For a module $M$ over PID which is not finitely generated, do we still have the isomorphism $M\cong Tor(M)\oplus M/Tor(M)$? If not can you give a counter-example?

  • 1
    $\begingroup$ @Bernard: That presumably means the torsion submodule, the set of $m\in M$ that are annihilated by some nonzero element of the ring. $\endgroup$ – Eric Wofsey Jul 26 '18 at 22:25
  • $\begingroup$ @EricWofsey; Why, yes of course! $\endgroup$ – Bernard Jul 26 '18 at 22:28

No. For instance, over the ring $\mathbb{Z}$, let $M=\prod\mathbb{Z}/(p)$ where the product ranges over all primes $p$. Then the torsion subgroup $T$ of $M$ is just the direct sum $\bigoplus \mathbb{Z}/(p)$, set set of elements of $M$ which have only finitely many nonzero coordinates. The quotient $M/T$ is then divisible, since for any $m\in M$ and any nonzero $a\in\mathbb{Z}$, we can divide $m$ by $a$ after modifying only finitely many coordinates of $m$ (namely, those corresponding to primes that divide $a$).

However, the only element of $M$ which is divisible by every nonzero integer is $0$, since if $m\in M$ is divisible by a prime $p$ then its $p$-coordinate must be $0$. So $M$ has no nontrivial divisible subgroups, and in particular it does not have a subgroup isomorphic to $M/T$.


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.