Let $G$ be a connected Lie group and $\Sigma$ a closed oriented surface. We know that principal $G$-bundles $P$ can be topologically classified by a characteristic class $c(P)\in H^2(\Sigma,\pi_1G)\cong\pi_1G$.

The following is my question:

Let $G$ be a semisimple connected (or even compact) Lie group, $\beta\colon\pi_1(\Sigma)\to G$ a group homomorphism, and $\Sigma$ a closed oriented surface. Consider the universal cover $\tilde{\Sigma}\to\Sigma$ which is a principal $\pi_1(\Sigma)$-bundle. Form the associated bundle $\tilde{\Sigma}\times_\beta G$ which is necessarily a principal $G$-bundle over $\Sigma$. Hence, it should have a characteristic class $c(\tilde{\Sigma}\times_\beta G)\in H^2(\Sigma,\pi_1G)\cong\pi_1G$. Is there a way to compute this characteristic class in terms of $\beta$?

The difficulty here is that $\pi_1(\Sigma)$ is not connected, and hence I cannot use the functoriality of the characteristic class. Another one is that I am not sure how to compute the induced map $B\beta\colon B\pi_1(\Sigma)\to BG$ concretely.

I've asked this question on MSE and haven't got any answers yet.


1 Answer 1


Note sure if you can call it "calculating" but there is a rather straightforward construction. Let us fix some generators $A_i, B_i$ of $\pi_1(\Sigma)$ (i.e. a canonical system of curves such that slicing $\Sigma$ along them gives the fundamental polygon). Consider the central extension of $\pi_1$ $$ \mathbb{Z} \to \Gamma \to \pi_1(\Sigma),$$ where $\Gamma$ is generated by $A_i, B_i$ and the central element $J = \Pi_i [A_i, B_i]$ (with the commutator being $[A, B] = ABA^{-1}B^{-1}$). Since this extension is universal, every group homomorphism $\beta: \pi_1(\Sigma) \to G$ lifts to a group homomorphism $\hat{\beta}: \Gamma \to \tilde{G}$ where $\tilde{G}$ is the universal covering group. Then $\hat{\beta}(J) \in \pi_1(G)$ is the characteristic class of the bundle coming from $\beta$.

I think more details about this construction can be found in Atiyah and Bott's classical work on the Yang-Mills equation over surfaces.

  • $\begingroup$ Actually, I asked this question exactly because I was reading Atiyah and Bott's paper. Therefore, could you explain why $\hat{\beta}(J)$ is the characteristic class of the bundle? They didn't offer many details actually. $\endgroup$
    – YYF
    Nov 6, 2017 at 15:05
  • $\begingroup$ Well, it is kind of like in Chern-Weil theory. There you integrate the curvature over the 2-dimensional fundamental polygon. However, now that everything is discrete, this integral has to be replaced with the holonomy of the boundary curve. Now the "holonomy" is $\hat{\beta}$ and the boundary curve $J$. Admittedly, this reasoning is a bit heuristic but should give you a general idea (sorry, I don't have the time right now for a more detailed answer). $\endgroup$ Nov 7, 2017 at 12:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.