I read the proof of Stokes theorem for manifolds by do Carmo's book and I'm trying understand an example (the Divergence theorem) given after the proof of Stoke's theorem, but I didn't understand. A preliminary definition of star Hodge operator $*$ given by do Carmo which is used in the example can be found here. The example is this. The map $i$ is the inclusion map $i: \partial M \longrightarrow M$.

I would like to know why $i^* *\omega(e_1,e_2) = \omega (N)$.

I thought that $i^* *\omega(e_1,e_2) = *\omega(di(e_1),di(e_2)) = *\omega(e_1,e_2) = \omega(N)$, but this seems incorrect to me, because we can indentify $\partial M$ as $\mathbb{R}^2$ and $M$ as $\mathbb{R}^3$ and see the inclusion map as

$$i: \mathbb{R}^2 \longrightarrow \mathbb{R}^3$$

$$(x_2, x_3) \mapsto (0,x_2,x_3)$$

due to this proposition of do Carmo's book, then

$di_p = \begin{pmatrix} 0 & 0\\ 1 & 0\\ 0 & 1 \end{pmatrix}$, $di_p (e_1) = (0,1,0) = a_2$ and $di_p (e_2) = (0,0,1) = a_3$, therefore

$$i^* *\omega(e_1,e_2) = *\omega(di(e_1),di(e_2)) = *\omega(a_2,a_3) = \omega(a_1),$$

where $a_1 = (1,0,0)$, but don't know what to do now, because I don't know if $a_1 = N$.

I will appreciate any comment in order to elucidate the reason why $i^* *\omega(e_1,e_2) = \omega (N)$ is valid. Thanks in advance!


A mistake I noticed is that you say you can identify $\partial M$ with $\mathbb{R}^{2}$, when really it should be $T_{p}\partial M$ with $\mathbb{R}^{2}$ for any point $p \in \partial M$ (and similarly for $T_{p}M$ with $\mathbb{R}^{3}$).

So with that in mind, let the inclusion map $i:\partial M \rightarrow M$ be given by $$i(x_{1}, x_{2}) = (x_{1}, x_{2}, 0),$$ then in letting $p = (x_{1}, x_{2})$, in a neighbourhood $U \subset \partial M$ with $p \in U$, let $\{e_{1}, e_{2}\}$ be an orthonormal frame for $T_{p}\partial M$. Then $$d_{p}i:T_{p}\partial M \longrightarrow T_{i(p)}M,\qquad (e_{1}, e_{2}) \longmapsto (e_{1}, e_{2}, 0). $$ The tangent space $T_{i(p)}M$ can be seen to decompose as $$T_{i(p)}M = T_{p}\partial M\ \oplus\ N_{i(p)}M,$$ where $N_{i(p)}M$ is the normal space to $M$ at the point $i(p)$, i.e. the orthogonal complement to $T_{p}\partial M$, which is one-dimensional and spanned by $N$, say. Then $e_{1}, e_{2}, N$ forms an orthonormal frame for $T_{i(p)}M$.

So then, in a calculation similar to what you have already done, $$i^{\ast} \star \omega(e_{1},e_{2}) = \star \omega \big(d_{p}i(e_{1}), d_{p}i(e_{2})\big) = \star \omega (e_{1}, e_{2}) = \omega(N),$$ in considering $\omega$ to be a 1-form on $T_{i(p)}M \cong \mathbb{R}^{3}$, $\star \omega$ for be a 2-form on $T_{i(p)}M \cong \mathbb{R}^{3}$, and finally $i^{\ast}\star\omega$ to be a 2-form on $T_{p}\partial M \cong \mathbb{R}^{2}$.

  • $\begingroup$ thanks for correct me! Why $di_p$ maps $(e_1,e_2)$ to $(e_1,e_2,N)$? Are you assuming that the map $i$ is as I defined? $\endgroup$ – George Nov 8 '18 at 21:57
  • $\begingroup$ Sorry, just realised I had made a mistake regarding that part! I have edited it so hopefully it makes sense now. I had a look at Do Carmo's book, so it should be the same as what you have. $\endgroup$ – BenCWBrown Nov 8 '18 at 22:33
  • $\begingroup$ When do Carmo prove the Stokes theorem, he states that "the inclusion map $i$ can be written as: $x_1 = 0$, $x_j = x_j$" for the case that a coordinate neighborhood intersects $\partial M$, so $i$ is defined as I defined in the OP, but I can't see why $di_p$ will be exactly what you said, because I think $e_1 \mapsto (0,1,0)$, $e_2 \mapsto (0,0,1)$ as I computed in the OP. $\endgroup$ – George Nov 8 '18 at 22:51
  • $\begingroup$ I think you wanted other thing where you wrote $(e_1,e_2) \mapsto (e_1,e_2,0)$, because this is not make sense since we identify $T_p \partial M \cong \mathbb{R}^2$ and $T_p M \cong \mathbb{R}^3$ and, in this sense, $e_1, e_2 \in \mathbb{R}^2$ and not coordinates as you put when you wrote $(e_1,e_2) \mapsto (e_1,e_2,0)$ $\endgroup$ – George Nov 8 '18 at 23:02
  • $\begingroup$ You will get the same result whether the inclusion map has $x_{1} = 0$ or $x_{3} = 0$, since the orientation of the orthonormal frame will be the same in either case. I have added more clarification to the inclusion map as its derivative now, hopefully that differentiates between what is a coordinate point and what is a tangent vector. $\endgroup$ – BenCWBrown Nov 8 '18 at 23:19

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.