2

I am trying to resolve an exercise and there are 2 point that are missing in order to finalize:

Suppose $A$, $B$, $C$, and $P$ are $R$-modules, and $f:A \rightarrow B$ and $g:B\rightarrow C$ are both $R$-module morphisms.

1) $\forall \phi : C \rightarrow P$ morphism, if $\phi \circ g = 0 \Rightarrow \phi = 0$, for a morphism $\phi : C \rightarrow P$, does this imply that $g$ is surjective? Why?

2) If $\phi \circ g \circ f = 0$ $ \forall \phi : C \rightarrow P$ morphism does this mean that $g \circ f = 0$? Why?

zipirovich
  • 14,670
  • 1
  • 26
  • 35
roi_saumon
  • 4,196
  • 1
    In the first point, should the implication hold for some morphism $\phi$ or for all morphisms $\phi$? And is $P$ some fixed $R$-module, or should the implication hold for all $R$-modules $P$? – Servaes Dec 05 '18 at 19:30
  • for all! I will edit, thanks – roi_saumon Dec 05 '18 at 19:31

2 Answers2

1

For the first point consider for $\phi$ the quotient morphism $\pi:\ C\ \longrightarrow\ \operatorname{coker}g$.

For the second point consider for $\phi$ the identity morphism $\operatorname{id}:\ C\ \longrightarrow\ C$.

Servaes
  • 63,261
  • 7
  • 75
  • 163
  • Oh, so nice! Especially the one with the cokernel. How did you think about that? Is it a standard trick? – roi_saumon Dec 05 '18 at 21:27
  • Well if $\phi\circ g=0$ then $\operatorname{im}g\subset\ker\phi$ and so $\phi$ factors over the morphism $C\ \longrightarrow\ \operatorname{coker}g$. So the implication holds for all $\phi$ if and only if it holds for the factor map $\pi:\ C\ \longrightarrow\ \operatorname{coker}g$. – Servaes Dec 05 '18 at 21:42
0

1) No, not necessarily. Here's a counterexample. Let $B=\mathbb{Z}$ and $C=P=\mathbb{Q}$ as $R=\mathbb{Z}$-modules. Further, let $g:B\to C$, i.e. $g:\mathbb{Z}\to\mathbb{Q}$, be the inclusion map $g(n)=n$. Then for any $\varphi:C\to P$, i.e. for any $\varphi:\mathbb{Q}\to\mathbb{Q}$, $\varphi\circ g=0$ implies $\varphi=0$ (basically, because $\varphi\circ g=0$ implies $\varphi(1)=0$ implies $\varphi=0$). And yet, $g\neq0$.

2) No, not necessarily. Here's a counterexample. Let $A=B=C=\mathbb{Z}$ and $P=\mathbb{Z}_2$ as $R=\mathbb{Z}$-modules. Further, let $f:\mathbb{Z}\to\mathbb{Z}$ be the identity map $f(n)=n$ and $g:\mathbb{Z}\to\mathbb{Z}$ be multiplication by two map $g(n)=2n$. Then for any $\varphi:\mathbb{Z}\to\mathbb{Z}_2$. we have $\varphi\circ g\circ f=0$, even though $g\circ f\neq0$.

zipirovich
  • 14,670
  • 1
  • 26
  • 35