Let $\pi_P:P\rightarrow M$ a principal $G$ (right action) bundle. Let $F$ be a manifold with a left action of $G$. Then we have the notion of associated fibre bundle over $M$ whose fibre is $F$. I do not know if there is any connection in the other way which is an inverse for the above construction.

Suppose we start with a $Gl(n,\mathbb{R})$ bundle. We have associated vector bundle whose fibre is $\mathbb{R}^n$ which is a vector bundle.

Suppose we start with a vector bundle of rank $n$, we have the notion of what is called frame bundle which is a principal $Gl(n,\mathbb{R})$ bundle.

This says that there is a correspondence between vector bundle of rank $n$ and principal $Gl(n,\mathbb{R})$ bundle.

Do we have such correspondence in case of fibre bundles? Given a fibre bundle, $\pi:E\rightarrow M$, with fibre $F$, can I produce a principal $G$ bundle with an action of $G$ on $F$ such that associated fibre bundle is precisely the one I started with?

  • $\begingroup$ You could look at Steenrod, The Topology of Fiber Bundles, for the theory on topological spaces. $\endgroup$
    – Ben McKay
    Mar 13, 2019 at 16:39
  • $\begingroup$ @BenMcKay ok ok. I will see :) $\endgroup$ Mar 14, 2019 at 2:58

1 Answer 1


Yes. There is a bundle $\mathrm{Fr}(E)\to M$ whose fibre at $m\in M$ is the space $\mathrm{Iso}(F,E_m)$, where $E_m$ is the fibre of $E$ over $m$. Then $G=\mathrm{Aut}(F)$ acts on the right of $\mathrm{Fr}(E)$ by composition. That this is a locally trivial bundle follows from the fact $E$ is locally trivial. Moreover this action makes $\mathrm{Fr}(E)$ a principal $G$-bundle. The canonical action of $G$ on $F$ means we can form the associated bundle $\mathrm{Fr}(E)\times_G F$, and this is (IIRC) isomorphic to $E$.

I didn't specify what kind of isomorphisms should be used. In the case that $E$ is a vector bundle, we take linear isomorphisms, and this is just the usual frame bundle, with $\mathrm{Aut}(\mathbb{R}^n) = GL(n,\mathbb{R})$. For an arbitrary continuous fibre bundle, we take homeomorphisms, and then we have $G=\mathrm{Homeo}(F)$. For a smooth fibre bundle we take diffeomorphisms (but then note that one lands in infinite-dimensional manifold territory, and one has to be a little bit more careful if the fibre $F$ is a non-compact manifold).

  • $\begingroup$ i might be misunderstanding something,,, isn’t it the case that $F$ is same as $E_m$? We are taking automorphisms group of fibre.. that is a Lie group.. corresponding notion is what you are calling as $Fr(E)\rightarrow M$.. is it the case?? Is there any specific reason to write $E_m$ and $F$ both and not just say $\text{Aut}(E_m,E_m)$?? In case of vector bundle, where $E_m$ is an $n$ dimensional real vector space, we have $\text{Aut}(E_m,E_m)=Gl(n,\mathbb{R})$... $\endgroup$ Mar 13, 2019 at 11:10
  • $\begingroup$ They are isomorphic, but not canonically. This isomorphism arises from the local trivialisations, which can be nontrivial. If $F=E_m$ for all $m$, then $E_m = E_{m'}$ for arbitrary pairs. The whole point of connections and parallel transport is that such canonical identifications do not exist. Automorphisms of the fibre are not a Lie group when you are in the topological category, which I left as a possible case covered by this construction. $\endgroup$ Mar 13, 2019 at 11:34
  • $\begingroup$ That is true, it is not canonical isomorphism... :) :) i understand why you write $\text{ISO}(F,E_m)$.... $\endgroup$ Mar 13, 2019 at 11:37
  • $\begingroup$ @David Roberts, in the topological context, I think that this works only if $G = Homeo(F)$ is a topological group, i.e. if $F$ is locally compact and locally connexe for example. $\endgroup$
    – ychemama
    Mar 14, 2019 at 9:53
  • $\begingroup$ @ychemama good point. I guess I was just allowing for the case when the fibre is a non-compact manifold, in the topological or the smooth category. $\endgroup$ Mar 14, 2019 at 10:51

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.