I'm giving a seminar talk on the ideal class group, and am looking for an example of a fractional ideal that is not invertible. Does know of a simple example in say $k[x,y]$ or $\mathbb{Z}[\sqrt{-3}]$. Such examples must exist but computation there is very hard.



A fractional ideal is invertible iff it is a projective module. To find an example of a non-invertible one it is therefore enough to exhibit an example of an ideal which is not projective.

A simple example is the ideal $(x,y)$ in $R=k[x,y]$.

Let us check this ideal is not projective «by hand» showing that the natural map $\phi:(a,b)\in R\oplus R\mapsto ax+by\in (x,y)$, which is surjective, does not admit a section $\psi:(x,y)\to R\oplus R$. We suppose otherwise, that is, that there is such a $\psi$, and we let $\psi(x)=(f,g)$ and $\psi(y)=(r,s)$. We have $0=\psi(xy-yx)=x\psi(y)-y\psi(x)$, so that $yf=xr$ and $yg=xs$. It follows that $r$ and $s$ are multiples of $y$, and $f$ and $g$ are multiples of $x$. Now, as $\phi\circ\psi$ is the identity map of $(x,y)$, we have $x=\phi(\psi(x))=\phi(f,g)=xf+gy$, but now we know that $xf+gy$ is in the ideal $(x^2,y)$, which does not contain $x$: this is impossible.

  • $\begingroup$ Mariano, is there a super easy way $(x,y)$ isn't projective? I know it's not even flat, by the classic computation of $\text{Tor}$. Even more ridiculous, is that it would have to be free by Quillen-Suslin, which is ridiculous since it's not principal. But, is there quick and easy way to see this directly? Thanks! :) $\endgroup$ – Alex Youcis Oct 22 '13 at 6:21
  • $\begingroup$ This was the calculation with Tor I was referring to. I'm wondering if there is a way to see the non projectivity without having to restort to a Tor calculation. Thanks! $\endgroup$ – Alex Youcis Oct 22 '13 at 6:25
  • $\begingroup$ One should be able to prove directly that the obvious map $k[x,y]^2\to I$ is not split. $\endgroup$ – Mariano Suárez-Álvarez Oct 22 '13 at 6:28
  • $\begingroup$ @AlexYoucis, I added a simle argument. $\endgroup$ – Mariano Suárez-Álvarez Oct 22 '13 at 8:32
  • 2
    $\begingroup$ Invoking Quillen-Suslin for this is absurd... $\endgroup$ – Mariano Suárez-Álvarez Oct 23 '13 at 5:21

Hint: Consider the ring $\mathbb{Z}[\sqrt{5}]$. Since this is a non-maximal order in the Dedekind domain $\mathbb{Z}[\varphi]$, you know it must be not integrally closed, and so must have a non-invertible fractional ideal.

To actually find one, a good place to start might be at $I=(2)$ (why?). Indeed, let $\mathfrak{P}=(2,1-\sqrt{5})$. Show that $\mathfrak{P}$ is maximal (just compute the quotient ring). Then show that $\mathfrak{P}\ne I$, but


  • $\begingroup$ What about the fractional ideal $J=2^{-1}R$ is that not a fractional ideal such that $(2)J=R$? $\endgroup$ – TheNumber23 Oct 23 '13 at 5:11
  • $\begingroup$ Yes, the non-invertible ideal is $\mathfrak{P}$... $\endgroup$ – Alex Youcis Oct 23 '13 at 7:51
  • $\begingroup$ Oh ok, sorry for the confusion. $\endgroup$ – TheNumber23 Oct 23 '13 at 17:01

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.