# PID and finitely generated module

I am trying to prove the following statements:

Let $R$ be a PID and $M$ a finitely generated $R$-module. Prove:

(a) $M$ is torsion module iff $\operatorname{Hom}_R(M,R)=0$

(b) $M$ is an indecomposable module iff $M \cong R$ or $\exists \space p \in R$ irreducible and $n \in \mathbb N$ such that $M \cong R/ \langle p^n \rangle$

I got stuck with one implication in (a) and I don't know what to do in (b).

In (a), if $M$ is a torsion module, then take $f \in \operatorname{Hom}_R(M,R)$. Suppose $\{x_1,...,x_n\}$ is a set of generators of $M$, let $a_i \in R \setminus \{0\} : a_ix_i=0$, now take $x \in M$, we have $x=b_1x_1+\cdots+b_nx_n$. Define $a=\prod_{i=1}^na_i$, then $$0=f(ax)=af(x)$$Since $R$ is an integral domain and $a \neq 0$, it must be $f(x)=0$. It follows $f=0$.

I would appreciate suggestions for the other direction and for part (b). Thanks in advance.

• Are you allowed the structure theorem? That gives you a decomposition of $M$ and it's somewhat clear what all the homomorphisms out of that look like. – Hoot Nov 14 '14 at 3:13

## 2 Answers

(a) As you probably know $M=t(M)\oplus F$, where $t(M)$ is the torsion submodule of $M$ and $F$ is free of finite rank. Then $\operatorname{Hom}(M,R)=\operatorname{Hom}(t(M)\oplus F,R)\simeq \operatorname{Hom}(t(M),R)\oplus \operatorname{Hom}(F,R)$. As you already proved $\operatorname{Hom}(t(M),R)=0$, while $\operatorname{Hom}(F,R)\neq 0$ unless $F=0$.

(b) If $M=R$ then it is indecomposable: if $R=I+J$ with $I\ne 0$ and $J\ne 0$, then $0\ne IJ\subseteq I\cap J$, so we can't have $I\cap J=0$.
If $M=R/(p^n)$ then it is indecomposable: the ideals of $R/(p^n)$ are $(p^i)/(p^n)$ with $i=0,1,\dots,n$, so they can intersect trivially.

If $M$ is indecomposable, then from $M=t(M)\oplus F$ we must have $M=t(M)$ or $M=F$. If $M=F$, then $F$ is of rank one, so $M\simeq R$. If $M=t(M)$, then $M$ must be cyclic. (Here I've used the structure theorem.) So $M\simeq R/(a)$. If $a=bc$ with $(b,c)=1$, then $R/(a)\simeq R/(b)\oplus R/(c)$, a contradiction. It follows that $a=p^n$, a power of a prime element.

• In (b) I assume that you meant "if $R=IJ$". I could follow your answer except for this affirmation: "the ideals of $R/(p^n)$ are $(p^i)/(p^n)$ with $i=0,1,...,n$", I couldn't prove this is true, would you help me with that? – user16924 Nov 14 '14 at 15:35
• Oh, and also, in "If $M=F$, then $F$ is of rank one...", why can't $F \cong R^m$ with $m>1$? – user16924 Nov 14 '14 at 15:40
• @user16924 In (b) I suppose that $R=I+J$ and $I\cap J=0$. The ideals of a quotient ring $R/I$ are of the form $J/I$ with $J\supseteq I$. Since $R$ is principal, $J=(a)$ and from $(a)\supseteq (p^n)$ we get $a\mid p^n$. If $F$ is free, $F=R^m$ with $m>1$, then $F$ is clearly decomposable: think to the simplest case $m=2$ and see that $R^2=(R\times \{0\})+(\{0\}\times R)$, $(R\times \{0\})\cap(\{0\}\times R)=\{(0,0\}$. – user26857 Nov 14 '14 at 16:38
• Could you suggest me how to show that the ideals of $R/(p^n)$ are of the form $(p^i)/(p^n)$? – user16924 Nov 15 '14 at 1:27
• @user16924 This is explained in the comment above, the second and third sentences. (Sorry for not making this clear enough.) – user26857 Nov 15 '14 at 1:39

Let $k$ be the field of fractions of $R$. I claim that $M$ is torsion iff $k\otimes_R M=0$. One direction is clear. Now suppose $M$ is f.g. and pick a set of generators $x_1,\ldots,x_n$. We can order the $x_i$ such that $\{x_1,\ldots,x_r\}$ is l.i. and such that $\{x_1,\ldots,x_r,x_k\}$ is l.d. for each $k>r$. In particular there is $a_k$ nonzero such that $a_kx_k\in \langle x_1,\ldots,x_r\rangle$. Let $a=a_1\cdots a_k$. We have a morphism $\eta:M\to F$ where $F\simeq R^r$ given by $x\mapsto ax$. You should check that $\ker \eta={\rm tor}\,M$. Thus $\overline M=M/{\rm tor}\,M$ injects into the free $R$-module $F$, since $R$ is a PID, we know that $\overline M$ is free. But we have a surjective morphism $\pi:M\to \overline M$ that sends $x$ to its class modulo ${\rm tor}\; M$. Since $\overline M$ is free, we have a splitting $M=\ker\pi\oplus F'$ where $F'\subseteq M$ is free of rank $r$ and $\ker \pi={\rm tor}\,M$. If $M$ is not torsion then $r>0$, and $k\otimes_R M\simeq k^r\neq 0$. Can you take it from here? This in fact gives that $M={\rm tor}\,M\oplus F$ with $F\simeq R^r$, $r=\dim_k(k\otimes_R M)$, so that ${\rm tor}\, M\neq M\implies r>0$ and hence ${\rm Hom}_R(M,R)\simeq {\rm Hom}_R(R^r,R)$ since you've already shown the torsion part vanishes.

N.B.: Here we're using the "big gun" that submodules of free modules over PIDs are free.

• (For the part $(b)$ I am hoping you can use the structure theorem) – Pedro Tamaroff Nov 14 '14 at 3:43
• I am trying to understand your answer but I have two doubts: I don't understand what $k\otimes_R M$ is, and when you wrote $A^r$, did you mean $k^r$ or $R^r$? – user16924 Nov 14 '14 at 4:02
• @user16924 If you haven't heard about tensor products, think of it as the localization of $M$ at the prime ideal $(0)$. I meant $R^r$, sorry. – Pedro Tamaroff Nov 14 '14 at 4:16
• Hmm, I am just learning this stuff, haven't read about tensors and localization (yet). – user16924 Nov 14 '14 at 4:23
• @user16924 In time then, you'll come back to this answer and understand it. – Pedro Tamaroff Nov 14 '14 at 5:22