I am reading this paper and at the top of page 5 the author makes reference to categories consisting of flat complex vector bundles over $X$ where $X$ is an arbitrary topological space. However, the only notion of flat vector bundle which I have seen uses a connection on the vector bundle, which in turn requires that $X$ have a cotangent bundle, hence $X$ must be a smooth manifold.

My questions are therefore as follows:

  1. What is the definition of a flat vector bundle over a topological space?
  2. What is the action of $\pi_1(X)$ on such a vector bundle? (I need to understand $H^\bullet(X;\text{Hom}_X(V,W))$ for flat vector bundles $V,W$ over $X$.)
  • 3
    $\begingroup$ For sensible spaces, vector bundles with discrete structure group are a good candidate $\endgroup$ Sep 28, 2015 at 2:36

3 Answers 3


A flat vector bundle over a topological space is a bundle whose transition functions can be taken to be locally constant; equivalently, over a path-connected space, it's the same data as a principal $G$-bundle ($G = GL_n(\mathbb{R})$ or $GL_n(\mathbb{C})$ as appropriate) where $G$ is given the discrete topology. Over a reasonable space $X$ this is the same thing as a functor from the fundamental groupoid $\Pi_1(X)$ to vector spaces.

  • 7
    $\begingroup$ I disagree with Qiaochu: it's not "a bundle whose transition functions can be taken to be locally constant". That would be a bundle with a property, whereas flat vector bundles are bundles equipped with extra structure. One way to describe that structure is that there's a certain distinguished class of local sections which are singled out, which one calls the "flat sections". The flat sections should form a locally constant sheaf, and the stalks of that sheaf should be the fibers of the vector bundle. $\endgroup$ Oct 1, 2015 at 20:26
  • $\begingroup$ @Andre: my apologies for being imprecise. Of course the issue is that I haven't said what a morphism of flat vector bundles is, and of course the correct notion is not the same as a morphism of underlying vector bundles; I think you also need a local constancy condition here and then things are fine? The discrete structure group definition is maybe clearer (in particular, it's also clear in this formulation that flatness is a structure rather than a property) but I wanted to avoid a formulation which doesn't allow vector bundles to jump in rank across different components. $\endgroup$ Oct 2, 2015 at 2:05

According to Kamber-Tondeur (1967), a principal $G$-bundle over a space $X$

is flat if it is induced from the universal covering bundle of $X$ by a homomorphism $\pi_1X\to G$. In the differentiable case this is equivalent to the existence of a connection with curvature zero [15, Lemma 1].


A vector bundle is called flat, if its associated principal bundle is flat.


A way to rephrase QY's answer is to say that a vector bundle E is flat if there is a local system L such that $L \otimes_{\mathbb{C}} C_X \simeq E$ where $C_X$ is the trivial vector bundle (or, said differently, $C_X$ is the bundle whose sections on an open U are given by $C_X(U)$ - the continuous functions $f\colon U \to \mathbb C$).

  • 2
    $\begingroup$ I don't understand this argument. Doesn't this just amount to the (true) remark that "Another name for a flat vector bundle is a local system"? $\endgroup$
    – Tom Church
    Sep 29, 2015 at 0:08
  • 1
    $\begingroup$ @Tom: I think pro wants to identify a local system with its sheaf of flat sections but wants to identify a flat vector bundle with its sheaf of continuous sections. $\endgroup$ Sep 29, 2015 at 3:35
  • 1
    $\begingroup$ @TomChurch I am not saying anything deep. I like sheaves: while the local system L is a Q-sheaf of finite rank, the trivial vector bundle C_X is infinite dimensional. When you don't have Kahler differentials this is the only way I know to say that a vector bundle is flat (this and by using representations of $\pi_1$ as QY said). $\endgroup$
    – pro
    Sep 29, 2015 at 4:50

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.