7
$\begingroup$

In the book 'Tensor Categories' by Pavel Etingof, Shlomo Gelaki, Dmitri Nikshych and Victor Ostrik on page 10 it says:

'Conversely, it is well known (and easy to show) that any exact faithful functor $F : \mathcal{C} \rightarrow \text{Vec}$ is represented by a unique (up to a unique isomorphism) projective generator $P$.'

But I could not find any proof of that fact. Can someone tell me how to prove it or where I can find a proof?

Note: Here $\mathcal{C}$ is a finite k-linear abelian category for some field k.

$\endgroup$
3

1 Answer 1

5
$\begingroup$

Let $\mathcal C$ be the category of finite dimensional left modules over a finite dimensional ring $R$. Let $G: \mathcal C \to \mathrm{Vec}$ be an exact and faithful functor to finite dimensional vector spaces. We use $V^*$ to denote the dual vector space. For motivation, notice that if we had a representing object $M$, we would have $$G(R^*) = Hom_R(M,R^*) = Hom_R(R, M^*) = M^*.$$

Now $G(R^*)$ is a right $R$ module via the action of $R$ by left multiplication. So we define $P:= G(R^*)^*$ to be the dual left module and consider the functor $Hom_R(P,-)$.

This functor is tautologically left exact, and it takes the injective left module $R^*$ to $Hom_R(P,R^*) = Hom_R(R, G(R^*)) = G(R^*)$. Any other finite left module $M$ admits an injective presentation $$0 \to M \to R^{* \oplus a} \to R^{* \oplus b} \to $$ dual to the presentation of $M^*$ as a right $R$ module. So by left exactness, we see that there is a natural isomorphism $Hom_R(P,-) \simeq G(-)$.

Thus $G$ is represented by $P$. Since $G$ is right exact $P$ is projective, and since it is faithful $P$ is a generator.

$\endgroup$
5
  • $\begingroup$ Assuming that by $G(R)^*$ you mean the vector space dual, then $P=G(R)^*$ is not usually projective. Even when $G=\text{Hom}_R(R,-)$ is the forgetful functor, $G(R)^*\cong R^*$ which is not projective unless $R$ is self-injective. $\endgroup$ May 11, 2020 at 7:02
  • $\begingroup$ @JeremyRickard Thanks, I was missing a dual. I edited the post to fix the mistake-- I think it is correct now. $\endgroup$ May 11, 2020 at 15:28
  • $\begingroup$ Why is the Hom-functor tautologically right exact? I thought since Hom is right adjoint to the tensor functor, it usually just is left exact. $\endgroup$
    – S.Farr
    May 11, 2020 at 19:01
  • $\begingroup$ @S.Farr Sorry, I wrote it backwards-- I am using left exactness in the argument. $\endgroup$ May 11, 2020 at 19:03
  • $\begingroup$ @PhilTosteson, good, then it makes sense. I think you also need to change the 'left exact' to 'right exact' in the last line. $\endgroup$
    – S.Farr
    May 11, 2020 at 19:07

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.