8
$\begingroup$

The categories of modules over commutative rings are especially notable Abelian categories. Wanting to extend this class a bit, I thought of this question:

Let $R$ be a commutative ring with $1$. Does there exist a Grothendieck topos $E$ such that $\mathrm{Ab}(E) \simeq R\text{-}\mathrm{Mod}$ (equivalence of categories)?

See the related question I was able to find: Is every Grothendieck category with a generator a category of sheaves?

Improve this question
$\endgroup$
2
  • 10
    $\begingroup$ If $E$ is a topos which has a point $x$, then the topos $Set$ is a retract (in the category of topoi) of $E$. It follows that $Ab = Ab(Set)$ is a retract of $Ab(E)$. So for instance $Ab(E)$ can't be $Vect_k$ for $k$ a field. I suspect that the hypothesis of $E$ having a point can be lifted by arguments along the lines here (or maybe the result there can be used directly by passing to hearts of $t$-structures). In short, I'm pretty sure the answer is no in general. $\endgroup$
    – Tim Campion ♦
    Feb 3 at 19:06
  • $\begingroup$ Here are some examples of commutative rings whose module categories are the abelian group objects in a topos. 1. If $M$ is a commutative monoid, then the category of $\mathbb ZM$-modules is equivalent to the category of abelian group objects in the topos of $M$-sets. 2. If $X$ is a compact totally disconnected space and $R$ is the ring of locally constant functions from $X$ to $\mathbb Z$, then the category of $R$-modules is equivalent to the category of sheaves of abelian groups on $X$, which is of course the abelian group objects in the topos of sheaves of sets on $X$. $\endgroup$
    – Benjamin Steinberg
    Feb 4 at 16:20

1 Answer 1

Reset to default
7
$\begingroup$

No — in particular, not if $R$ has any non-integer rationals.

Briefly: From $\newcommand{\Z}{\mathbb{Z}}\newcommand{\Ab}{\mathrm{Ab}}\newcommand{\RMod}{{R\text{-}\mathrm{Mod}}}\RMod$, we can recover $R$ as the ring of endomorphisms of the identity functor, $\newcommand{\End}{\mathrm{End}}\newcommand{\id}{\mathrm{id}}R \cong \End(\id_\RMod)$. But for a topos $\newcommand{\E}{\mathcal{E}}\E$, $\End(\id_{\Ab(\E)})$ retracts onto $\Gamma(\Z_\E)$ (global sections of the integers of $\E$); and this will never have any non-trivial rationals.

Unwinding this more elementarily: Suppose $n \in \Z$ is invertible in $R$, with $n \neq \pm 1$. Then multiplication by $\frac{1}{n}$ gives a natural endomorphism $\mu_{\frac{1}{n}}$ of the identity functor on $\RMod$, satisfying $n \cdot \mu_{\frac{1}{n}} = 1$ (where $1$ is the identity endomorphism of the identity functor).

But the $\mathbf{Ab}$-enrichment of any Abelian category is determined by the category structure (using the biproducts). So an equivalence $\RMod \cong \Ab(\E)$ would transfer $\mu_{\frac{1}{n}}$ to an endomorphism $\mu'$ of $\id_{\Ab(\E)}$ with $n \cdot \mu' = 1$; then applying this at $\Z_\E$, and taking global sections, would give an inverse for $n$ in $\Gamma(\Z_\E)$. But this can never exist unless $\E$ is the trivial topos, since as $n$ is non-unital you can prove constructively “$n$ is not invertible in $\Z$”, so this holds in the internal language of $\E$. So then $\Ab(\E)$ is also trivial, and the equivalence $\RMod \cong \Ab(\E)$ implies $R \cong 1$.

So we’ve shown: if an integer $n \neq \pm 1$ is invertible in $R$, and $\RMod \cong \Ab(\E)$, then $R$ is the zero ring.

Improve this answer
$\endgroup$
7
  • 1
    $\begingroup$ I think it is not true that the endomorphism of the identity gives the global section of the integer: consider the case of the topos of M-set for M a commutative monoid $\endgroup$
    – Simon Henry
    Feb 4 at 17:08
  • $\begingroup$ Great! Could you expand on why $n$ can't be invertible in $Ab(\mathcal E)$ unless $\mathcal E$ is trivial? $\endgroup$
    – Tim Campion ♦
    Feb 4 at 17:09
  • 1
    $\begingroup$ @TimCampion : you can proove constructively that integers are not invertible. $\endgroup$
    – Simon Henry
    Feb 4 at 17:13
  • 1
    $\begingroup$ @SimonHenry: Oops, good catch — yes, $\Gamma(\mathbb{Z})(\mathcal{E})$ isn’t generally isomorphic to $\mathrm{End}(\mathrm{id}_{\mathrm{Ab}(\mathcal{E})})$ as I’d claimed. But it’s still a retract, which is enough for the argument to go through. $\endgroup$
    – Peter LeFanu Lumsdaine
    Feb 4 at 17:27
  • 1
    $\begingroup$ @JensHemelaer: For the map $\newcommand{\End}{\mathrm{End}}\newcommand{\id}{\mathrm{id}}\newcommand{\Ab}{\mathrm{Ab}}\newcommand{\E}{\mathcal{E}}\newcommand{\Z}{\mathbb{Z}}\End(\id_{\Ab(\E)}) \to \Gamma(\Z_\E)$: given an endomorphism of $\id_{\Ab(\E)}$, apply it at $\Z_\E$ and take global sections to get an ab-gp endomorphism of $\Gamma(\Z_\E)$, then apply it to $1$. Conversely, any $A$ in $\Ab(\E)$ has a natural $\Z_\E$-action $\Z_\E \times A \to A$; so a global section $s : 1 \to \Z_\E$ gives a natural map $s_A : A \to A$. Checking these give a retraction is a follow-your-nose computation. $\endgroup$
    – Peter LeFanu Lumsdaine
    Feb 7 at 10:17

Your Answer

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

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