Let $\pi\colon P\to X$ be some $G$-principal bundle; $\{U_\alpha\}$ a cover of $X$; and $\{s_\alpha\colon U_\alpha\to\pi^{-1}(U_\alpha)\}$ a collection of local sections.

Claim: The pullback $\pi^*P$ of $P$ over itself is trivial, as witnessed by the global section $\pi^* s_\alpha$.

My question: What does the above statement mean (if it makes sense at all)? That is, we know that the local sections differ by some transition $g_{\alpha\beta}\colon U_\alpha\beta\to G$, so how do we obtain a single global section from these local ones on the pullback?


The claim is correct, the local sections are irrelevant.
Proof of claim:
The pulled-back bundle $\pi^*(P) $ consists of those ordered pairs $(p,p')\in P\times P$ which satisfy $\pi(p)=\pi(p')\in P$.
The projection $\Pi:\pi^*(P)\to P $ of that bundle to its base is given by the formula $\Pi(p,p')=p.$
This projection has a canonical global section $\sigma:P\to \pi^*(P)$ given by $\sigma (p)=(p,p)$.
And thus the $G$-bundle $\pi^*(P)$ is trivial, like all principal bundles having a global section.

  • $\begingroup$ Sorry, I didn't make my question very clear: I know that this bundle is trivial, but I want to understand what happens to these local sections $\endgroup$ – Tim May 17 '18 at 18:54
  • $\begingroup$ These local sections $s_\alpha$ give rise to local sections $S_\alpha:\pi^{-1}(U_\alpha)\to \pi^*P: p \mapsto (s_\alpha(\pi p),s_\alpha(\pi p))$ of the pulled-back bundle $\pi^*P$. But these new local sections $S_\alpha$ will not glue unless the old sections $s_\alpha$ already glued, in which case the old $G$-bundle $P$ already was trivial and there was no need for your construction... $\endgroup$ – Georges Elencwajg May 17 '18 at 19:12

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.