Let $M$ be a manifold and $\pi : E \to M$ a rank $n$ vector bundle on $M$. We can define a connection on $E$ in two ways:

  • We can specify the covariant derivatives $\nabla_X s$ or
  • We can choose a connection form $\Phi \in \Omega^1({\rm Fr} \, E) \otimes \mathfrak{gl}_n$ where ${\rm Fr} \, E$ is the frame bundle.

I have seen both of the following called curvature:

  • $R(X,Y,s) = \nabla_X \nabla_Y s - \nabla_Y \nabla_X s - \nabla_{[X,Y]} s$
  • $ \Omega = d \Phi + \frac{1}{2} [\Phi,\Phi]$

The first is an ${\rm End}(E)$-valued 2-form on $M$ and the second is a $\mathfrak{gl}_n$-valued 2-form on ${\rm Fr} \, E$

Question 1: How are $R$ and $\Omega$ related?

By direct computation, if $ \Omega = 0$, then the horizontal subbundle is integrable, so the connection is flat. A slightly more concrete question than question 1 is

Question 2: Why does $R = 0$ imply that the connection is flat? (By flat, I mean that the horizontal subbundle $ H \subseteq T ({\rm Fr} \, E)$ is integrable)


There are many textbooks on Differential geometry answering your question in detail, for example the Kobayashi-Nomizu book, or in not so much detail, e.g. Roe's "Elliptic operators...".

Concerning question 1: Basically, $\Omega$ and $R$ are the same: $R$ is a 2-form on $M$ with values in the endomorphism bundle of $E,$ but the endomorphism bundle is the bundle associated to adjoint representation. Hence, after pulling back to $Fr E$ you can think of $R$ as a horizontal equivariant 2-form, and this 2-form is exactly $\Omega.$

Question 2: $\Omega$ measures the integrability of the horizontal distribution: if you compute $\Omega(X,Y)$ for horizontal vector fields (which are by definition in the kernel of $\Phi$) all what remains is $$\Omega(X,Y)=-\Phi([X,Y]).$$ If $R$ vanishes, $\Omega$ vanishes, and hence the commutator of horizontal fields is horizontal, again.

| cite | improve this answer | |
  • $\begingroup$ Thanks a lot! It took me a while to digest this, but I understand now. $\endgroup$ – Daniel Barter Oct 18 '15 at 1:36

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.