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

$\renewcommand{\im}{\mathop{\rm im}}\DeclareMathOperator{\coker}{coker}$Let $A\xrightarrow{f}B\xrightarrow{g}C$ be a complex in an abelian category, I.e. $gf=0$. Let $H:=\coker(\im(f)\to \ker(g)).$

Then $H\xrightarrow{i}\coker(f)\xrightarrow{p}\im(g)$ is exact.

Could you give me a proof of it?

I can say that $i$ is a monomorphism and $p$ is epi. I know that $\im(f)=\ker(\coker(f))=\coker(\ker(f))$ and that $\im(f)\to \ker(g)$ is monomorphic.

But I want $\im(i)=\ker(p).$ I cannot prove even that $pi=0.$

share|cite|improve this question
See Kashiwara-Schapira, (8.3.4), page 178. This thread might be interesting (although it doesn't contain an answer to your specific question). – commenter Oct 27 '12 at 13:22
Do you accept usage of the nine lemma here? Your desired sequence is obtained by taking $\im(f)$-quotient of the exact sequence $0\to \ker(g)\to B\to \im(g)\to 0$ which may be the bottom row of an $3\times 3$ square with exact cols and two exact top rows. – Ben Oct 27 '12 at 16:06
@commenter Thank you for letting know me the book. – Tom Oct 28 '12 at 6:00
up vote 2 down vote accepted

Let us denote the objects by capital and corresponding arrows by small letters: ${\rm coim} f:A\to {\rm Coim} f$. Sorry, I write from left to right.

First of all, as in any Abelian category, $f={\rm coim} f\cdot {\rm im}f$, so $$fg=0 \iff {\rm coim}f\, {\rm im}f\cdot{\rm coim}g\,{\rm im}g=0 \iff {\rm im}f\cdot {\rm coim g}=0,$$ as ${\rm coim} f$ is epi and ${\rm im}g$ is mono, moreover, we also have ${\rm ker}g={\rm ker}({\rm coim}g)$ and ${\rm coker}f={\rm coker}({\rm im}f)$. Hence wlog. we can assume that $f$ is mono ($f={\rm im}f$) and $g$ is epi ($g={\rm coim} g$), and this way $A={\rm Im}f$ and $C={\rm Coim}g$ is also assumed.

We can then draw a (bit simplified) diagram with $$ A\underbrace{\overset{j}\to {\rm Ker}g \to B}_f \overbrace{\to{\rm Coker}f \underset{p}\to C}^g$$ where $H={\rm Coker}j$. As $j\cdot{\rm ker}g\cdot{\rm coker}f=f\cdot{\rm coker}f=0$, we get the $i:H\to{\rm Coker}f$. Now $${\rm coker}j\cdot i\, p= {\rm ker}g\cdot {\rm coker}f\cdot p= {\rm ker}g\cdot g=0$$ hence, $ip=0$, because ${\rm coker}j$ is epi.

For the exactness, it is enough to prove that $p={\rm coker}i$. For this: if $iv=0$ for any arrow $v$, we have $0={\rm coker} j\cdot iv={\rm ker} g\cdot{\rm coker}f\cdot v$. But, as $g$ is epi, $g={\rm coim}g={\rm coker}({\rm ker}g)$, this goes through $g$: we get a $t$ such that $gt={\rm coker}f\cdot v$. Then, using that ${\rm coker}f$ is epi, we arrive to $pt=v$.

share|cite|improve this answer
Thank you very much for writing the detailed answer. I understood all steps without any trouble! – Tom Oct 28 '12 at 3:31
Only for our statement, the first argument showing $f$ mono $g$ epi does not contribute to shorten the total length of the proof, doesn't it? However the argument was first for me and very informative, thank you so much. – Tom Oct 28 '12 at 5:57

$\renewcommand{\im}{\mathop{\rm im}}\DeclareMathOperator{\coker}{coker}$You can use Freyd-Mitchell Embedding theorem to reduce to the case of modules. Note that $H = \ker(g)/\im(f)$, so you have $$\ker(g)/\im(f) \xrightarrow{i}\coker(f) = B/\im(f) \xrightarrow{p} \im(g).$$ Now, clearly $pi=0$, so then we have that $\im i \subset \ker p$. For the opposite inclusion suppose that $x \in B /\im(f)$ is in $\ker p$. Take a representative element of the coset in B, and argue that it must lie in $\ker(g)$. Argue that you always can lift a coset x in $B/\im(f)$ that lies in the kernel of p to an element of $\ker(g)/\im(f)$.

share|cite|improve this answer
Thank you. I know the proof for modules. I hope a direct proof without their big theorem. – Tom Oct 27 '12 at 13:03
Happy to help! Good luck. – Dedalus Oct 27 '12 at 13:03
Do you have a direct proof? – Tom Oct 27 '12 at 13:04

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.