0

The fiber bundle is defined as

A fiber bundle is defined as the tuple $(E, B, \pi)$ where $\pi: E \to B$ is a continuous surjective map from topological space $E$ to topological space $B$. Furthermore, there is a topological space $F$, called the fiber, such that for each $b\in B$ there is an open neighborhood $U\subset B$ with $b\in U$ such that $\pi^{-1}(U)\subset E$ is homeomorphic to $B\times F$ via homeomorphism $\varphi: \pi^{-1}(U)\to B\times F$ and $\text{proj}_1\circ \varphi = \pi$.

$\pi^{-1}(\{b\}) \subset E$ is called the fiber of $b$.

One definition of a principal bundle is

The tuple $(E, B, \pi)$ is a $G$-principal bundle if it is a fiber bundle and if there is a continuous action $\lhd:E\times G \to E$ which preserves the fibers in $E$ (that is $x\in \pi^{-1}(b)$ then $x \lhd g \in \pi^{-1}(b)$ for $g\in G$) and which acts freely and transitively on the fibers in $E$.

But another definition or property I've seen for principal bundles over smooth manifolds is (e.g. in the Schuller lectures)

The tuple $(E, B, \pi)$ is a $G$-principal bundle if there is a smooth free action $\lhd: E\times G \to G$ and $\pi:E\to B$ is isomorphic as a smooth bundle to $\rho: E\to E/G$ where $\rho$ is the normal equivalence relation projection map where $\rho(e)$ for $e\in E$ maps $e$ to $[e]$, the equivalence class of $e$.

Since $\pi:E\to B$ is a smooth bundle then $E$ and $B$ are manifolds, $\pi$ is a smooth map, and $\varphi:\pi^{-1}(U)\to B\times F$ is a diffeomorphism.

Wikipedia also says that similar properties fully characterize the principal bundle.

Here are my questions.

  • Is there any reason the second principal bundle definition can't be relaxed to topological fiber bundles instead of smooth fiber bundles?
  • If the second principal bundle definition can and is weakened to be for topological fiber bundles then are the two definitions equivalent?
  • If the first principal bundle definition is strengthened to apply to smooth manifolds then are the two definitions equivalent?
Jagerber48
  • 1,631
  • What kind of group is $G$? For definition 1 to be correct as it stands, you need to assume something on the order of "$G$ is a compact Lie group and $E$ a manifold," see this answer. For definition 2, you'll also at least need that $G$ is a compact Lie group. If you want $G$ to be something more general (I guess the most general you can go is "arbitrary Lie group"), or $E$ in definition 1, then you'll need to add conditions. – Ben Steffan Apr 01 '24 at 21:25
  • Ah, sorry, the fiber is what needs to be a manifold in definition 1. – Ben Steffan Apr 01 '24 at 21:38
  • My suggestion is to avoid using Schuller's lectures as a source for learning anything. Wikipedia is also suboptimal. There are many textbooks covering this material. – Moishe Kohan Apr 01 '24 at 21:56
  • @BenSteffan I guess for the first definition I would just say $G$ is a topological group so that "continuous" makes sense. For the second definition I would say $G$ is a Lie group so that the group action being smooth makes sense. You're suggesting these definitions aren't restrictive enough? – Jagerber48 Apr 01 '24 at 22:12
  • @MoisheKohan I've found the Schuller lectures to be very good and useful for a lot of topics. However, when it comes to bundles specifically, it seem that he uses some non-standard definitions which have led to a lot of confusion (lots of questions on this site and responses by you saying the lectures are a bad source). Would you mind pointing me towards a few good references where I can learn about Principal bundles? – Jagerber48 Apr 01 '24 at 22:13
  • There generally seems to be a lot of confusion online about the definition of Principal bundle which is very frustrating. I'm not sure if the literature/textbook references are better or not. – Jagerber48 Apr 01 '24 at 22:14
  • Steenrod's book, or Kobayashi-Nomizu . I wrote several answers on MSE dealing with terminological confusion regarding fiber bundles, I will try to locate these. – Moishe Kohan Apr 01 '24 at 22:18
  • @Jagerber48 Yes. If you want to define something that matches the usual definitions of principal $G$-bundles people use and avoids pathologies, then you'll need to add more conditions, in particular something like properness of the action. Your frustration is entirely understandable; there are a sizeable number of competing definitions out there and a not insubstantial amount of them are "wrong" in more or less subtle ways. – Ben Steffan Apr 01 '24 at 22:19
  • https://math.stackexchange.com/questions/399035/equivalence-of-definitions-of-principal-g-bundle/2135581#2135581 – Moishe Kohan Apr 01 '24 at 22:19
  • https://math.stackexchange.com/questions/560371/orbit-space-of-a-free-proper-g-action-principal-bundle/1754706#1754706 – Moishe Kohan Apr 01 '24 at 22:21
  • https://math.stackexchange.com/search?q=user%3A84907+fiber+bundle – Moishe Kohan Apr 01 '24 at 22:23
  • https://math.stackexchange.com/questions/3095741/doubt-in-the-definition-of-principal-bundle/3099251#3099251 – Moishe Kohan Apr 01 '24 at 22:23
  • Thank you very much for hunting down all those answers and references! It seems very complicated :(. I wish there was one place I could look to get a succinct definition of Principal bundle. Instead it feels like I need to learn 4 different definitions and all of the machinery to compare between the different definitions which is a much larger task than just digesting one definition. – Jagerber48 Apr 01 '24 at 22:26
  • @MoisheKohan I've asked a question looking for help on understanding the Kobayashi Nomizu definition of principal bundle here: https://math.stackexchange.com/questions/4891961/understanding-the-definition-of-principal-bundle-kobayashi-and-nomizu it's similar to one of your linked answers, but I think my questions are really just very basic, coming from someone who's still an "immature" reader of topological details. – Jagerber48 Apr 02 '24 at 17:33

0 Answers0