7
$\begingroup$

Let $R$ be a noetherian domain and let $\mathcal{O}$ be an $R$-algebra that is finitely generated and projective as an $R$-module. The set of invertible fractional ideals of $\mathcal{O}$ is a group under multiplication. Is this group always abelian?

$\endgroup$
7
  • 4
    $\begingroup$ It looks like some of your post got cut off. I also think you meant $R$-module, rather than $\mathcal{O}$-module. Also, what is the definition you use for an invertible ideal in a non-commutative ring? $\endgroup$ Apr 8, 2015 at 15:25
  • 4
    $\begingroup$ What's a fractional ideal at all in a noncommutative ring? $\endgroup$ Apr 8, 2015 at 21:34
  • 2
    $\begingroup$ @FernandoMuro $K$ being the field of fractions of $R$ and $A = K\otimes_R\mathcal{O}$, I would guess a fractional ideal is a lattice $I\subset A$ (finitely generated $R$-submodule with $KI=A$) with left and right order equal to $\mathcal{O}$. Does this work? $\endgroup$
    – Aurel
    Apr 8, 2015 at 23:09
  • 3
    $\begingroup$ @Aurel: conventions differ, but you might ask only that $\mathcal{O}$ is contained in the left and right orders; if $I$ is invertible, then equality holds, so this doesn't matter for the question. (@FernandoMuro: It is also equivalent to take a fractional ideal $I$ to be a $\mathcal{O}$-sub-bimodule of $K \otimes_R \mathcal{O}$ of the form $I=cJ$ where $c \in K^\times$ and $J \subseteq \mathcal{O}$ is a two-sided ideal. Hence the ``fractional''.) $\endgroup$ Apr 9, 2015 at 2:00
  • 3
    $\begingroup$ @FernandoMuro The base ring $R$ is assumed to be commutative: it is a domain. So the field of fractions means what it usually does. (There are notions of rings of quotients for noncommutative rings, but they don't play a role here.) $\endgroup$ Apr 9, 2015 at 14:24

1 Answer 1

7
$\begingroup$

No; it is not abelian in general.

Let $R$ be a discrete valuation ring with maximal ideal generated by an element $\pi$, let $k := R / \pi R$ be the residue field of $R$ and let $\mathcal{O}$ be the inverse image in $M_2(R)$ of the scalar matrices in $M_2(k)$. Note that $\mathcal{O}$ is free of rank $4$ as an $R$-module, with basis $\{e_{11} + e_{22}, \pi e_{11}, \pi e_{12}, \pi e_{21}\}$ say. So $\mathcal{O}$ is, in particular, finitely generated and projective as an $R$-module.

For every $g \in GL_2(R)$, you have the invertible fractional ideal $\mathcal{O}g$. Note that $\mathcal{O}g \cdot \mathcal{O} h = \mathcal{O} gh$ since $GL_2(R)$ normalises $\mathcal{O}$ inside $M_2(R)$. Note also that $\mathcal{O} g = \mathcal{O} h$ if and only if $gh^{-1} \in \mathcal{O}^\times$, which is the preimage in $GL_2(R)$ of the group of scalar matrices in $GL_2(k)$. So we obtain an injective group homomorphism from $PGL_2(k) \cong GL_2(R) / \mathcal{O}^\times$ into the group $I(\mathcal{O})$ of invertible fractional ideals of $\mathcal{O}$.

In the positive direction, if $K$ is the field of fractions of a Dedekind domain $R$ and if $\mathcal{O}$ is a maximal order in some finite dimensional separable $K$-algebra, then $I(\mathcal{O})$ is abelian.

See Theorem 12 and Theorem 13(c) of this paper by Frőhlich, which gives a thorough discussion of the Picard group of a noncommutative ring.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.