Let $G$ be a Lie group, $M$ be a smooth manifold and $\pi:P\rightarrow M$ be a principal $G$-bundle. Given a connection $1$-form $A:P\rightarrow \Lambda^1_{\mathfrak{g}}T^*P$ on the principal bundle, fixing a point $x\in M$ we have the notion of holonomy map $\Omega(M,x)\rightarrow \text{Aut}(\text{fiber of P at x})$

Recall : Given a loop (path) $\gamma$ in $M$, based at $x$, and a point $u\in \pi^{-1}(x)$, the connection gives a path $\gamma^*_u$ in $P$ starting at $u$.Varying $u$ over $\pi^{-1}(x)$, we get paths $\{\gamma^*_u:[0,1]\rightarrow P|u\in \pi^{-1}(x)\}$. This gives a map $\Phi_x:\pi^{-1}(x)\rightarrow \pi^{-1}(x)$ defined as $u\mapsto \gamma^*_u(1)$. I was wondering if $\Phi_x$ is a smooth map (recall that $\pi^{-1}(x)$ is an embedded submanifold of $P$).

Is the parallel transport/displacement map $\Phi_x:\pi^{-1}(x)\rightarrow \pi^{-1}(x)$ a smooth map for each $x\in M$? I am sure this is true but could not prove it now. Any help is appreciated.

  • $\begingroup$ Of course, you do not even need to assume that the bundle is principal. This should be in volume 1 of Kobayashi-Nomizu. The proof boils down to applying a theorem from ODEs on smoothness of the solution of the Cauchy initial value problem for 1st order ODEs. $\endgroup$ – Moishe Kohan Sep 23 '19 at 16:13
  • $\begingroup$ @MoisheKohan I have seen in Kobayashi and Nomizu, I do not see mention of smoothness of the map $\tau:\pi^{-1}(x)\rightarrow \pi^{-1}(y)$ anywhere,, $\endgroup$ – Praphulla Koushik Sep 23 '19 at 16:42
  • $\begingroup$ See Appendix I (the main theorem in the appendix). You apply this theorem locally, to charts covering the given curve in $M$. $\endgroup$ – Moishe Kohan Sep 23 '19 at 17:18
  • $\begingroup$ I think in the principal bundle case, it is even easier as it is enough to note that $\pi^{-1}(x)$ is a smooth $G$-torsor and the map $\Phi_x$ is $G$-equivariant. $\endgroup$ – Tobias Diez Sep 24 '19 at 8:27
  • 1
    $\begingroup$ Fix a point $x \in X$, then $G \ni g \mapsto x \cdot g \in X$ is a diffeomorphism. This allows you to write your map as a map $G \to G$, which by $G$-equivariance is smooth. In your case, this map is $g \mapsto a g$ where $a \in G$ is fixed. $\endgroup$ – Tobias Diez Sep 24 '19 at 9:47

As I said in a comment, this holds even for general smooth bundles (which need not be not $G$-bundles). You can find a proof for instance in Theorem 9.8, page 80, in:

I.Kollar, P.Michor, J.Slovak, "Natural operators in differential geometry", Springer-Verlag, 1993.

But the key is a local theorem from ODEs (proof of which you can find in any graduate-level textbook on ODEs):

Assuming that $\Phi$ is a vector field on an open subset $U$ of $R^n$, the Cauchy problem of the form $$ x'(t)= \Phi(x), x(t_0)=v\in R^n $$ has unique smooth short-term solutions on relatively compact open subsets in $U$. See for instance Appendix I of Kobayashi-Nomizu "Foundations of Differential Geometry".

  • $\begingroup$ Thank you. I checked Natural operators in differential geometry.. did not understand completely, I might need day or two but I am sure I will get it :) $\endgroup$ – Praphulla Koushik Sep 24 '19 at 2:26

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.