2
$\begingroup$

Let $P\to M$ be a $G$-principal bundle, $G$ a topological group, $\omega$ the connection and $V$ a vector space. We define $d_\omega: \Omega^k_G(P, V)\to\Omega^{k+1}_G(P, V)$ the exterior covariant derivative such that for a vector-valued differential form $\alpha\in\Omega^k_G(P, V)$, we get $d_\omega\alpha=d\alpha+\omega\wedge_\rho\alpha$, where $\rho:\mathfrak g \to \operatorname{End}(V)$ is a representation of the algebra of $G$.

Question: defining the curvature as $\Omega=d\omega+\frac{1}{2}[\omega,\omega]$, is there any representation in which $\Omega=d_\omega\omega$?

My guess: no.

The point is that I have found several forums (as well as in Wikipedia's entry) in which the curvature was expressed as the covariant derivative of the connection, but in the adjoint representation we do not get the $\frac{1}{2}$ in front. Moreover if we notice, we have for matrix groups, $\frac{1}{2}[\omega,\omega]=\omega\wedge\omega$ and thus one might argue that we could get the curvature by using the fundamental representation.

What do you think?

$\endgroup$

2 Answers 2

4
$\begingroup$

The definition of the exterior covariant derivative isn't $d_\omega\alpha=d\alpha+\omega\wedge_\rho\alpha$.

It is $$ d_\omega\alpha(X...)=d\alpha(\mathrm hX...), $$ where $\mathrm h$ is the horizontal projection.

We then state several things:

  • A $k$-form with values in $V$ $\alpha\in\Omega^k(P,V)$ is a pseudotensorial $k$-form of type $\rho$, if it is right-equivariant in the sense that $r_g^\ast\alpha=\rho(g^{-1})\alpha$.
  • A $k$-form $\alpha$ with values in $V$ is a tensorial $k$-form of type $\rho$, if it is a pseudotensorial $k$-form of type $\rho$ and is horizontal in the sense that $\alpha(X_1,...,X_k)=0$ whenever all elements are vertical.
  • The covariant exterior derivative of a pseudotensorial form is a tensorial form of one higher degree
  • If $\omega$ is the connection form, then it is a pseudotensorial 1-form of type $\text{Ad}$, and we have $d_\omega\omega=\Omega$, so the curvature is the covariant exterior derivative.
  • If $\alpha$ is a tensorial $k$-form of type $\rho$, then we have the "generalized Cartan structure equation" $$ d_\omega\alpha=d\alpha+\omega\wedge_\rho\alpha, $$ however this is valid only if $\alpha$ is a tensorial form, not when it is a pseudotensorial one.
  • In particular, if $\alpha\in\Omega^k(P,\mathfrak g)$ is a tensorial $k$-form of type $\text{Ad}$, then we have $$ d_\omega\alpha=d\alpha+\omega\wedge_{\text{Ad}}\alpha\equiv d\alpha+[\omega\wedge\alpha], $$ where $[\cdot\wedge\cdot]$ is the Lie-exterior product. One may expand (without using the generalized structure equation) that for the curvature we have $$ \Omega=d\omega+\frac{1}{2}[\omega\wedge\omega], $$ where the anomalous factor of 1/2 comes from the fact that $\omega$ is a pseudotensorial form and thus the generalized structure equation is not valid.
$\endgroup$
6
  • $\begingroup$ How do you get the last formula? (When you wrote "one may expand...") $\endgroup$
    – Bellem
    Commented Apr 24, 2019 at 12:21
  • 1
    $\begingroup$ Btw, I guess that for being horizontal/tensorial you only need to get zero whenever at least one vector is vertical, not all vectors... $\endgroup$
    – Bellem
    Commented Apr 24, 2019 at 12:30
  • 2
    $\begingroup$ @Wideep Yes it seems you are correct about horizontality. The formula is derived in Kobayashi/Nomizu and "Gauge Theory and Variational Principles" by Bleecker. And in most books that treat connections on principal bundles. $\endgroup$ Commented Apr 24, 2019 at 18:04
  • $\begingroup$ Thanks, are you sure about KN? Because I could not find the derivation there... $\endgroup$
    – Bellem
    Commented Apr 24, 2019 at 20:12
  • $\begingroup$ @Wideep It's page 77, theorem 5.2 in vol 1. $\endgroup$ Commented Apr 24, 2019 at 21:54
0
$\begingroup$

Yes, this is the case in synthetic differential geometry if we view the connection 1-form $\omega$ on a principal $G$-bundle as a $G$-valued simplicial 1-form. Then the exterior (Darboux) derivative of $\omega$ is indeed the curvature. Theorem 6.7.1 in Anders Kock's Synthetic Geometry of Manifolds explains how this relates to the usual curvature formula for classical differential forms.

$\endgroup$

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.