We know that for a PID $R$, any finitely generated module is of the form $\frac{R}{(a_1)} \oplus \dots \oplus \frac{R}{(a_s)} $. I was wondering if the converse of this statement is true, that is, is it true that for a domain $R$, if any f.g. module is isomorphic to $\frac{R}{I_1} \oplus \dots \oplus \frac{R}{I_s}$ for ideals $I_i \triangleleft R $, must $R$ be a PID?

I know that a counter example to the above statement must be non Noetherian (and infact, it must be that any non-principal ideal is not f.g.) because if $I = (a_1, ..., a_n)$ is a non principal f.g. ideal, that it is a torsion free $R$ module, but it isn't free, since $x_1, ... x_m$ are not linearly independent if $m > 1$ since $x_2 x_1 + (-x_1)x_2 = 0$, hence they do not form a basis.

PS: I would generally like $R$ to be commutative unitial and an integral domain, but I would love to know even for other cases.

  • 3
    $\begingroup$ An íntegral domain in which every finitely generated ideal is principal is called a Bezout domain. It is apparently am open question of the elementary divisor theorem os true in all Bezout domains see alpha.math.uga.edu/~lorenz/Bezout.pdf $\endgroup$ – Benjamin Steinberg Jan 24 '19 at 21:27
  • $\begingroup$ @AnweshRay Yes, and this is what I stated in the second paragraph. The question regards weirder rings then, where the only non-principal ideals are those which are not f.g. and that there is at least one such ideal. $\endgroup$ – Adi Ostrov Jan 24 '19 at 23:20
  • 1
    $\begingroup$ As Benjamin Steinberg has hinted, a reasonable conclusion here should be that $R$ is a Bezout domain, rather than a PID. One of the first examples of a non-PID Bezout domain, $k[x^{1/p^{\infty}}]$, is an elementary divisor ring that is not a PID, hence provides a counterexample to the statement. $\endgroup$ – Victor Protsak Jan 25 '19 at 4:14
  • $\begingroup$ Possible duplicate: mathoverflow.net/questions/31275 $\endgroup$ – Victor Protsak Jan 25 '19 at 6:30
  • 3
    $\begingroup$ @VictorProtsak I don’t think that possible duplicate is the same question. I don’t see why having Smith normal form implies anything about the structure of finitely generated, but not finitely presented, modules. $\endgroup$ – Jeremy Rickard Jan 25 '19 at 8:20

There seems to be quite some literature about rings with this property, sometimes under the name "FGC domains".

From Googling, not personal knowledge:

In Theorem 14 of Kaplansky, Irving, Modules over Dedekind rings and valuation rings, Trans. Am. Math. Soc. 72, 327-340 (1952). ZBL0046.25701, it is proved that an almost maximal valuation domain has this property.

"Maximal" means that a system $x\equiv a_i\pmod{I_i}$ of congruences has a solution if each pair of the congruences has a solution, and "almost maximal" is the same, but only for such systems of congruences with $\bigcap_iI_i\neq0$. I'm not sure what the simplest example is that is not a PID, but I guess the completion of $R=k[x^\alpha\mid 0<\alpha\in\mathbb{Q}]$ with respect to the set of ideals $x^\alpha R$ is a maximal valuation domain.

Osofsky constructed an example of an FGC domain that is neither a PID nor a valuation domain, and work of many people, culminating in Brandal, Willy; Wiegand, Roger, Reduced rings whose finitely generated modules decompose, Commun. Algebra 6, 195-201 (1978). ZBL0368.13005, led to the theorem that FGC domains are precisely the almost maximal Bezout domains.

And there's a whole Springer Lecture Notes volume by Brandal on the not-necessarily-domain case:

Brandal, Willy, Commutative rings whose finitely generated modules decompose, Lecture Notes in Mathematics. 723. Berlin-Heidelberg-New York: Springer-Verlag. 116 p. (1979). ZBL0426.13004.


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.