Let $R=k[x_1,\ldots,x_n]$. According to the first answer, every finitely generated flat module over an integral domain is necessarily projective.

Therefore, the only hope to find a flat non-projective $R$-module $M$ is when $M$ is not finitely generated. Can anyone please suggest such modules?

Actually, here there are some examples of flat non-projective modules (but not over a polynomial ring), and the following general claim, due to Bass: Flat modules are projective iff the ring is perfect.

(A somewhat relevant question can be found here).

EDIT: I have posted here my last comment as a question, which already has an answer.


migrated from mathoverflow.net Jul 28 '15 at 20:27

This question came from our site for professional mathematicians.

  • 3
    $\begingroup$ Projective $R$-modules are not divisible. Every projective $R$-module is a direct summand of a free $R$-module, and every nonzero element in a free $R$-module obviously fails to be divisible. Thus, a divisible, flat $R$-module, e.g., the fraction field, is not projective. $\endgroup$ – Jason Starr Jul 28 '15 at 0:40
  • $\begingroup$ Thanks!! So given any integral domain $D$, the field of fractions of $D$ is a a non-f.g. non-projective flat module over $D$? I am curious if there exist other examples? (I guess yes). BTW, if you would like to post your comment as an answer, I will accept it. $\endgroup$ – user237522 Jul 28 '15 at 1:34
  • $\begingroup$ How did we end up here, user237522? Somebody out there does not appreciate your question (or maybe they do not appreciate my answer). $\endgroup$ – Jason Starr Jul 29 '15 at 3:07
  • $\begingroup$ @JasonStarr I do not know :-) I think it is more plausible that someone did not appreciate my question (than your answer). It seems to me that the claim in your answer ("divisible implies non-projective") is not so known (at least not to me, though I am not an expert) and hence appropriate for MO. $\endgroup$ – user237522 Jul 29 '15 at 10:42
  • 1
    $\begingroup$ Well, it is better to learn, to discover, and to have fun than it is to be appreciated. I hope that you have fun learning about some algebra. $\endgroup$ – Jason Starr Jul 29 '15 at 13:11

Let $R$ be an integral domain that is not a field. Then $R$ contains a nonzero element that is not invertible. For every nonzero element $r$ that is not invertible, $r$ is not a multiple of $r^2$: if it were, then $r=r^2s$ for some $s\in R$. Since $R$ is an integral domain, $1=rs$, contradicting that $r$ is not invertible.

Now, for $I$ any nonzero ideal in $R$, if $I$ equals $R$, then $I$ is only divisible by invertible elements of $R$. By hypothesis, there exists a nonzero, noninvertible element $q$ of $R$, and $I$ is not divisible by $q$. On the other hand, if $I$ does not equal $R$, then $I$ contains a nonzero element $r$ that is not invertible. Thus $r$ is not divisible by $q=r^2$. Since $r$ is not divisible by $q$, also $I$ is not divisible by $q$. Thus, for every nonzero ideal $I$ in $R$, there exists a nonzero, noninvertible element $q$ of $R$ such that $I$ is not divisible by $q$.

Finally, for any free $R$-module $M$, say free on a basis $S$, for every nonzero element $m$ of $M$, the ideal $I$ generated by the finitely many nonzero $S$-coefficients of $m$ is nonzero. Thus there exists a nonzero element $q$ of $R$ such that $I$ is not divisible by $q$. Therefore also $m$ is not divisible by $q$. Therefore, also, for every nonzero $R$-submodule $P$ of $M$, for every nonzero element $m$ of $P$, there exists nonzero, noninvertible $q$ in $R$ such that $m$ is not divisible by $q$. In particular, if $P$ is a direct summand of $M$, then $P$ is not divisible.

Thus, every divisible $R$-module is not projective. In particular, the fraction field of $R$ is divisible. Therefore the fraction field of $R$ is a flat $R$-module that is not projective.

  • $\begingroup$ Thank you very much for your answer; the elaboration really helped me. I wonder if there exists an example of a non-finitely generated non-projective flat $R$-module which is not divisible, where $R$ is an integral domain. I guess there is such an example? and what about the special case $R=k[x_1,\ldots,x_n]$?. (Should I post this as a separate question?). $\endgroup$ – user237522 Jul 28 '15 at 23:43

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.