Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Proposition Claim: Let $A$ be a finite-dimensional algebra over a field $k$. Suppose $T,T'$ are simple $A$-modules and that there is a non-split extension $M$ of $T$ by $T'$. Then $T$ and $T'$ are composition factors of the same projective indecomposable module.

Proof. We have a short exact sequence $$0\to T'\xrightarrow{\iota_1} M\xrightarrow{\pi_1} T\to 0.$$

Let $P_T$ be the projective indecomposable corresponding to $T$, so $P_T/\mathrm{rad}(P_T)\cong T$, i.e. we have another short exact sequence $$ 0\to \mathrm{rad}(P_T)\xrightarrow{\iota_2}P_T\xrightarrow{\pi_2}T\to 0. $$

By the universal property of projective modules, using the surjective maps $\pi_1$ and $\pi_2$, there is a unique map $\varphi:P_T\to M$ satisfying $\pi_1\circ \varphi=\pi_2$.

Since the second sequence is exact, we know $\pi_2\circ\iota_2=0$ and so $\pi_2\circ\varphi\circ\iota_1=0$. Notice that $\ker\pi_2=T'$.

Hence by the universal property of $\ker\pi_2$, there exists a unique nonzero map $\psi:\mathrm{rad}(P_T)\to T'$. Since $T'$ is simple, $\psi$ is surjective and by the first isomorphism theorem $$ \mathrm{rad}(P_T)/\ker\psi\cong T'. $$ Hence $P_T$ has a filtration $$ P_T\supset \mathrm{rad}(P_T)\supset \ker\psi\supset 0 $$ with composition factors $T$ and $T'$ as required.

share|cite|improve this question
Consider writing the projective cover of M as a direct sum of PIMs. – Jack Schmidt Sep 2 '11 at 3:44
rad(PT′) need not be isomorphic to T, it will just have T as a top composition factor. – Jack Schmidt Sep 2 '11 at 3:45
Please do consider examples! Path algebras and their quotients provide an immense wealth of them ---if you are not familiar with them, do yourself a favor and read the first chapter of the book [Simson, Daniel; Skowronski, Andrzej; Assem, Ibrahim (2007), Elements of the Representation Theory of Associative Algebras, Cambridge University Press], for example. Most humans learn from examples (I am told by a few of his former students that Maurice Auslander was an exception to this, though...) – Mariano Suárez-Alvarez Sep 2 '11 at 3:48
You cannot conclude that $\psi$ is surjective using solely the fact that $T'$ is simple: $\psi$ could very well be zero... (Also: one usually only uses the term composition series when the subquotients of the sequence of submodules are simple: this is not true in the series $P_T\supset \mathrm{rad}(P_T)\supset \ker\psi\supset 0$ at the end of your edit) (Please, when you edit a question, do not remove old text, because otherwise you more or less render old comments and answers unintelligible: add text at the end instead) – Mariano Suárez-Alvarez Sep 2 '11 at 6:10
$\psi$ is nonzero because it fits into the commutative diagram no? (I have said that $\psi$ is nonzero in my proof). Changing to the word "filtration" is enough to complete the proof, since we may just refine this filtration to a composition series with the right top two factors. – Clinton Boys Sep 2 '11 at 6:24
up vote 0 down vote accepted


0 ---> T' ---> M ---> T ---> 0 

is a non-split extension with $T$ and $T'$ simple. Let $P\to T$ be the projective cover of $T$, whose kernel is $\mathrm{rad}(P)$. You should show that there is a commutative diagram with exact rows

0 ---> rad(P) ---> P ---> T ---> 0 
         |         |      |
         |         |      |
         V         V      V
0 --->   T'   ---> M ---> T ---> 0 

in which the arrow $T\to T$ is the identity and the map $\mathrm{rad}(P)\to T'$ is not zero.

Can you conclude from this that $T'$ is a composition factor in $P$?

For an example where your argument fails, consider the path algebra of the quiver $$\overset1\bullet\to\overset2\bullet\to\overset3\bullet$$ There is an extension of the simple $S_1$ by the simple $S_2$, but the radical of the projective cover $P_1$ of $S_1$ has length $2$, so it cannot be isomorphic to $S_2$ (depending on your conventions on modules, you might need to interchange $1$s and $3$s in what I wrote....)

share|cite|improve this answer
I see where your map $P\to M$ comes from (basically the def. of projective). Where does your map $\mathrm{rad}(P)\to T'$ come from? – Clinton Boys Sep 2 '11 at 4:35
@user1375: once you have constructed the map $P\to M$ so that the rightmost square commutes, you can see that the composition $\mathrm{rad}(P)\to P\to M\to T$ is zero: it follows from this that the image of the composition $\mathrm{rad}(P)\to P\to M$ is contained in the kernel of the map $M\to T$. &c. (BTW: Please consider changing your user name to something less generic, as it is much comfortable to communicate with someone with a more identifiable name!) – Mariano Suárez-Alvarez Sep 2 '11 at 4:38
Sorry, still not following. Are you using some universal property to construct this map? – Clinton Boys Sep 2 '11 at 4:45
If a map $f:X\to Y$ has its image in the kernel $K$ of $g:Y\to Z$, and we write $\iota:K\to Y$ the inclusion of thay kernel into $Y$, then there exists a unique map $\bar f:X\to K$ such that $\iota\circ\bar f=f$. That is the characteristic property of the kernel of a map that I am using. – Mariano Suárez-Alvarez Sep 2 '11 at 5:12
Right, the universal property of the kernel. Not sure how I missed that. OK so now we need some module $X$ such that $P/X\cong T'$? – Clinton Boys Sep 2 '11 at 5:32

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.