I'd be very happy if the question When do Kan extensions preserve limits/colimits? has been fully answered. But it seems not.

I have a more specific question though. Let $C$ be a site (essentially small to be safe) equipped with some subcanonical Grothendieck topology, $PSh$ be its category of presheaves. Then it's well-known that for any functor $C \to B$ with $B$ cocomplete, the left Kan extension $PSh \to B$ along the Yoneda embedding $C \to PSh$ necessarily preserves colimit. Is it true if I replace $PSh$ by the category of sheaves $Sh$? In other words, does the left Kan extension along $C \xrightarrow{Yoneda} PSh \xrightarrow{sheafification} Sh$ preserve colimit? If in general not, is there any sufficient condition we can impose on the functor $C \to B$?


Usually it does not. We can take $C \to B$ to be the universal example of a functor to a cocomplete category, namely the Yoneda embedding $C \to PSh$. Then the left Kan extension of $y : C \to PSh$ along $j : C \to PSh \to Sh$ is the inclusion $Sh \to PSh$, and this usually does not preserve colimits. The calculation of this left Kan extension can be verified using the coend formula: $$ (Lan_j y)(X) = \int^{c : C} Sh(jc, X) \otimes yc = \int^{c : C} X(c) \otimes yc = X $$ where the second equality used the fact that the topology is subcanonical.

  • $\begingroup$ Thanks for the counterexample! Is there any sufficient condition on the functor $C \to B$ that can guarantee the left Kan extension is colimit preserving? $\endgroup$ – gregodom Jun 10 '20 at 14:08
  • $\begingroup$ E.g. if $F:C \to B$ is a subfunctor of a functor $F':C \to B$ whose left Kan extension is colimit preserving, is the left kan extension of $F$ colimit preserving? $\endgroup$ – gregodom Jun 10 '20 at 14:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.