4
$\begingroup$

Let $E$ be a real analytic vector bundle on an analytic manifold $M$. Assume that $E$, as a smooth vector bundle, is a trivial bundle.

Is $E$ a trivial analytic vector bundle?

I need to the answer to this question for the following question:

Analytic version of the Cartan lemma

$\endgroup$
1
  • 2
    $\begingroup$ Yes, this is a special case of the work of Grauert-Morrey theorem on real-analytic vector bundles on real-analytic manifolds (showing that the theory is much closer to the $C^{\infty}$ case than to the complex-analytic case). Roughly speaking, real-analytic geometry turns out to be closely related to the theory of Stein spaces. $\endgroup$
    – user76758
    Mar 31, 2014 at 2:37

2 Answers 2

2
$\begingroup$

Here's just a sketch, maybe someone else can fill in the details. Most of the steps will require the use of the nontrivial fact that $C^\omega(M)$ is dense in $C^\infty(M)$. Let's assume $M$ is compact.

Step 1: Construct a real-analytic embedding of vector bundles $E\to M\times\mathbb R^N$ for some finite $N$. This gives a real-analytic classifying map $M\to\operatorname{Gr}_n(\mathbb R^N)$.

Step 2: Increase $N$ to $N'$ so that the composition $\gamma:M\to\operatorname{Gr}_n(\mathbb R^N)\to\operatorname{Gr}_n(\mathbb R^{N'})$ is smoothly null-homotopic.

Step 3: Approximate this smooth null-homotopy of $\gamma$ with a real-analytic homotopy from $\gamma$ to a map $\gamma':M\to\operatorname{Gr}_n(\mathbb R^{N'})$ which is contained in an $\epsilon$-ball over which the universal bundle is real-analytically trivial.

I suspect this argument also works for $M$ non-compact (taking $N=\infty$) though I haven't checked carefully.

$\endgroup$
4
$\begingroup$

Here's an easier strategy:

Consider the real-analytic vector bundle $\operatorname{Hom}(\mathbb R^n,E)$ where $\mathbb R^n$ denotes the trivial bundle of rank $n=\operatorname{rk}E$. By assumption, it has a smooth section which is a fiberwise isomorphism. Approximate this smooth section by a real-analytic section (this is where we use a hard theorem to tell us such an approximation exists). Now the property of being a fiberwise isomorphism is an open condition, so the real-analytic section is also a fiberwise isomorphism, and thus gives the desired real-analytic trivialization.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.