# Fiber product of sheaves

If one has a topological space $X$ and three presheaves resp. sheaves $F$ and $G$ and $H$ of abelian groups on it with morphisms of presheaves resp. sheaves $F\rightarrow H$, $G \rightarrow H$, then I wonder if one can consider a fiber product of $F$ and $G$ over $H$ in the category of presheaves and sheaves $F \times_HG$

Well, one would perhaps define it in the category of presheaves just as

$F\times _HG (U)$:= all (s,t) $\in F(U)\times G(U)$ going to the same element in $H(U)$.

And for sheaves then take the associated sheaf of this. Or perhaps it is already a sheaf if $F,G,H$ are sheaves?

Just give me some comment if this makes sense and if this concept is of relevance anywhere in Algebraic Geometry. I have never seen it indeed except for perhaps, if you want so, in the case of Schemes, where you consider a scheme as Zariski-Hom-sheaf.

-

Let $X$ be a topological space. The category of sheaves of sets on $X$, $\textrm{Sh}(X)$, is an example of a Grothendieck topos and is, in particular, complete and cocomplete. Therefore fibre products, or pullbacks as they are known in general category theory, exist in $\textrm{Sh}(X)$. Of course, one needs to verify that the forgetful functor $\textbf{Ab}(\textrm{Sh}(X)) \to \textrm{Sh}(X)$ preserves pullbacks, but this is straightforward.
Your proposed construction works and in fact does not require sheafification since limits and colimits of presheaves are constructed sectionwise, and the inclusion of categories $\textrm{Sh}(X) \hookrightarrow \textrm{Psh}(X)$ preserves all limits. (This is because it has a left adjoint, namely the associated sheaf functor.)
What I said holds for a general Grothendieck topos – see any reference on general topos theory. The problem with the category of sheaves over a large site is that these are, in general, not toposes. In some cases this is not a problem; for example, if you are only interested in schemes locally of finite type over a base field $k$, then you can look at the Zariski/étale/fpqc/fppf/etc. site on the category of finitely presented $k$-algebras. –  Zhen Lin Dec 19 '11 at 1:19