In what sense is a constant sheaf of abelian groups with stalks G isomorphic to G?

327 Views Asked by At

In the English translation of Serre's FAC (link to the PDF can be found in this mathoverflow discussion in the top answer) Serre gives his first example of a sheaf of abelian groups, the constant sheaf, on page 8 in section 1: Operations on Sheaves. The following is verbatim from the translation, however do note that Serre defines a sheaf from the étalé space point of view, the definition is also given on page 8 or can be found online.

Example of a sheaf. For $G$ an abelian group, set $\mathscr{F}_x = G$ for all $x \in X$; the set $\mathscr{F}$ can be identified with the product $X \times G$ and, if it is equipped with the product topology of the topology of $X$ by the discrete topology on G, one obtains a sheaf, called the constant sheaf isomorphic with $G$, and often identified with $G$.

For the most part I understand this example. I am able to verify that the group operation maps are continuous, and the projection map is a local homeomorphism, so $\mathscr{F}$ is indeed a sheaf using the definition that Serre gives. Then for the sake of exploration I studied the sections of this sheaf and found that $\Gamma(U,\mathscr{F})$ is the group of locally constant functions $s\colon U \to \mathscr{F}$. Further, if $U$ is connected they become constant functions and then $\Gamma(U,\mathscr{F}) \cong G$, but if not connected $\Gamma(U,\mathscr{F}) \cong \oplus \, G$ where you get a summand for each connected component of $U$. None of this stuff is too exciting, and can be found on the wikipedia page for the constant sheaf. But what I do not get is

In what sense would this constant sheaf $\mathscr{F}$ be isomorphic to $G$? Why would we want to identify it with $G$? I see no way that the sheaf itself acts like $G$, and really it is more like no matter where you look on this sheaf (stalks, sections) you are either going to find $G$, or find copies of $G$. I checked the french version and it is not a translation error. Can anyone provide some context or explanation for this?

Here is another example of where Serre uses this language in a way that is confusing to me:

Every sheaf of abelian groups can be considered a sheaf of $\mathbb{Z}$ - modules, $\mathbb{Z}$ being the constant sheaf isomorphic to the ring of integers.

Also, here is another (less important) question that occurred to me while typing my original question up that I don't think is quite worthy of its own post (but I'll move it if someone complains):

Does anything weird happen when the subset $U$ has countably many or uncountably many connected components? Usually we would want to switch to a direct product of $G$ but that seems not to be what we would want, since we wouldn't want to require the sections to be nontrivial on only finitely many components?