Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Consider a principal bundle $P\rightarrow M$ and the associated vector bundle $P\times_{\rho}V$ over $M$ such that $(p,v)=(pg^{-1},\rho(g)v)$. The connection on the principal bundle $A$ defines a covariant derivative on $P\times_{\rho}V$ as follows.

View a section $s$ on $P\times_{\rho}V$ as a $G$-equivariant map $s^{P}:P\rightarrow V$. This means $S^{P}(pg^{-1})=\rho(g)s^{P}(p)$. This defines a $G$-equivariant homomorphism $s^{P}_{*}$ from $TP$ to $V$.

Let $x\in M$ and $v\in TM_{x}$, $H_{A}$ be the horizontal vector bundle in $TP$ defined by $A$. Then for any $p$ be the inverse image of $x$, there is a unique horizontal vector $v_{A}\in H_{A}|_{p}$ being the horizontal lift of $v$ such that $\pi_{*}v_{A}=v$. The covariant derivative $\nabla_{A}$ sends section $v$ to the equivalence of the pair $(p,s^{P}_{*}v_{A})$. A routine exercise showed this is not dependent on the choice of $p$.

Taubes claim in his book Differential Geometry that locally we can write the covariderive $\nabla_{A}$ on $P\times_{\rho}V$ as $$x\rightarrow (x,ds_{U}+\rho_{*}(a_{U})s_{U})$$where $\rho_{*}:\mathcal{G}\rightarrow End(V)$ is the differential of $\rho$ at the identity. I do not understand how he derived this identity given the above definition of $\nabla_{A}$. Even if we use the canonical identification $$\phi_{U}^{*}H_{A}=(x,-g^{-1}a_{U}(v)g)\in TU\oplus \mathcal{G}$$ where $\phi_{U}^{*}$ is the local trivialization. I still do not know how to derive the desired identity. Note in particular the connection 1-form on $P$ is given by $$g^{-1}dg+g^{-1}a_{U}g$$ where $a_{U}:U\rightarrow \mathcal{G}\otimes T^{*}M$. However I do not know how to put all these formulas together.

share|cite|improve this question

Actually more is true: globally we can write $\nabla s = ds^P + \rho_*(a) s^P \in \Omega^1(P)\otimes \mathfrak g$ where $a$ is the (global) connection 1-form on $P$. Notice that if $v_A$ is horizontal then by definition $a(v_A) = 0$ so that $\nabla s (v_A) = ds^P(v_A) = s^P_* v_A$, agreeing with what you wrote.

Now you may ask then why we didn't just define $\nabla s$ as $d s^P$. The point is that we want $\nabla s$ to Lie in $\Omega^1(M) \otimes \mathfrak g$, where we view $\Omega^1(M) \subset \Omega^1(P)$ by the pullback of the projection. This identifies $\Omega^1(M)$ with those forms in $\Omega^1(P)$ which are right invariant and horizontal (i.e. they annihilate $\ker \pi_*$). The problem is that $ds^P$ is not horizontal: let $X^*$ be the vector field on $P$ given by $p \mapsto p\exp(tX)$ for $X \in \mathfrak g$. Then $$ ds^P(X^*_p) = \frac{d}{dt}\vert_{t=0} s(p\exp(tX)) = \frac{d}{dt}\vert_{t=0} \rho(\exp(-tX)) s(p) = -(\rho_* X) s(p). $$ However, this is cancelled by the $\rho_* a$ since $a(X^*) = X$ by definition of a connection form.

To answer the question in the title, the curvature 2-form is just $\rho_* \Omega$ where $\Omega \in \Omega^2(M)\otimes \mathfrak g$ is the curvature form of the connection on $P$.

share|cite|improve this answer
well done. One question remains - you explained the motivation, not the derivation of the formula. Can you tell me how to derive it? – Bombyx mori Oct 16 '12 at 13:14

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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