Let $C$ be a category, $D$ be a Grothendieck topos, and suppose we have a fully faithful, left-exact functor $F:C\rightarrow D$. Let $Lan_{y}F:PShv(C)\rightarrow D$ be the Yoneda extension of $F$. Since both $C,D$ are locally presentable and $Lan_{y}F$ is cocontinuous, $Lan_{y}F$ has a right adjoint, call it $G$. Is there an abstract way of seeing if $G$ is fully faithful?
-
5$\begingroup$ Well, you're making this more complicated than it is, because the right adjoint $D \to Set^{C^{op}}$ just takes an object $d$ to $\hom_D(F-, d): C^{op} \to Set$. This goes back to Kan's 1958 paper where adjoint functors were first introduced. Off-hand I don't have an answer to your question though... or maybe I do, it's that $F$ is dense. See ncatlab.org/nlab/show/dense+functor $\endgroup$– Todd Trimble ♦Dec 26, 2016 at 3:17
1 Answer
Assuming $C$ is a small category, the right adjoint to $Lan_y(F): PShv(C) = Set^{C^{op}} \to D$ is just the functor $D \to Set^{C^{op}}: d \mapsto \hom_D(F-, d)$; this basic "generalized nerve" construction traces back to the original 1958 paper by Kan where adjoint functors were first introduced:
- Daniel Kan, Functors involving c.s.s complexes, Transactions of the American Mathematical Society, Vol. 87, No. 2 (Mar., 1958), pp. 330--346 (jstor).
Meanwhile, the condition that the nerve $N_F$ defined by $N_F(d) = \hom_D(F-, d)$ is fully faithful (as in the case of the classical nerve induced by the inclusion $\Delta \hookrightarrow Cat$) is much considered in the categorical literature: we say that $F$ is dense, since it amounts to every object of $D$ being a colimit of objects coming from $C$ in a canonical way. A pretty good set of references is given in the nLab article on dense functors; some of the older references (such as by Isbell under the name "adequacy") pay particular attention to the case where $F$ itself is an embedding of a full subcategory.