For me, a local system on a topological space $X$ is a locally constant sheaf on $X$, i.e., a sheaf of sets $\mathscr{F}$ for which there is an open cover $X=\bigcup_i U_i$ so that $\mathscr{F}_i\cong\underline{\mathscr{F}(U_i)}_{U_i}$, the constant sheaf on $U_i$ associated to the set $\mathscr{F}(U_i)$.

The local systems I'm interested in arise as follows (and perhaps they all arise this way, maybe under some additional connectedness assumptions on $X$). Let $G$ be a group acting (on the left, let's say) freely and properly discontinuously on $X$ (I'm not sure if the definition of the latter term is universally agreed upon, but the condition I want is that each $x\in X$ admits an open $x\in U\subseteq X$ such that $gU\cap U=\emptyset$ for $g\neq 1$) and $V$ an $R$-module (for some commutative ring $R$) with an $R$-linear action of $G$. Then the natural projection $q:G\setminus(X\times V)\rightarrow G\setminus X$ (here I regard $V$ as having the discrete topology and let $G$ act on $X\times V$ diagonally) is then a covering space (because of the assumption on the action of $G$ on $X$), and all the fibers are in bijection with $V$. So the sheaf of continuous sections $\mathscr{F}_V$ on $G\setminus X$ is going to be locally on $X$ isomorphic to the constant sheaf $\underline{V}_X$. Now, my understanding is that something like this is called a local system of $R$-modules, and my question is:

What is the $R$-module structure on the set of sections $\mathscr{F}_V(U)$ for $U\subseteq X$ open?

The only thing that I can think of (and I guess I'm motivated by the vector space structure on the sections of a vector bundle) is that for each $x\in X$, the restriction of $X\times V\rightarrow G\setminus(X\times V)$ to $\{x\}\times V$ is a bijection onto the fiber over $Gx$ of $q$. So one can make the fiber into an $R$-module by declaring that this bijection should be an isomorphism of $R$-modules. Then the $R$-module structure on $\mathscr{F}_V(U)$ would be defined pointwise (like for vector bundles), i.e., if $s,t\in\mathscr{F}_V(U)$, so that $s(u),t(u)\in q^{-1}(u)$, then $(s+t)(u)=s(u)+t(u)$, the addition taking place in the fiber $q^{-1}(u)$, and likewise, for $r\in R$, $(rs)(u)=rs(u)$, the scalar multiplication taking place in $q^{-1}(u)$.

I think that with this definition, if $U$ is an open set of $X$ over which the covering $G\setminus(X\times V)\rightarrow G\setminus X$ can be trivialized, then the induced isomorphism $\mathscr{F}_V\vert_U\cong\underline{V}_U$ will be one of sheaves of $R$-modules, which suggests I have the right definition...but I haven't checked the details. Do I have the right definition of the $R$-module structure?


$X \times V$ is a constant local system of $R$-modules, just using the $R$-module structure on $V$. (A section on $U$ is just a locally constant $V$-valued function on $U$, with the $R$-module structure defined pointwise.)

The action of $G$ on $V$ is $R$-linear, and so $G\backslash (X \times V)$ is a (typically no longer constant) local system of $R$-modules too. A section of this local system over an open set $U$ of $G\backslash X$ is a $G$-equivariant locally constant $V$-valued function on the preimage of $U$ in $X$, again with the $R$-module structure defined pointwise.

In short, you have the correct $R$-module structure.

  • $\begingroup$ Ah, I see. I like this description better than mine. Less fiber-by-fiber and sort of more global. Thank you. $\endgroup$ – Keenan Kidwell Aug 18 '13 at 2:07

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.