Doubt showing commutativity of Pushforward and Lie Bracket

differential-geometrymanifoldspullbackpushforwardVector Fields

In this journey of understanding some of differential geometry I started to do some proof of results. And now I on doubt about this one which is the commutativity between Pushforward and the Lie bracket without draw upon local coordinates:

Let $M$ and $N$ two differential manifolds, and
$$ \varphi: M \longrightarrow N $$
A diffeomorphism between the two.

So if I have a map $g: N \rightarrow \mathbb{R} $ then I can define a function $\varphi^*g: M \rightarrow \mathbb{R} $ the Pullback of $g$.

And with the same philosophy, given a vector field on $M$, $X: C^\infty(M) \rightarrow C^\infty(M)$ the Pushforward vector field on $N$ is defined as: $\varphi_*X(g) = X(\varphi^*g), \ g \in C^\infty(N)$

(I made all this intro because I'm not even 100% sure about if I got this concepts right, so any comment is welcome).

So my idea of proof is first start with the pushformward of the Lie bracket, with $f \in C^\infty(N)$:
$$ \varphi_*[X,Y](f) = [X,Y](\varphi^*f) = X(Y(\varphi^*f)) – Y(X(\varphi^*f))$$
Since the two terms are the same under a swaping of $X$ and $Y$ I will continue with the first one.

So, by the definition of pushforward $Y(\varphi^*f) = \varphi_*Y(f)$, but $\varphi_*Y(f) \in C^\infty(N)$ so it cannot be an argument for $X(\cdot{})$ therefore the expression
$$ X(Y(\varphi^*f)) = X(\varphi_*Y(f))$$
Cannot the be true.

My intuition tells me that the solution of this is that, since $\varphi_*Y(f)$ is another map on $N$ I can first "pull-it back" to $M$ before feed it into $X$. Having:
$$X(Y(\varphi^*f)) = X(\varphi^*(\varphi_*Y(f))) = \varphi_*X(\varphi_*Y(f))$$
Letting me completing my "proof" after a few straightforward steps.

Is this correct? Is so, what is the justification for incorporate that last pullback? How can I know that the pullback of $\varphi_*Y(f)$ if the map on $M$ that I'm looking for?

Sorry for the bad English and thanks in advance!

Best Answer

(I made all this intro because I'm not even 100% sure about if I got this concepts right, so any comment is welcome).

So far, the definitions and setup seem fine.

Is this correct? Is so, what is the justification for incorporate that last pullback?

Yes, it is correct that $Y(\varphi^*f) = \varphi^*(\varphi_*Y(f))$.

Here is a rough proof.

I'm mostly copying this proof from Proposition 8.16 of John Lee's book "Smooth Manifolds"

To prove it, take a point $p \in M$. Then $Y(\varphi^*f)$ is a function on $M$, which at the point $p$ is: $$ Y(\varphi^*f)(p) = (\varphi_*Y)_{\varphi(p)}(f) $$ Here, $(\varphi_*Y)_{\varphi(p)}$ is the pushforward vector field evaluated at $\varphi(p)$. This last expression is equal to $\varphi_*Y(f)(\varphi(p))$. That is, $\varphi_*Y(f)$ is a function on $N$, and evaluating it at $\varphi(p)$ is equal to $(\varphi_*Y)_p(f)$. Now, this last expression, $\varphi_*Y(f)(\varphi(p))$, is the composition $\Big( \varphi_*Y(f) \Big) \circ \varphi$, evaluated at $p$. By definition of pullback of functions, this is equal to $\varphi^*(\varphi_*Y(f))$.

Related Question