4
$\begingroup$

This is a follow-up to this question: Reduction to graph subgroups for Bredon homology when the $G_1\times G_2$ is $G_2$-free

In his (very nice) answer Gregory Arone stated the following fact. Let $Q:\mathcal{C}_0\to\mathcal{C}$ be a functor between locally small categories. Then there is adjunction between functor categories $$ Q_!\colon[\mathcal{C}_0,\mathcal{D}]\leftrightarrows[\mathcal{C},\mathcal{D}]\colon Q^\ast, $$ where $Q^\ast$ is a precomposition with $Q$ and $Q_!$ is the left Kan extension (I hope I have the correct notation). Now if we have functors $F\colon\mathcal{C}_0\to\mathcal D$ and $G\colon\mathcal{C}^{op}\to\mathcal{D}$, then there is an isomorphism of coends: $$ F\otimes_{\mathcal{C_0}}Q^\ast G\cong Q_!F\otimes_{\mathcal{C}}G. $$

So my question is: how to prove this isomorphism? And is it something coming from having just an djoint pair of functors, or it applies only to this particular setting - i.e., we have a functor and its left Kan extension?

P.S. If this is a simple fact, reference will also do.

$\endgroup$
4
  • $\begingroup$ Do you know the universal property of $F\otimes_{C_0}H$? Namely, $\hom(F\otimes_{C_0}H, d)\cong \hom(F, \mathrm{Hom}(H,d))$ where $\mathrm{Hom}$ is the internal hom of $D$ (provided it exists, let me assume all the niceties I could want for $D$) $\endgroup$ Commented Jun 25, 2021 at 8:12
  • $\begingroup$ Well, I would be happy with some reference - because on the LHS here I have $\hom$ in the category $\mathcal{D}$, but on the RHS I have $\hom$ in what category? This should be natural transformation of functors over $\mathcal{C}_0$, so what is the functor in the codomain? $\endgroup$ Commented Jun 25, 2021 at 8:20
  • $\begingroup$ $\mathrm{Hom}(H,d)$ would be defined by $c\mapsto \mathrm{Hom}(H(c),d)$. I don't know a reference, though. $\endgroup$ Commented Jun 25, 2021 at 8:28
  • $\begingroup$ Ok, I see. Having this statement would be very nice, and there is a post on Math.SE where somebody prove that coends admit right adjoint - but then it still requires showing that this right adjoint is indeed internal hom in $\mathcal{D}$. I was looking for reference on this, but also wasn't able to find any... $\endgroup$ Commented Jun 25, 2021 at 8:32

1 Answer 1

6
$\begingroup$

Knowing that $Q_!F : y\mapsto \int^z \hom(Qz,y)\otimes Fz$ (this is often called "pointwise formula" for Kan extensions) it is easy to derive the isomorphism in question: $$ \begin{align} Q_!F\;\otimes_{\cal C_0} G& := \int^y Q_!F(y)\otimes Gy \\ &\cong \int^y \int^z \hom(Qz,y)\otimes Fz \otimes Gy \\ &\cong \int^z Fz\otimes \left( \int^y \hom(Qz,y)\otimes Gy\right)\\ &\cong \int^z Fz\otimes GQ(z) \\ & = F \;\otimes_{\cal C_0} GQ = F \;\otimes_{\cal C_0} Q^*G \end{align} $$

$\endgroup$
6
  • $\begingroup$ Lovely, thanks! So this is really a feature of the Kan extension rather than the adjoint pair? Also, as I understand the copower in a symmetric monoidal category is given by the sym.mon.product? $\endgroup$ Commented Jun 25, 2021 at 9:40
  • 2
    $\begingroup$ As far as I know, this is a feature of the particular weight that expresses Kan extensions as weighted co/limits; in a dual fashion one proves $\{Q^*A,B\}\cong \{A, \text{Ran}_QB\}$ if $\text{Ran}_QB : x\mapsto \int_y [\hom(x, Qy), By]$ and $\{E,F\}=\int_x [Ex, Fx]$. $\endgroup$
    – fosco
    Commented Jun 25, 2021 at 9:54
  • $\begingroup$ Ok, I actually need a hand in understanding one more passage here. I.e., where the isomorphism from the third to fourth line comes from? In other words, why can I describe $GQ(z)$ as a coend? $\endgroup$ Commented Jun 29, 2021 at 11:56
  • 1
    $\begingroup$ it's called "ninja yoneda lemma", or just "yoneda lemma"! $\endgroup$
    – fosco
    Commented Jun 29, 2021 at 17:12
  • $\begingroup$ Yup, now I see - I needed to wrap my head around it. Thank you, sifu! $\endgroup$ Commented Jun 29, 2021 at 17:14

You must log in to answer this question.

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