Let $M$ be an oriented $m$-dimensional manifold. Suppose the support of $\omega$ is in an open subset $U$ of $M$, and $\phi \colon U \to R^m$, $\psi \colon U \to R^m$ are two different charts on $M$ defining the same orientation.

In coordinates {$r_1,r_2,...r_m$} on U, we can see $\omega$ as $f dr_1\wedge dr_2\wedge \cdots \wedge dr_m$. Then the pull back $({\phi}^{-1})^{*}\omega = (f{\phi}^{-1}) dr_1{\phi}^{-1}\wedge \cdots \wedge dr_m{\phi}^{-1}$, say $(f{\phi}^{-1})dx_1\wedge dx_2\wedge \cdots \wedge dx_m,$ where define $dr_i{\phi}^{-1}$ to be $dx_i$

Similarly, $({\psi}^{-1})^{*}\omega = (f{\psi}^{-1}) dr_1{\psi}^{-1}\wedge \cdots \wedge dr_m{\psi}^{-1}$, say $(f{\psi}^{-1})dy_1\wedge dy_2\wedge \cdots \wedge dy_m$.

As we know $$\int_{\phi(U)}({\phi}^{-1})^{*}\omega = \int_{\psi(U)}({\psi}^{-1})^{*}\omega.$$ Note $$\int_{\phi(U)}({\phi}^{-1})^{*}\omega = \int_{\phi(U)}f({\phi}^{-1})(x)dx_1\wedge dx_2\wedge \cdots \wedge dx_m := \int_{\phi(U)}(f{\phi}^{-1})(x)dx,$$ and similarly $$\int_{\psi(U)}({\psi}^{-1})^{*}\omega = \int_{\psi(U)}(f{\psi}^{-1})(y)dy.$$ And $\int_{\psi(U)}(f{\psi}^{-1})(y)dy$ may not be equal to $\int_{\phi(U)}(f{\phi}^{-1})(x)dx$ with the reason in the picture below:

enter image description here

So it seems that it leads to a contraction, that is $\int_{\phi(U)}({\phi}^{-1})^{*}\omega = \int_{\psi(U)}({\psi}^{-1})^{*}\omega$ first and then they are not equal. Can you tell me where I went wrong? Thanks.


I think I see the issue.

Let $\omega$ be an $m$-form on $M$, supported in $U \subset M$. Let $\tau \colon U \to \mathbb{R}^m$ be an oriented chart, with coordinates $(r_1, \ldots, r_m)$. On $U \subset M$, we can write $\omega = f\,dr_1 \wedge \cdots \wedge dr_m$.

Let $\phi, \psi \colon U \to \mathbb{R}^m$ be oriented charts, also on $U \subset M$. It is true that $$\int_{\phi(U)} (\phi^{-1})^*\omega = \int_U \omega = \int_{\psi(U)} (\psi^{-1})^*\omega.$$ Let $(z_1, \ldots, z_m)$ be the standard coordinates on $\mathbb{R}^m$. It is also true that \begin{align*} \int_{\phi(U)} (\phi^{-1})^*\omega & = \int_{\phi(U)} (\phi^{-1})^*(f dr_1 \wedge \cdots \wedge dr_m) \\ & = \int_{\phi(U)} (f \circ \phi^{-1}) \cdot (\phi^{-1})^*(dr_1 \wedge \cdots \wedge dr_m) \\ & = \int_{\phi(U)} (f \circ \phi^{-1}) \cdot (d(r_1 \circ \phi^{-1}) \wedge \cdots \wedge d(r_m \circ \phi^{-1})) \\ & = \int_{\phi(U)} (f \circ \phi^{-1}) \cdot (d(z_1 \circ \tau \circ \phi^{-1}) \wedge \cdots \wedge d(z_m \circ \tau \circ \phi^{-1})) \end{align*} and similarly, \begin{align*} \int_{\psi(U)} (\psi^{-1})^*\omega & = \int_{\psi(U)} (\psi^{-1})^*(f dr_1 \wedge \cdots \wedge dr_m) \\ & = \int_{\psi(U)} (f \circ \psi^{-1}) \cdot (\psi^{-1})^*(dr_1 \wedge \cdots \wedge dr_m) \\ & = \int_{\psi(U)} (f \circ \psi^{-1}) \cdot (d(r_1 \circ \psi^{-1}) \wedge \cdots \wedge d(r_m \circ \psi^{-1})) \\ & = \int_{\psi(U)} (f \circ \psi^{-1}) \cdot (d(z_1 \circ \tau \circ \psi^{-1}) \wedge \cdots \wedge d(z_m \circ \tau \circ \psi^{-1})) \\ \end{align*}

The point is now that $d(z_1 \circ \tau \circ \phi^{-1}) \wedge \cdots \wedge d(z_m \circ \tau \circ \phi^{-1})$ is not necessarily equal to the euclidean volume form $dz_1 \wedge \cdots \wedge dz_m$. Similarly, $d(z_1 \circ \tau \circ \psi^{-1}) \wedge \cdots \wedge d(z_m \circ \tau \circ \psi^{-1})$ is not necessarily equal to $dz_1 \wedge \cdots \wedge dz_m$, either. That's the issue.

Addendum: Now, I have a feeling that this might leave you unsatisfied. That is, you might say something like: Well, can't I just define $x_i = r_i \circ \phi^{-1} = z_i \circ \tau \circ \phi^{-1}$ and $y_i = r_i \circ \psi^{-1} = z_i \circ \tau \circ \psi^{-1}$? Because if I do that, then I will get $$\int_{\phi(U)} (\phi^{-1})^*\omega = \int_{\phi(U)} (f \circ \phi^{-1})\,dx_1 \wedge \cdots \wedge dx_m$$ and $$\int_{\psi(U)} (\psi^{-1})^*\omega = \int_{\psi(U)} (f \circ \psi^{-1})\,dy_1 \wedge \cdots \wedge dy_m,$$ and indeed you do. However, that's not what your book means by by $dx$ and $dy$. What your book means is that $dx$ and $dy$ are both the euclidean volume form $dz_1 \wedge \cdots \wedge dz_m$.

The euclidean volume form is defined to be the $m$-form $dz_1 \wedge \cdots \wedge dz_m$ on $\mathbb{R}^m$, where $(z_1, \ldots, z_m)$ are the standard Cartesian coordinates. So, for example, in $\mathbb{R}^2$, the euclidean volume form is $$dz_1 \wedge dz_2 = r\,dr \wedge d\theta.$$ In $\mathbb{R}^3$, it is $$dz_1 \wedge dz_2 \wedge dz_3 = r\,dr \wedge d\theta \wedge dz_3 = \rho^2\sin\phi \,d\rho \wedge d\theta \wedge d\phi.$$

  • $\begingroup$ Thank you, that's a very good answer! $\endgroup$ – 6666 Jun 16 '16 at 21:54

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.