Let $C$ be a small category, and $D$ any category, and $D^C$ the functor category of functors from $C$ to $D$.

The diagonal functor $\Delta\colon D\to D^C$ takes an object $a\in D$ to the constant functor at that object: $\Delta(a)(x) = a$ for all $x\in C.$

The limit functor $\lim\colon D^C \to D$ takes a functor $F\colon C\to D$ to its limit, $\lim F$, the initial cone over $F$.

There is an adjunction $\operatorname{Hom}_{D^C}(\Delta(a),F) = \operatorname{Hom}_{D}(a,\lim F).$ This is more or less just a restatement of the universal property of the limit.

Now if we set $D=\text{Sets},$ then $D^C=\text{Sets}^C=\hat{C}$ is the category of presheaves on $C$. $\Gamma(F)=\operatorname{Hom}_{\hat{C}}(1,F)$ is the global sections functor. In this setting we again have an adjunction with the diagonal functor: $\operatorname{Hom}_{\hat{C}}(\Delta(a),F) = \operatorname{Hom}_{\text{Sets}}(a,\Gamma(F)).$

By the uniqueness of adjoint functors, we can conclude that $\Gamma\cong\lim$, yes? Assuming that's true, that seems rather odd, can I have some context to that isomorphism to make it seem more natural or less surprising? How does it look for enriched presheaves, when $D$ is not $\text{Sets}$?

Is there some significance to the fact that the limit of a functor is just the set of natural transformations from the constant functor at the terminal object?


Your situation is actually a special case of a general fact :

Proposition : If $G:\mathcal{X}\to \mathbf{Sets}$ is a functor that has a left adjoint $F$, then $G$ is represented by $F(1)$ (where $1$ is just a one-element set).

The proof is easy : just notice that $\operatorname{Hom}_\mathbf{Sets}(1,\_)$ is naturally isomorphic to the identity functor on $\mathbf{Sets}$, and then the adjunction gives you directly an isomorphism $$G(x)\cong \operatorname{Hom}_\mathbf{Sets}(1,G(x))\cong \operatorname{Hom}_{\mathcal{X}}(F(1),x)$$ natural in $x$.

In your case $G=\lim:\mathcal{X}\to \mathbf{Sets}$ is right adjoint to $\Delta$, thus it must be represented by $\Delta(1)$, i.e. is must be isomorphic to $\operatorname{Hom}_{[C,\mathbf{Sets}]}(\Delta(1),\_)$, which is how you've defined the global sections functor.

  • $\begingroup$ I guess it was the choice to elide the presence of a diagonal functor in the sheaf notation. All we're really saying is, via the first adjunction, that $\operatorname{Hom}(\Delta 1,F)=\operatorname{Hom}(1,\lim F)$, and then for any set $S$, $\operatorname{Hom}(1,S)\cong S.$ This is just an obvious fact about sets that they are determined by their elements, which we might call the axiom of extensionality or the fact that Set is a cocomplete well-pointed category. I guess my question turned out kind of dumb. $\endgroup$ – ziggurism Dec 4 '19 at 16:45

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.