2
$\begingroup$

I was looking at these notes on Tannakian categories. Let me briefly recall the notion of tensor functors: Let $(\mathcal{C},\otimes)$ and $(\mathcal{C'},\otimes')\DeclareMathOperator{\uphom}{\underline{Hom}} \DeclareMathOperator{\hom}{Hom}$ be two rigid tensor categories. (He defines the notion of rigid tensor categories in page 2 in his notes). A tensor functor $(\mathcal{C},\otimes)\rightarrow (\mathcal{C'},\otimes')$ is a pair $(F,c)$ where $F:\mathcal{C}\rightarrow \mathcal{C'}$ is a functor and $c$ is a collection of isomorphisms $c_{X,Y} : F(X)\otimes F(Y)\rightarrow F(X\otimes Y)$, functorial in $X$ and $Y$, that “commute with the associativity and commutativity constraints” in the obvious way, and send identity objects to identity objects.

Next, the author states the following fact, which is not clear to me:

Claim: One checks that if $\mathcal{C}$ and $\mathcal{C'}$ are rigid, then the axioms force the induced map $F(\uphom(X,Y))\rightarrow \uphom(FX,FY)$ is an isomorphism, where $\uphom(X,Y)$ is the object that represents the functor $\mathcal{C}^{op}\rightarrow\operatorname{Sets}$ $$T\mapsto \hom(T\otimes X, Y)\,.$$

Could someone please explain/ suggest some reference where I can find a proof of this fact?

Thanks in advance!

$\endgroup$

2 Answers 2

5
$\begingroup$

Your $F$ preserves dual objects because it preserves the duality situations: quadruples $(X,Y,\alpha : X\otimes Y \rightarrow I, \beta: I \rightarrow Y \otimes X)$.

It remains to recall the formula for the internal hom: $$ \underline{\mathrm{Hom}} (X,Y) = X^\ast \otimes Y. $$

$\endgroup$
4
  • 1
    $\begingroup$ I'm new into these stuff, I would be grateful if you could explain a bit more. I am unfamiliar with the notion of 'duality situations', could you please elaborate this part? $\endgroup$ Commented Sep 7, 2023 at 13:53
  • $\begingroup$ @Hajime_Saito Look up "rigid monoidal category." Bugs is giving a shorthand for an evaluation map α, a coevaluation map β, and dual objects X and Y. Monoidal functors preserve duality. $\endgroup$ Commented Sep 7, 2023 at 21:45
  • $\begingroup$ See Definition 2.1 in ncatlab.org/nlab/show/dualizable+object $\endgroup$
    – Bugs Bunny
    Commented Sep 8, 2023 at 11:57
  • $\begingroup$ Hi, i’d be really happy if you could add why this is enough (together with the isomorphism you write in the last sentence)? I know how to show that duality is preserved, but then I get stuck. $\endgroup$
    – Ben123
    Commented Jul 15 at 5:38
3
$\begingroup$

$\DeclareMathOperator{\uphom}{\underline{Hom}}$

Looking at the notes, I can see that $\uphom(X,Y)$ is the internal hom of the category.

This has the universal property that that maps into it $ Z\to \uphom(X,Y)$ are in bijection with maps $ Z\otimes X \to Y$.

A suitable proof of your statement might go like this:

The map $$ F(\uphom(X,Y))\to \uphom(FX,FY)$$ corresponds to a map (from property above) $$ F(\uphom(X,Y))\otimes FX\to FY$$ which in turn corresponds to a map (from $F$ being a tensor functor) $$ F(\uphom(X,Y)\otimes X)\to FY$$

If the counit of the adjunction $-\otimes X \dashv \uphom(X,-)$ is called $ev:\uphom(X,Y)\otimes X\to Y$, then one may check that the map above is $F(ev)$. And since the isomorphism between homs defined by the adjunction takes (naturally) its counit to the identity morphim on $\uphom(X,Y)$. Then the functor $F$ must take $ev$ to an isomorphism $F(ev)$.

--- edit ---

Here are the corners of a relevant square depicting the chase of $ev$. Where square brackets denote the external hom sets. $$ [\uphom(X,Y)\otimes X, Y] \cong [\uphom(X,Y),\uphom(X,Y)] $$ $$ [F(\uphom(X,Y)\otimes X), F(Y)] \cong [F(\uphom(X,Y)),F(\uphom(X,Y))] $$ There should be vertical arrows with labels $F$ which I'll won't draw now. :)

$\endgroup$
4
  • $\begingroup$ Hm, could you please elaborate a little on this answer? I am with you until the last two sentences, where I get lost. $\endgroup$
    – Ben123
    Commented Jun 10 at 23:48
  • $\begingroup$ I edited the last sentence to be a bit more explanatory. Also added the corners of what should be a commutative square. You can chase $ev$ and $id$ and $F(ev)$ and the desired isomorphism that I refer to in the previous paragraph. $\endgroup$ Commented Jun 11 at 6:19
  • $\begingroup$ Thanks for the update, I am still unsure about it, I guess I have to read up on adjunctions. But I appreciate the effort. =) $\endgroup$
    – Ben123
    Commented Jun 11 at 11:02
  • $\begingroup$ Hm, tell me his, please sir; why does this square commute (i.e. by which property)? I am referring to the one in your edit. $\endgroup$
    – Ben123
    Commented Jun 14 at 9:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .