Everything is smooth.

Consider the ring $\Omega^{p}(M)$ of sections of the exterior powers of the cotangent bundle $\Lambda^{p}(T^{*}M)$ over a manifold $M$.

Purely in terms of the smooth structure, we define the exterior derivative as the map $\operatorname{d}:\Omega^{p}(M)\to\Omega^{p+1}(M)$ whose action on $\phi\in\Omega^{p}(M)$ is defined by $$\operatorname{d}\!\phi\bigg(\bigotimes_{k=0}^{p}X_{k}\bigg) = \sum_{i=0}^{p}(-1)^{i}X_{i}\left(\phi\Big(\bigotimes_{k\neq i}X_{k}\Big)\right) + \sum_{i<j}(-1)^{i+j}\phi\Big(\left[X_{i},X_{j}\right],\bigotimes_{k\neq i,j}X_{k}\Big)$$ where the $X_{k}$ are $p+1$ sections of the tangent bundle $TM\to M$, and the tensor products are taken in order.

On the other hand, if we have a connection on the frame bundle we may inherit a covariant derivative $\nabla$ in $TM$ and then in every tensor bundle $T^{n}_{m}M$, including the subbundles $\Lambda^{p}(T^{*}M)$. If the connection is torsion free, then we may express the exterior derivative in terms of the covariant derivative by (using abstract index notation)

$$(\operatorname{d}\phi)_{a_0\dots a_p} = (p+1)\nabla_{[a_0}\phi_{a_1\dots a_p]} \in\Omega^{p+1}(M)$$

for every $\phi\in\Omega^{p}(M)$. This construction is well defined for vector valued forms (elements of $\Omega^{p}(M,E)$), however at the moment I'm just interested in the following...


Is there a coordinate-free proof (say, using abstract index notation) to see that the second definition is equivalent to the first, when $\phi\in\Omega^{p}(M)$?

I've just finished a proof using Penrose graphical notation, however it is rather lengthy. Maybe I was not ingenuous enough. Does anybody here know of a good one?


Althought there are better ways to do this, I wanted to close the question posting my own answer. This is a refined version of the proof I was refering to when I posted the question.

We start with $$(\operatorname{d}\phi)\left(\bigotimes_{k=0}^{p}X_{k}\right) = \sum_{i=0}^{p}(-1)^{i}X_{i}\left(\phi\left(\bigotimes_{k\neq i}X_{k}\right)\right) + \sum_{i<j}(-1)^{i+j}\phi\left(\left[X_{i},X_{j}\right],\bigotimes_{k\neq i,j}X_{k}\right)$$

Let's denote the two terms in the RHS by $(A)$ and $(B)$.

Now we expand $(A)$ as follows

First part Second part So at the end we recognize that

$$(A) = (p+1)\nabla_{[a_0}\phi_{a_1\dots a_p]}\left(\bigotimes_{k=0}^{p}X_{k}\right)^{a_0 \dots a_p} - (B)$$


$$(\operatorname{d}\phi)\left(\bigotimes_{k=0}^{p}X_{k}\right) = (A) + (B) = (p+1)\nabla_{[a_0}\phi_{a_1\dots a_p]}\left(\bigotimes_{k=0}^{p}X_{k}\right)^{a_0 \dots a_p}$$

Since the $X_i$ are arbitrary, we finally have

$$(\operatorname{d}\phi)_{a_0 \dots a_p} = (p+1)\nabla_{[a_0}\phi_{a_1\dots a_p]} \hspace{2cm} \square$$


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.