I am currently a physicist studying differential geometry I am trying to proof the expression below.

Given that for a map $\phi$ : $M$ $\to$ $M$ the pull-back $\phi$*$\omega$ $\in$ $T^\ast_p M$ of a 1-form $\omega $ $ \in$ $T^\ast_p M$ is defined by :

($\phi$*$\omega$)$(v)$ = $\omega$($\phi_{*}v$) where $v$ $\in$ $T_{p}M$.

How would we proof this in a coordinate basis $dx^{\mu}_{p}$, $\phi^{*}\omega$ has components:

$(\phi^{*}\omega)_{\nu} = \frac{\partial x^{'\mu}}{\partial x^{v}}\omega_{\mu}$

where $\mathbf{\omega} = \omega_{\mu}dx^{\mu}_{\phi(p)}$ and $x^{'\mu} = x^{\mu} \bullet \phi $.

EDIT and also prove that if $\phi$ is a diffeomorphism, then the push-forward is $\phi$*$\omega$ $\in$ $T^{\ast}_{\phi(p)} M$ of a 1-form $\omega$ $\in$ $T^{\ast}_{p} M$ is defined by:

$(\phi_{*}\omega)(v) = \omega(\phi^{*}v)$ for any $v \in T^{\ast}_{\phi(p)} M$. Prove that in the coordinate basis $dx^{\mu}_{\phi(p)}, \phi_{*}\omega$ has components :

$(\phi_{*}\omega)_{\nu} = \frac{\partial x^{\mu}}{\partial x^{'v}}\omega_{\mu}$.

To clarify things please find the extract of the notes I am reading: extract



We want to show that $$\phi^* \omega = \frac{\partial (x')^\mu}{\partial x^\nu} \omega_\mu dx^\nu$$ where I use the usual summation convention and have written $\omega = \omega_\mu dx^\mu$ in local coordinates.

Since the pull-back is linear, it will be enough to check that $$(\phi^* dx^\mu)_\nu = \frac{\partial (x')^\mu}{\partial x^\nu}$$ where my subscript refers to the coordinate representation with respect to coordinates $x^\nu$.

Now $$(\phi^* dx^\mu)_\nu = (\phi^* dx^\mu)\bigg(\frac{\partial}{\partial x^\nu} \bigg) = dx^\mu \bigg( \phi_* \frac{\partial}{\partial x^\nu} \bigg) = \bigg( \phi_* \frac{\partial}{\partial x^\nu} \bigg)_\mu.$$ Finally, we have $$\bigg( \phi_* \frac{\partial}{\partial x^\nu} \bigg)_\mu = \bigg( \phi_* \frac{\partial}{\partial x^\nu}\bigg)(x^\mu) = \frac{\partial}{\partial x^\nu}(x^\mu \circ \phi) = \frac{\partial (x')^\mu}{\partial x^\nu}$$ which gives the desired result.

  • $\begingroup$ Thanks Rhys. I have edited the question we also have to prove the pushforward of the 1-form $\omega$ and we are explicitly told the proof should show subscripts p and $\phi_{p} $ on all basis vectors $\frac{\partial}{\partial x^{\mu}}$ and 1-forms $dx^{\mu}$. Is an analogous proof to what you done for the second part still applies ? And is it OK to ignore the indices p and $\phi(p)$? Thanks $\endgroup$ – kevint Feb 16 at 21:51
  • $\begingroup$ @kevint If I'm honest, the notation in your edit is a bit confusing to me, I think there are some typos. However, it seems like your definition of the push-forward will just be the pull-back by $\phi^{-1}$ and then this will follow by the first result by replacing $\phi$ with $\phi^{-1}$. $\endgroup$ – Rhys Steele Feb 16 at 21:59
  • 1
    $\begingroup$ @kevint By the way, in future you shouldn't really edit questions to ask further things after you receive an answer. Instead just ask a new question! Since you're a new user, I don't mind answering here this time. $\endgroup$ – Rhys Steele Feb 16 at 22:00
  • $\begingroup$ Understood. I have included an extract of the notes where this exercises are asked to be proved at the bottom of the page. $\endgroup$ – kevint Feb 16 at 22:12

I find this all easier to understand if I write $\phi$ as a map between two different manifolds (where the two might coincidentally be the same manifold).

Also, it all comes from abstract linear algebra: If $A_*: V \rightarrow W$ is a linear map (i.e., a pushforward), then there is a natural dual map $A^*: W^* \rightarrow V^*$ (i.e., a pullback), such that for any $v \in V$ and $\ell \in W^*$, $$ (A^*\ell)(v) = \ell(A_*v). $$ If you choose bases for $V$ and $W$ and use the dual bases for $V^*$ and $W^*$, then you can write $A_*$ and $A^*$ as matrices.

Now you can apply this to the differential of a map $\phi: M \rightarrow N$, which is a linear map $\phi_*: T_pM \rightarrow T_{\phi(p)}N$, which is analogous to the map $A_*$ above and defined as follows: Given $v \in T_pM$, there exists a curve $c: (-\delta,\delta) \rightarrow M$ such that $c(0) = p$ and $c'(0) = v$. You can compose $\phi$ with $c$ to get a curve in $N$ and define $$ \phi_*v = \left.\frac{d}{dt}\right|_{t=0}\phi(c(t)) \in T_{\phi(p)}N. $$ Since $\phi_*: T_pM \rightarrow T_{\phi(p)}N$ is a linear map like $A_*$ above, there is a dual map $\phi^*: T_{\phi(p)}^*N \rightarrow T_p^*M$.

When you push forward a vector field $v$, you're just applying the linear map $\phi_*$ to $v(p)$ for each $p \in M$. Notice that if $\phi$ is either not injective or not surjective, $\phi_*v$ is not a vector field on $N$. Similarly, the pullback of a differential form $\omega$ on $N$ is simply applying $\phi^*$ to $\omega(\phi(p))$ for each $p$. Notice that, contrast to the pushforward, the pullback of a smooth differential form on $N$ is a smooth differential form on $M$.

Since everything above was defined without using local coordinates, we now know they don't depend on any choice of coordinates.

If you now choose local coordinates on $M$ near $p \in M$ and on $N$ near $\phi(p)$, then you get a basis of $T_pM$ by holding all but one coordinate on $M$ fixed and differentiating the curve with respect to the remaining coordinate. You can do the same using the coordinates on $N$. You can now write $\phi_*: T_pM \rightarrow T_{\phi(p)}N$ as a matrix, just as for $A_*$ above. Using the corresponding dual bases, you can write $\phi^*: T^*_{\phi(p)}N \rightarrow T^*_pM$ as matrices, just as described for $A^*$. Now you can check that the matrices for $\phi_*$ and $\phi^*$ are essentially the Jacobian matrix of partial derivatives of $\phi$ written with respect to the local coordinates on $M$ and $N$.

Finally, to minimize confusion, I recommend never talking about the pushforward of a differential form or the pullback of a vector field. If $\phi$ is a diffeomorphism, then there is a pushforward of vector fields on $N$ by $\phi^{-1}_*$ and a pullback of differential forms on $M$ by $\phi^{-1}$. Such precision in language makes it much less likely you'll get confused or make mistakes.


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.