3
$\begingroup$

Let $P$ and $Q$ be categories, and suppose $P$ has a terminal object $\ast_p$. A parametric right adjoint, or pra, $F$ from $P$ to $Q$ is a functor $F:P\to Q$ such that the functor $$F\,/\,{\ast_p}: P\,/\,{\ast_p}\to Q\,/\,F(\ast_p)$$ induced by slicing over $*_p$ is a right adjoint. In particular, since right adjoints preserve terminal objects, note that every right adjoint $P\to Q$ is a pra.

Suppose that $C$ and $D$ are categories, and let $P:=C{-}\mathsf{Set}$ and $Q:=D{-}\mathsf{Set}$ be the associated copresheaf categories. A functor $f\colon C\to D$ induces a pullback functor $$\Delta_f\colon D{-}\mathsf{Set}\to C{-}\mathsf{Set},$$ which has both a left adjoint $\Sigma_f$ and right adjoint

$$ \Pi_f\colon C{-}\mathsf{Set}\longrightarrow D{-}\mathsf{Set}. $$

In particular, each of $\Delta_f:Q\to P$ and $\Pi_f:P\to Q$ is a parametric right adjoint.

Remark: A functor $C{-}\mathsf{Set}\longrightarrow D{-}\mathsf{Set}$ is a pra iff it is isomorphic to one of the form $\Delta_e\,\overset{\circ}{,}\,\Pi_f\,\overset{\circ}{,}\,\Sigma_g$, where $$ C\xleftarrow{e}\bullet\xrightarrow{f}\bullet\xrightarrow{g}D $$ are categories and functors, where $(e,f)$ forms a two-sided discrete fibration, and where $g$ is a discrete opfibration. This fact is due to Mark Weber.

Definition: Let $\mathsf{pra}$ denote the category whose objects are categories $\text{Ob}(\mathsf{pra})=\text{Ob}(\mathsf{Cat})$ and for which a morphism $C\to D$ is a pra between their copresheaf categories $$ \mathsf{pra}(C,D):=\{F\colon C{-}\mathsf{Set}\longrightarrow D{-}\mathsf{Set}\;\mid\;F\text{ is a pra}\}. $$ We refer to $\mathsf{pra}$ as the category of categories and copresheaf-pras. It is in fact a bicategory whose 2-morphisms are natural transformations.

Question 1: Inside the bicategory $\mathsf{pra}$ of categories and copresheaf-pras, can one characterize those morphisms $F$ that are of the form $F=\Delta_f$ for some functor $f\colon D\to C$?

As pointed out by Simon Henry in an answer below, it is equivalent to ask:

Question 2: Inside the bicategory $\mathsf{pra}$ of categories and copresheaf-pras, can one characterize those morphisms $F$ that are of the form $F=\Pi_f$ for some functor $f\colon C\to D$?

$\endgroup$
2
  • $\begingroup$ I can't find in Mark Weber's paper the precise statement of this results you mentioned in remarks, but don't you need to assume that the categories under consideration are Cauchy complete for this to hold ? I also feel like Cauchy completeness will probably be helpful to give nice purely categorical characterization of these functors (as any category is isomorphic to its Cauchy completion in the category pra). $\endgroup$ Commented Jul 10, 2020 at 13:16
  • $\begingroup$ Thanks Simon. I was misremembering the result on page 674. I've edited the text. $\endgroup$ Commented Jul 10, 2020 at 15:22

1 Answer 1

2
$\begingroup$

A first remark is that question 1 and 2 are equivalent as in the category pra you have $\Delta_f \dashv \Pi_f$. So if you have a characterization of one class you characterize the other as their left/right adjoint functors

Now, the copresheaf functor from pra to Cat can be recovered as the global section functor:

Indeed, the terminal object of pra is the small category $\emptyset$, indeed $\emptyset$-Set is the terminal category and the unique functor $C$-Set $\to 1$ is always a (parametric) right adjoint functor.

Moreover, any functor $\emptyset$-Set $ =1\to C$-Set is parametric right adjoint, so you get that pra$(\emptyset$-Set,$C$-Set$) \simeq C$-Set.

So, at least assuming Cauchy completeness you can characterize the $\Delta_f$ and $\Pi_f$ as the adjunction $h \dashv g$ such that the action of $h$ on global section has a further left adjoint.

I don't think you can give a purely categorical characterization of these functor without assuming Cauchy-completness as every small category is isomorphic to its Cauchy-completion in pra: a purely categorical construction cannot distinguish between a $\Delta_f$ functor and something that is a $\Delta_f$ for an $f$ between the Cauchy completion.

$\endgroup$

You must log in to answer this question.

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