Let $\mathcal{A}$ be a $k$-linear abelian symmetric tensor category with unit $\mathcal{O}_A$; here $k$ is a comm. ring. By that I assume implicitly that $\otimes$ is finitely cocontinuous in each variable.

Assume that $\mathcal{L}$ is a line object on $\mathcal{A}$, i.e. it is invertible with respect to $\otimes$ and the symmetry $\mathcal{L} \otimes \mathcal{L} \cong \mathcal{L} \otimes \mathcal{L}$ is the identity.

Question: Is every epimorphism $\mathcal{O}_A \to \mathcal{L}$ an isomorphism?

Remark that this is true in the following two cases:

a) $\mathcal{A}$ is a category of modules over a ring, or a category which "locally" looks like this (for example $\text{Qcoh}(X)$ for some scheme $X$).

b) $2$ is invertible in $k$ (which I learned from t3suji).

Is it true in general, or is there a nasty counterexample over $k=\mathbb{F}_2$? What about the special case $\mathcal{L} = \mathcal{O}_A$?


Assuming that $\otimes$ preserves finite colimits in each separate argument, I think it's true in general.

Since it's very easy to get twisted around here, let me abstract the situation a bit. Recall that if $\alpha: F' \to F$ is a map between left adjoints in a bicategory (with right adjoints $G'$ and $G$ respectively), then the mate $\alpha^\ast: G \to G'$ is defined by the composite

$$G \stackrel{\eta' G}{\to} (G'F')G \stackrel{(G'\alpha) G}{\to} (G'F)G \cong G'(FG) \stackrel{G'\varepsilon}{\to} G'$$

where $\eta'$ and $\varepsilon$ are the appropriate unit and counit of adjunctions. Here of course we are thinking of our monoidal category as a one-object bicategory.

If $\pi: \mathcal{O} \to L$ is an epi, then I claim its mate $\pi^\ast: L^{-1} \to \mathcal{O}$ is a mono. For, if $f, g: A \to L^{-1}$ are distinct arrows, then the two composites

$$A \stackrel{\overset{f}{\to}}{\underset{g}{\to}} L^{-1} \stackrel{\pi^\ast}{\to} \mathcal{O}$$

are equal to the composites

$$A \cong \mathcal{O} \otimes A \stackrel{\pi \otimes A}{\to} L \otimes A \stackrel{\overset{L \otimes f}{\to}}{\underset{L \otimes g}{\to}} L \otimes L^{-1} \stackrel{\varepsilon}{\cong} \mathcal{O}$$

by the usual yoga of mates and adjunctions; note that $L \otimes f$ and $L \otimes g$ are distinct since tensoring with $L$ is an equivalence, and note that $\pi \otimes A$ is an epi because tensoring with $A$ is right exact. So these two composites are distinct. Hence $\pi^\ast$ is monic.

On the other hand, the mate $\pi^\ast$ is also epic, because it is given by the composite

$$L^{-1} \cong \mathcal{O} \otimes L^{-1} \stackrel{\pi \otimes L^{-1}}{\to} L \otimes L^{-1} \stackrel{\varepsilon}{\cong} \mathcal{O}$$

where the middle arrow is epi again because tensoring with an object is right exact.

Since $\pi^\ast$ is mono and epi in an abelian category, it is an isomorphism, and therefore its mate $\pi$ is also an isomorphism.


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.