This is a rather nice question I got from this user via private communication.

Let $\mathcal{C} = Top$ the category of topological spaces. Let $\mathcal{C}^\prime$ be the category $Funct(\mathcal{C}^{op}, Sets)$. Say $F \in \mathcal{C}^\prime$ is a sheaf, with the usual patch up condition. Does this imply that F is representable?

An elementary way of putting this without using the category theory terminology: Say a set $X$ is endowed with the following data:

For every topological space $Y$, a subset $M(Y,X) \subset Maps(Y,X)$, such that,

  1. whenever $f: Y \rightarrow X$ in $M(Y,X)$, and $g : Z \rightarrow Y$ continuous, $f \circ g : Z \rightarrow X$ belongs to $M(Z,X)$.

  2. If ${U_i}$ is an open cover of $Y$, and $f: Y \rightarrow X$ is a set map, then $f \in M(Y,X)$ if and only if $f|_{U_i} \in M (U_i, X)$ for all $i$.

Now, endow $X$ with the finest topology such that for all $f: Y \rightarrow X \in M (Y,X)$ [for all topological space $Y$], $f$ becomes continuous. In other words, declare $U \subset X$ to be open if for all $f \in M(Y,X)$ for all $Y \in ob (Top)$, $f^{-1} (U)$ is open. [it is clear that that is a topology].

It is clear that under this topology, $f \in M(Y,X)$ is continuous. Does the converse hold? ie. if $f: Y \rightarrow X$ is continuous, then is $f \in M(Y,X)$?

  • $\begingroup$ In the elementary version of your question you assume that $F$ is a subfunctor of a representable functor $Hom(-,X)$. I don't think that this comes for free. $\endgroup$ – Martin Brandenburg Jan 30 '10 at 22:11
  • $\begingroup$ The second version can be explained even to undergraduates. Also even that one is intractable for me. $\endgroup$ – Anweshi Jan 30 '10 at 23:16
  • $\begingroup$ ok put $X=F(pt)$, then there is a canonical natural map $F(Y) \to Maps(Y,X)$, but I don't see why it is injective. $\endgroup$ – Martin Brandenburg Jan 30 '10 at 23:40

There are lots of counterexamples. Take a property of functions defined between topological spaces, such as bounded, and define the sheaf of functions which have this property locally. This works in most cases, I think.

Define $F(Y) := \{f : Y \to \mathbb{R} : f \text{ is locally bounded}\}$. Clearly this is a subsheaf of $Maps(-,\mathbb{R})$. Assume $F$ is representable, then $X:=F(pt)=\mathbb{R}$ is the representing object and carries the finest topology, making all the $f \in F(Y)$ continuous. Assume $U$ is a nonempty open subset of $X$, and take $Y$ to be an arbitrary bounded interval around which intersects $U$, endowed with the indiscrete topology, and $f : Y \to X$ the inclusion. Then $f \in F(Y)$, thus $f$ is continuous, meaning $Y \subseteq U$. As we can vary $Y$, we get $U = X$. Thus $X$ carries the indiscrete topology. But then $X$ represents the functor $Maps(Y,\mathbb{R})$, which is larger than $F$.

There is another reason that $F$ is not representable, namely $F$ does not preserves colimits. A function $f : Y/\sim \to \mathbb{R}$, whose composition with $Y \to Y/\sim$ is locally bounded, does not have to be locally bounded. Take $Y = \mathbb{R}$ and collapse $\mathbb{Z}$ to one point, and let the supremum of $\mathbb{R} \to \mathbb{R}$ around $z \in \mathbb{Z}$ be $|z|$, but taking the same values at $z$.

A reasonable question is now: Let $F : Top^{op} \to Sets$ a continuous sheaf. Is then $F$ representable? If the canonical map $F \to Maps(-,F(pt))$ is injective, the solution set condition is fulfilled and we may apply Freyds Representability Theorem. EDIT: Ok it follows from SAFT.


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.