5
$\begingroup$

I'm reading Appendix C of "Characteristic Classes" by Milnor & Stasheff and something confuses me. When proving that the Pfaffian of the curvature form (of an oriented $2n$-plane bundle $\xi$ over a smooth manifold $M$, with a metric and a connection compatible with that metric) represents a multiple of the Euler class of $\xi$, the authors say: enter image description here (here $K$ stands for curvature tensor)

The notation $Pf(K(\tilde{\gamma}))$ suggests that $\tilde{\gamma}$ has a connection on it, thus the base space must be a smooth manifold. On the other hand, $\tilde{\gamma}$ is required to be "universal", so it should be the tautological bundle over the oriented Grassmannian $\tilde{G}_{2n}(\mathbb{R}^{\infty})$, but then the base space is not a manifold, seemingly a contradiction. So what should this bundle $\tilde{\gamma}$ be?

$\endgroup$
9
$\begingroup$

In the second line of the quoted passage, it says that $\tilde\gamma$ should be `universal in dimensions $\leq 4n$'. This means that it's universal for bundles over complexes of dimension at most $4n$, and so we can take a finite dimensional oriented Grassmannian $\tilde{G}_{2n}(R^N)$ for sufficiently large $N$. This is a manifold, and the restriction of $\tilde\gamma$ has a connection.

You can read about $n$-universal bundles in Steenrod's book, Fiber Bundles.

$\endgroup$
  • $\begingroup$ Thank you for the reference. Some more questions: How does the structure of $H^*(\tilde{G}_{2n}(\mathbb{R}^N);\mathbb{R})$ guarantees that $a^2=b^2$ implies $a=\pm b$? Steenrod's book only considers finite complexes, so how is $\tilde{\gamma}$ universal for bundles over manifolds of dimension $\leq 4n$?(I guess it's possible to extend bundle maps on finite subcomplexes to a countable complex using a weak topology argument?) $\endgroup$ – AaronS May 10 '18 at 13:45
  • $\begingroup$ A presentation for the cohomology ring by generators (Pontrjagin and Euler classes) is known; if that implication on the cup products is true it ought to follow directly by a bit of algebra. And yes also for your second question. $\endgroup$ – Danny Ruberman May 11 '18 at 18:34

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.