For $A$-modules and homomorphisms $0\to M'\stackrel{u}{\to}M\stackrel{v}{\to}M''\to 0$ is exact. Prove if $M'$ and $M''$ are fintely generated then $M$ is finitely generated.
-
2Do this for vector spaces. Then do exactly the same for modules. – Mariano Suárez-Álvarez Nov 11 '12 at 09:47
-
2See the proof of the lemma of my answer to this question. http://math.stackexchange.com/questions/231058/noetherian-rings-and-modules – Makoto Kato Nov 11 '12 at 10:01
-
1You do not need $u$ to be injective, by the way. The result is true without that assumption. – Max Demirdilek Apr 13 '23 at 12:28
3 Answers
Suppose $M'$ is generated by $x_1,\dots,x_n$ and $M''$ is generated by $z_1,\dots,z_m$. Let $v(y_i) = z_i$ for $i = 1,\dots,m$. Let $x \in M$. Then there exist $b_1,\dots,b_m \in A$ such that $v(x) = b_1z_1 + \cdots + b_mz_m$. Then $v(x) = v(b_1y_1 + \cdots + b_my_m)$. Hence $x - (b_1y_1 + \cdots + b_my_m) \in \operatorname{Ker}v$. Since $\operatorname{Ker}v = \operatorname{Im}u$, there exist $a_1,\dots,a_n \in A$ such that $x - (b_1y_1 + \cdots + b_my_m) = a_1u(x_1) + \cdots + a_nu(x_n)$. Hence $M$ is generated by $u(x_1),\dots,u(x_n), y_1,\dots,y_m$.
- 53,190
- 44,216
$\require{begingroup} \begingroup$ $\def\coker{\operatorname{Coker}}$
Since $M',M''$ are finitely generated, we can pick surjections $s_{M'}:R^m\to M'$ and $s_{M''}:R^n\to M''$. Now consider the following commutative diagram:
$\require{AMScd}$ \begin{CD} 0 @>>>R^m @>>> R^m\oplus R^n @>>> R^n@>>>0 \\ @. @VV{s_{M'}}V @. @VV{s_{M''}}V\\ 0 @>>>M' @>{u}>>M @>{v}>> M'' @>>> 0\\ \end{CD}
where the vertical arrows are the surjections. Then as $M\to M''$ is a surjection and $R^n$ is projective (since free), we get map $w:R^n\to M$ so that the composition $v\circ w$ is the same as the surjection $s_{M''}:R^n\to M''$. We can then define a map from $s_M:R^{m+n}=R^m\oplus R^n\to M$ as follows: send $(x,y)\mapsto (u\circ s_{M'})(x) + w(y)$.
Now by snake lemma, we have exact sequence $\coker{s_{M'}}\to\coker{s_M}\to\coker{s_{M''}}$. Then $\coker{s_{M'}}, \coker{s_{M''}}$ are trivial by assumption, so is $\coker{s_{M}}$. $\endgroup$
- 1,076
-
Isn't it more direct to invoke the "four lemma" here instead of the "snake lemma"? Then we don't even need to consider cokernels. – Smiley1000 Nov 21 '24 at 09:55
I would consider this as a very special case of the Horseshoe lemma and prove it like that. This is essentially the same prove as in Makoto Kato wrote down, but his is written down more elementary.
- 9,750