Definition: Suppose $M\subseteq \mathbb{R}^n$ is a compact smooth k-manifold and $f:M\to \mathbb{R}$ is a continuous function. Since $S=\text{support}(f)$ is compact suppose $\phi:U\to V$ is a chart such that $S\subseteq V$ and $U$ is bounded. We define the integral of $f$ over $M$ to be $$\int_M fdV=\int_{U^{int}} (f\circ \phi)\text{Vol}(D\phi)$$

Suppose $x_1,...x_k\in\mathbb{R}^n$ form the vertices of a k-parallelepiped $P$, and $X$ be the matrix with columns $x_i$ then the volume of $P$ is $\sqrt{\det(X^TX)}$

Prove this definition is independent of coordinate chart $\phi$

Suppose $\psi:\widetilde U\to V$ is coordinate chart, then $\phi^{-1}\circ\psi:\widetilde U\to U$ is a change of variables.

Applying change of variables to the integral $$\int_{U^{int}} (f\circ \phi) \text{Vol}(D\phi)$$

$$\int_{\widetilde U^{\text{int}}} (f\circ\phi\circ(\phi^{-1}\circ\psi))\vert \det D(\phi^{-1}\circ\psi)\vert \text{Vol}(D\phi)$$ $$=\int_{\widetilde U^{\text{int}}} (f\circ\psi)\vert \det D\phi^{-1} D\psi\vert\text{Vol}(D\phi)$$

I'm not sure how to deal with the jacobians here. Since I don't really understand how you can take the determinant of $D\phi$ when $\phi$ is a map from $\mathbb{R}^k$ to $\mathbb{R}^n$ so it isn't a square matrix.

  • $\begingroup$ Please define $\text{Vol}(D\phi)$. $\endgroup$ – Ted Shifrin Mar 27 at 1:06
  • $\begingroup$ @Ted Shifrin Added definition. $\endgroup$ – AColoredReptile Mar 27 at 1:13
  • $\begingroup$ So now you have some linear algebra to do with what you have written. $\endgroup$ – Ted Shifrin Mar 27 at 1:26
  • $\begingroup$ @Ted Shifrin So I need to show that $\text{Vol}(D\psi)=\vert \det (D\phi^{-1})(D\psi)\vert\text{Vol}(D\phi)$? I guess my confusion is that $D\phi$ is an $n\times k$ matrix, is it not? Otherwise if it was the $k\times k$ independent rows of that matrix then I would have $\text{Vol}(D\phi)=\vert\det (D\phi)\vert$ and $\det (D\phi^{-1}) \det(D\phi)=1$ so $\vert \det (D\phi^{-1})(D\psi)\vert\text{Vol}(D\phi)=\vert\det (D\psi)\vert=\text{Vol}(D\psi)$ $\endgroup$ – AColoredReptile Mar 27 at 1:37

Applying the change of variables we get $$\int_{U^{int}} (f\circ \phi) \text{Vol}(D\phi)=\int_{\widetilde U^{\text{int}}} (f\circ\psi)\cdot\text{Vol}(D\phi)\circ\phi^{-1}\circ\psi\cdot\vert \det D(\phi^{-1}\circ\psi)\vert$$

So we should show

$$\text{Vol}(D\psi)=\text{Vol}(D\phi)\circ\phi^{-1}\circ\psi\cdot\vert \det D(\phi^{-1}\circ\psi)\vert$$

Writing this out, taking squares and using $|\det A|^2=\det (A\cdot A^T)$ and the multiplicativity of the determinant wee see that this is equivalent to

$$\det (D\psi^T\cdot D\psi) =\det\left(((D\phi)\circ\phi^{-1}\circ\psi)^T\cdot ((D\phi)\circ\phi^{-1}\circ\psi)\cdot D(\phi^{-1}\circ\psi)\cdot D(\phi^{-1}\circ\psi)^T\right).$$

Since $\psi=\phi\circ(\phi^{-1}\circ\psi)$ by applying the chain rule the right hand side becomes

$$\det\left(((D\phi)\circ\phi^{-1}\circ\psi)^T\cdot D\psi\cdot D(\phi^{-1}\circ\psi)^T\right).$$

Using the commutativity of the determinantant this is equal to $$ \det\left((D(\phi^{-1}\circ\psi)^T\cdot ((D\phi)\circ\phi^{-1}\circ\psi))^T\cdot D\psi\right)$$

which due to $(A\cdot B)^T=B^T\cdot A^T$ is

$$\det\left((((D\phi)\circ\phi^{-1}\circ\psi)\cdot D(\phi^{-1}\circ\psi))^T\cdot D\psi\right)$$

and so again by using the chain rule this is just $$ \det\left(D\psi^T\cdot D\psi\right)$$

which is what we wanted to show.

  • $\begingroup$ What happened to $D(\phi^{-1}\circ\psi)$ in your 2nd step? $\endgroup$ – AColoredReptile Mar 27 at 17:15
  • $\begingroup$ Sorry. You said $\psi=\phi\circ(\phi^{−1}\circ \psi)$ and applied the chain rule. And I assume simplified $(D\phi)\circ\phi^{-1}\circ\psi)\cdot D(\phi^{-1}\circ\psi)\to D\psi$. But I dont see how.$D(\phi^{-1}\circ\psi)=D(\phi^{-1})\circ\psi\cdot D\psi$ right? $\endgroup$ – AColoredReptile Mar 27 at 21:25
  • $\begingroup$ Maybe you get confused by the notation. $ ((D\phi)\circ\phi^{-1}\circ\psi)\cdot D(\phi^{-1}\circ\psi)=D\psi$ is an equation of matrix valued functions on $\tilde U$. $D(\phi^{-1}\circ\psi)$ is the function which maps an $x$ to $(D(\phi^{-1}\circ\psi))(x)$ which is the Jacobian of $\phi^{-1}\circ\psi$ at the point $x$ whereas $(D\phi)\circ\phi^{-1}\circ\psi$ maps an $x$ to $(D\phi)(\phi^{-1}(\psi(x))$ which is the Jacobian of $\phi$ at the point $\phi^{-1}(\psi(x))$. $\endgroup$ – Chiara Mar 27 at 23:33
  • $\begingroup$ The chain rule states that for all $x\in \tilde U$ $(D\phi)(\phi^{-1}(\psi(x))\cdot (D(\phi^{-1}\circ\psi))(x)=(D\psi)(x)$ which just means $((D\phi)\circ\phi^{-1}\circ\psi)\cdot D(\phi^{-1}\circ\psi)=D\psi$. I hope this helps! $\endgroup$ – Chiara Mar 27 at 23:42
  • $\begingroup$ Yes I see what you meant now. $\endgroup$ – AColoredReptile Mar 27 at 23:47

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.