Let $R$ be a ring. Let's work in the category of left $R$-modules. There is a way of looking at $\operatorname{Ext}_R^1(M,A)$ as classifying extensions of $M$ by $A$ up to equivalence.

But, if we look at the $\operatorname{Ext}_R^1(M, -)$ as the first derived functor of $\hom(M, -)$, for every short exact sequence of $R$-modules $$0 \to A \to B \to C \to 0 $$ we get a long exact sequence of form $$0 \to \hom(M,A) \to \hom(M,B) \to \hom(M,C) \to \operatorname{Ext}_R^1(M,A) \to \operatorname{Ext}_R^1(M,B) \to \operatorname{Ext}_R^1(M,C) \to \operatorname{Ext}_R^2(M,A) \to \ldots $$

In this formulation, we could say that $\operatorname{Ext}_R^1(M,A)$ in a way measures failure to lift homomorphisms $M \to C$ to homomorphisms $M \to B$. This gives a connection between extensions of $M$ by $A$ and lifting of homomorphisms $M \to C$.

For an example, if we have no non-trivial extension of $M$ by $A$, then $\operatorname{Ext}_R^1(M,A) = 0$ and we can always lift homomorphisms $M \to B$ to homomorphisms $M \to C$.

My question is: what would be a more concrete way of connecting these two notions? If I have a homomorphism $M \to C$ that I can't lift to homomorphism $M \to B$, do I automatically get an extension of $M$ by $A$? Is this described in some book/set of lecture notes?

  • $\begingroup$ Robinson's "Course in the Theory of Groups" has a chapter on low-dimensional group cohomology in terms of central extensions. $\endgroup$ – anomaly Aug 30 '18 at 13:03

You can construct the connecting morphism $\hom(M,C) \to \operatorname{Ext}^1(M,A)$ explicitly in terms of the short exact sequence, and this should give you an idea of what's going on. Indeed, given a map $u:M\to C$, "plug it in" the diagram of the sequence and form the pullback $E$ of $B\to C\leftarrow M$. The kernel of the pullback is naturally isomorphic to $A$, so you obtain an extension $\xi(u):A\to E\to M$, which lives in $\operatorname{Ext}^1(M,A)$. Its class gives a representative of $\partial(u)$.

Claim We can lift $u : M\to C$ to a map $h:M\to B$ if and only if the class of $\xi(u)$ is trivial.

Proof Suppose first the map can be lifted, so there is (calling the maps of your sec $f,g$ as usual) a map $h:M\to B$ such that $gh = u$. Recall that $E= \{ (b,m) : gb = um\}$. Since $gh=u$, this is just the set of $(b,m)$ such that $b-hm\in \ker g$. Now $\ker g = \operatorname{im} f\simeq A$ and we get a map $t: E\to A$ that sends $(fa,m)$ to $fa-hm$. This splits the inclusion $\operatorname{im} f\to E$, so by the splitting lemma the extension $\xi(u)$ is trivial.

Conversely, suppose that the extension is trivial, so there is a map $k: \operatorname{im}f\oplus M\to E$ such that $k(fa,0) = (fa,0)$ and $\pi k(fa,m) = m$. This means the map is of the form $k(fa,0) = (fa+hm,m)$ for some map $h$. But since this lands in $E$, we get that $g(fa+hm) =ghm = um$, so $h$ is a lift.$\hspace{8 cm}\blacktriangleleft$

It is important that you observe that not every class in $\operatorname{Ext}^1(M,A)$ arises as such obstruction. This happens, by the LES, if and only if the map $\operatorname{Ext}^1(M,f)$ is injective.

  • $\begingroup$ (It would be useful for you to carry out this in the dual situation, regarding lifts of maps $N\to A$, where the proof of the dual statement will require, of course, pushouts.) $\endgroup$ – Pedro Tamaroff Sep 3 '18 at 10:43

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.