Let $\mathsf{C}$ be a rigid abelian tensor category in the sense of Deligne and Milne's notes (p.9). Let $\mathbf{1}$ denote the identity object in $\mathsf{C}$ with respect to $\otimes$. The abelian group $\text{End}(\mathbf{1})$ has the structure of a commutative ring (see discussion immediately after Definition $1.15$). Let us suppose that this ring is actually a field, so by Remark $1.18$, $\mathbf{1}$ is a simple object. The proof of Proposition $1.19$ says that there is a criterion in $\mathsf{C}$:

$$X\neq 0 \iff X\otimes X^\vee \to \mathbf{1} \text{ is an epimorphism}.$$

Here the morphism $X\otimes X^\vee \to \mathbf{1}$ is the morphism obtained from $\operatorname{id}_{X^\vee}$ via the commutativity constraint $X\otimes X^\vee\cong X^\vee\otimes X$ and the bijection

$$\text{Hom}(X^\vee \otimes X, \mathbf{1})\cong\text{Hom}(X^\vee, X^\vee).$$

I have shown that if this morphism is an epimorphism then $X\neq 0$, but I'm having trouble working through the other implication (and I'm fairly sure this "criterion" wasn't stated previously). Could anyone show me how to prove that if $X\neq 0$ then the morphism is epic? Many thanks!

Edit: Now suppose $(\mathsf{C}', \mathbf{1}')$ is another rigid abelian tensor category with $\mathbf{1}\neq 0$ and let $F: \mathsf{C}\to\mathsf{C}'$ be an exact tensor functor. I must confess that even assuming the above criterion I am slightly mystified as to how this can be used to prove Proposition $1.19$:

$F$ is faithful

I understand that $F$ preserves the criterion, but I'm not sure how to relate this to showing it is injective on hom-sets between objects. Any tips on this would be greatly appreciated!


For an exact functor $F$ between abelian categories, being faithful is equivalent to the condition $$FX=0\Rightarrow X=0$$ on objects, since a morphism is zero if and only if its image is zero, and exact functors preserve images.

So, for your second question, the "criterion" proves this equivalent formulation of faithfulness.

  • 1
    $\begingroup$ This answer, combined with working through the details of the proposition, has taught me a lot of basic but useful things about abelian categories and it is nice to see how the tensor/abelian structures interact. Thank you! $\endgroup$ – Alex Saad Apr 24 '16 at 10:29

I think this answers the first part of my question (before the edit):

Suppose that the map $ev:X\otimes X^\vee \to \mathbf{1}$ is not an epimorphism. Then there exists an object $Z$ and distinct morphisms $\alpha, \beta: \mathbf{1}\to Z$ such that $\alpha\circ ev = \beta\circ ev$. Thus $ev: X\otimes X^\vee \to \mathbf{1}$ factors through the monomorphism $\ker(\alpha-\beta)\hookrightarrow \mathbf{1}$. Since $\text{End}(\mathbf{1})$ is a field, $\mathbf{1}$ is a simple object and therefore $\ker(\alpha-\beta)$ is isomorphic to either $\mathbf{1}$ or $0$ (as subobjects of $\mathbf{1}$ correspond to idempotents in $\text{End}(\mathbf{1})$). But if $\ker(\alpha-\beta) \cong \mathbf{1}$ then $\alpha-\beta$ is the zero map, which cannot happen since $\alpha\neq\beta$.

Hence $\ker(\alpha-\beta) = 0$ and so $ev: X\otimes X^\vee \to\mathbf{1}$ factors through the zero map $0\to\mathbf{1}$. By the tensor-hom adjunction, $\text{Hom}(X\otimes X^\vee, \mathbf{1})\cong \text{Hom}(X^\vee, X^\vee)$ is an isomorphism of abelian groups. Recall that $ev$ was defined as the image of $\text{id}\in\text{Hom}(X^\vee, X^\vee)$ under this adjunction, which implies $id_{X^\vee} = 0$. But then $X^\vee = 0$, so $X \cong (X^\vee)^\vee = 0$.


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.