I'd like to prove (i) implies (ii) where:

(i) Whenever $f: A \to B$ is injective and $A,B$ are finitely generated then $f \otimes \operatorname{id}: A \otimes P \to B \otimes P$ is injective.

(ii) If $f: M \to N$ is injective then $f \otimes \operatorname{id}: M \otimes P \to N \otimes P$ for arbitrary $R$-modules $M,N$.

Can you tell me if this is correct (I'm spelling out all the details because I want to be sure that I understand what I'm doing):

Let $M,N$ be arbitrary modules and let $f: M \to N$ be injective. Assume that $f \otimes \operatorname{id}$ is not injective so that there is an element $u = \sum_{i=1}^n m_i \otimes p_i$ in $M \otimes P$ such that $u \neq 0$ and $(f \otimes \operatorname{id}) (u) = \sum_{i=1}^n f(m_i) \otimes p_i = 0$ in $N \otimes P$.

Let $M_0$ be the module generated by $m_1, \dots , m_n$.

$\sum_{i=1}^n f(m_i) \otimes p_i = 0$ in $N \otimes P$ hence by proposition (2.13) in Atiyah-Macdonald we get finitely generated submodules $X \subset N$,$Y \subset P$ such that $\sum_{i=1}^n f(m_i) \otimes p_i = 0$ in $X \otimes Y$. Let $S$ be the (finite) set generating $X$.

Let $Z$ be the module generated by $S$ and $f(m_1), \dots , f(m_n)$. Then $Z$ is finitely generated and contains $f(M_0)$. Since $f : M \to N$ is injective we hence have that $f\mid_{M_0} : M_0 \to Z$ is injective and well-defined. By (i) we get that $f\mid_{M_0} \otimes \operatorname{id}$ is injective.

We have $(f\mid_{M_0} \otimes \operatorname{id})(u) = \sum_{i=1}^n f(m_i) \otimes p_i = 0$ in $X \otimes Y \subset Z \otimes Y$. Hence by injectivity of $f\mid_{M_0} \otimes \operatorname{id} : M_0 \otimes Y \to Z \otimes Y$ we get that $u = 0$ in $M_0 \otimes Y \subset M \otimes P$ and hence in $M \otimes P$.


The main idea of the proof is correct.

However, I don't understand why you want to assume that $f \otimes \operatorname{id}$ is not injective. Basically you take a $u \in M\otimes P$ such that $f \otimes \operatorname{id} (u)=0$ and then you show that $u=0$. So this is a straight-forward proof - no contraposition or contradiction is needed.

But this is only a minor thing. As I said above the remaining should be correct.

Note that an important point here is that zero stays zero if you "go up", i.e. $M'\subset M,N' \subset N$ and $u=0 \in M'\otimes N' \Rightarrow u=0 \in M\otimes N$. But if you "go down" with a non-zero-element then it may become zero, i.e. $M'\subset M,N' \subset N$ and $0 \neq u \in M'\otimes N'$ then not necessarily $0 \neq u \in M\otimes N$. (cf. Atiyah-Macdonald the remark ii) right before (2.13)).

  • $\begingroup$ Awesome, thank you very much! $\endgroup$ – Rudy the Reindeer Jun 11 '12 at 13:47

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.