I'm reading the proof of Stokes theorem at page 83 of "Godinho, Natàrio, An introduction to Riemannian geometry" and I can't understand a passage in it, probably because the definition of orientability is not very clear to me. The definition of orientability and orientation used in the book are:

Definition (Oriented manifold) Let $M$ be a $n$-differentiable manifold. An orientation for $M$ consists in a choice of orientation for all the $n$-dimensional real vector spaces $T_pM$ s.t. for every $p \in M$ there exist a parametrization $(U, \varphi)$ s.t. \begin{equation} \text{the linear map } \, \, d\varphi_x : T_xU \cong \mathbb{R}^n \to T_{\varphi(x)}M \, \, \text{ is orientation preserving} \quad \quad \quad (*)\end{equation} where $\mathbb{R}^{n}$ is equipped with the standard orientation (i.e. the one that assigns a positive sign to the standard basis). A manifold admitting an orientation is called oriented manifold. A parametrization $(U, \phi)$ is said to be compatible with the orientation if it satisfies (*).

I also know that

Lemma A $n$-differentiable manifold is orientable iff there exists an atlas s.t. the (matrices representing the) differentials of the overlapping maps have positive determinant.

And then I have this definition of induced orientation on $\partial M$:

Definition (Induced orientation) Let $M$ be a $n$-differentiable manifold with boundary $\partial M$ (which is a $(n-1)$-differential manifold). Chosen coordinates $(x^1, \dots, x^n)$ in $M$ around $p \in \partial M$ then $(x^1, \dots, x^{n-1})$ are coordinates in $\partial M$ around $p$. $ \{ \partial_1, \dots, \partial_n \}$ is a basis for $T_p(M)$ and $\{ \partial_1, \dots, \partial_{n-1} \}$ is a basis for $T_p (\partial M)$. We call induced orientation by $M$ on $\partial M$ the one for which a basis $\mathcal{B}$ of $T_p(\partial M)$ has positive sign if it does the basis $ \{ -\partial_n, \mathcal{B} \}$ of $T_pM$.

Then in the proof of Stokes theorem I have a parametrization of $M$ $(U, \varphi)$ with $U \subset \mathbb{H}^n$ s.t. $\varphi(U) \cap \partial M \ne \emptyset$. Then the book defines $$ \tilde{U} : = \{ (x^1, \dots, x^{n-1} ) \mid ((-1)^n x^1, \dots, x^{n-1}, 0) \in U \} $$ and the parametrization $$ \tilde{\varphi} : \tilde{U} \to \partial M \quad \quad (x^1, \dots, x^{n-1}) \mapsto \varphi ((-1)^n x^1, \dots, x^{n-1}, 0)) $$

My question : why is $(\tilde{U}, \tilde{\varphi})$ a parametrization of $\partial M$ preserving the induced orientation of $M$ on $\partial M$?

Edit (Just to see if I understood the suggestions of Balloon) Define $$ \overline{U} := \{ (x^1, \dots, x^{n-1}) \in \mathbb{R}^{n-1} \mid (x^1, \dots, x^{n-1}, 0) \in U \} \\ \overline{\varphi} := \varphi \biggr |_{\overline{U}} : \overline{U} \to \partial M \\ \tilde{U}:= \{ (x^1, \dots, x^{n-1} \in \mathbb{R}^{n-1} \mid ((-1)^n x^1, \dots, x^{n-1}) \in \overline{U} \} \\ \tau : \tilde{U} \to \overline{U} \quad \quad (x^1, \dots, x^{n-1}) \mapsto ((-1)^n x^1, \dots, x^{n-1})\\ \tilde{\varphi} := \overline{\varphi} \circ \tau : \tilde{U} \to \partial M $$ I want to prove $(\tilde{U}, \tilde{\varphi})$ is a parametrization of $\partial M$ preserving the standard orientation of $\mathbb{R}^{n-1}$ w.r.t. the orientation induced by $M$ on $\partial M$.

It is enough to prove that the determinant of the Jacobian of $\tilde{\varphi}$ (written w.r.t. the standard basis of $\mathbb{R}^{n-1}$ and a positive basis in $\partial M$) is positive at each point $x \in \tilde{U}$ . I observe that $$ d\tilde{\varphi}_x = d \overline{\varphi}_{\tau(x)} \circ d\tau_x $$

  1. $d \tau_x = (-1)^n I_{n-1} $ when I write it w.r.t. the standard basis (both) of $\mathbb{R}^{n-1}$ and then $\det(d\tau_x) = (-1)^n $ for any $x \in \tilde{U}$.

  2. I know that $d \overline{\varphi}_y (e^k) = \partial_k$ for every $k=1, \dots, n-1$ and every $y \in \overline{U}$. And then $d \overline{\varphi}_{\tau(x)}=I_{n-1}$ when I write it w.r.t. the standard basis of $\mathbb{R}^{n-1}$ and the basis $\mathcal{B}=\{\partial_1, \dots, \partial_{n-1} \}$ of $T_{\tilde{\varphi}(x)} \partial M$. Then, since $\mathcal{B}$ has $(-1)^n$ times the sign of a the positive basis in $T_{\tilde{\varphi}(x)} \partial M$, $d \overline{\varphi}_{\tau(x)}=(-1)^nI_{n-1}$ when I write it w.r.t. the standard basis of $\mathbb{R}^{n-1}$ and a positive basis of $T_{\tilde{\varphi}(x)} \partial M$. Then $\det(d \overline{\varphi}_{\tau(x)})= (-1)^n$ for any $x \in \tilde{U}$.

And then $\det(d\tilde{\varphi}_x) = (-1)^n (-1)^n = 1 >0$ for any $ x \in \tilde{U}$ when I write the jacobian matrix w.r.t. the standard basis of $\mathbb{R}^{n-1}$ and a positive basis in $T_{\tilde{\varphi}(x)} \partial M$.


You want your basis $(\varepsilon\partial_1,\dots,\partial_{n-1})$ of $T(\partial M)$ being such that $(-\partial_n,\varepsilon \partial_1,\dots,\partial_{n-1})$ is a direct basis of $TM$, where $\varepsilon=\pm1$ is a sign we will adjust. We will try to write this in term of the standard basis $(\partial_1,\dots,\partial_n)$, which is direct by assumption of your parametrization being orientation-preserving. If we want to put the $-\partial_n$ in the $n$-th position, we have to apply $(n-1)$ transpositions, changing each time the sign of the orientation to finally get \begin{align*} (-\partial_n,\varepsilon\partial_1,\dots,\partial_{n-1})\text{ is direct }&\iff (-1)^{n-1}(\varepsilon\partial_1,\dots,\partial_{n-1},-\partial_n)\text{ is direct}\\ &\iff\varepsilon(-1)^n(\partial_1,\dots,\partial_n)\text{ is direct.} \end{align*}

Then precomposing by the map your book describes make appear the correct $\varepsilon=(-1)^n$, which guarantees that the parametrization $\tilde{\varphi}:\tilde{U}\to\partial M$ is orientation preserving.

For example, the orientation induced by $\mathbb{H}^2$ on $\mathbb{R}=\partial\mathbb{H}^2$ is the standard one, while the one induced by $\mathbb{H}^3$ on $\mathbb{R}^2=\partial\mathbb{H}^3$ is the opposite of the standard one.

  • $\begingroup$ Thank you for your answer. The first part is clear to me, I understand why the basis $(\partial_1, \dots, \partial_{n-1})$ has $(-1)^n$ the sign of the positive orientation induced by $M$ on $\partial M$. What I can't understand is why $\tilde{\phi}$ is orientation preserving. I should prove that its differential at any point maps the standard basis of $\mathbb{R}^{n-1}$ is a positive oriented basis in $T_pM$ or something similar... $\endgroup$ – Bremen000 Oct 25 '18 at 18:38
  • $\begingroup$ As a shortcut to show that, you can note that $\tilde\varphi:\tilde U\to \partial M$ is the composition $\varphi\circ \tau$, where $\tau:\tilde U\to U$ is the the map $(x_1,\dots,x_{n-1})\mapsto((-1)^nx_1,\dots,x_{n-1})$. So $$\det (d\tilde\varphi)=\det(d\varphi)\det(d\tau)=(-1)^n(-1)^n=1,$$ so is orientation-preserving. $\endgroup$ – Balloon Oct 25 '18 at 18:47
  • $\begingroup$ Oh sure! Thank you! It was really simple! $\endgroup$ – Bremen000 Oct 25 '18 at 18:49
  • $\begingroup$ No problem, you are welcome! $\endgroup$ – Balloon Oct 25 '18 at 18:49
  • $\begingroup$ Ok I thought it was clear to me, but it is not. I understand why $\det(d\varphi) = (-1)^n$ but, in order to consider the composition it must be $\tau: (x_1, \dots, x_{n-1}) \mapsto ((-1)^n, \dots, x_{n-1}, 0 )$ and its jacobian matrix is not squared! I'm missing something, could you please write me the details? $\endgroup$ – Bremen000 Oct 26 '18 at 12:08

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.