Doubt on how to prove proposition about bundles

67 Views Asked by At

I've started studying bundles and fiber bundles and to get some practice I've tried to prove the following proposition: "Every vector bundle $(E,B,\pi,F,G)$ is associated to a given principal bundle $(E',B',\pi',G',G')$."

I've been told that this proposition is true, but as a first time trying to deal with this kind of thing I'm a little confused yet. Can someone give me just a little hint on how to start this proof? I don't want a full answer, because I want to do it by myself, I just want a hint on how to start.

Thanks very much in advance!

1

There are 1 best solutions below

1
On BEST ANSWER

Suppose you have a trivialization of $\pi: E \to B$ (with fiber $F$), so that you have a covering $U_\alpha$ of $B$ and a bundle isomorphism $\phi_\alpha$ between $\pi^{-1}(U_\alpha) \to U_\alpha$ and $p_\alpha =\text{proj}: U_\alpha \times F \to U_\alpha$; the isomorphism should respect of course the $G$-action on fibers. On intersections $U_\alpha \cap U_\beta$ you can then put these together to construct a bundle isomorphism from $(U_\alpha \cap U_\beta) \times F$ to itself, lying over the identity on $U_\alpha \cap U_\beta$. This in turn is given by a map $U_\alpha \cap U_\beta \to \text{Aut}(F)$ that factors through the action $G \to \text{Aut}(F)$. (Cf. Wikipedia on local trivialization; normally we require the action on $F$ to be faithful.)

So the structure of the bundle can be described in terms of a family of maps $\phi_{\alpha \beta}: U_\alpha \cap U_\beta \to G$, called Čech cocycle data; i.e., the original bundle is retrieved by taking all the trivial bundles $p_\alpha: U_\alpha \times F \to U$, and gluing them together by identifying the bundle restrictions of $p_\alpha, p_\beta$ over $U_\alpha \cap U_\beta$ along the isomorphism induced by $\phi_{\alpha \beta}$.

The point is that you can use the same cocycle data to create a principal bundle, replacing the action $G \to \text{Aut}(F)$ by the canonical action $G \to \text{Aut}(G)$ (say as left $G$-spaces). There is a nice description of the original bundle in terms of this principal bundle $E' \to B$, which can roughly be expressed as $E \cong E' \otimes_G F$, where the tensor product is meant to indicate identifications $(e g, f) \sim (e, g f)$ (NB: there is a left $G$-bundle map $E' \times G \to E'$, parallel to the fact that the right $G$-action of $G$ on itself is a left $G$-space map by virtue of the associative law).

More details can be found in Wikipedia (there is an article "associated bundle"), but this should be a start. Or did you want something more?