image description here

By construction, the two outermost vertical arrows agree with the ones given in the question. Applying the Snake Lemma now concludes the proof.

$\endgroup$

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.