Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm reading do Carmo's Riemannian Geometry, in ex6.11 d) he wrote that"every embedded hypersurface $M^{n} \subset \bar{M}^{n+1}$ is locally the inverse image of a regular value". Could anyone comment on how to show this?

To be more specific, let $\bar{M}^{n+1}$ be an $n+1$ dimensional Riemannian manifold, let $M^{n}$ be some $n$ dimensional embedded submanifold of $\bar{M}$, then locally we have that $M=f^{-1}(r)$, where $f: \bar{M}^{n+1} \rightarrow \mathbb{R}$ is a differentiable function and $r$ is a regular value of $f$.

Thank you very much!

share|cite|improve this question
Locally, the submanifold has a tubular neighborhood $U$ which looks like a product of a ball times your submanifold. Then you can take $f$ to be the projection on the ball. (If you want $f$ defined on the whole big manifold, just extend smoothly in any way) – Mariano Suárez-Alvarez May 8 '12 at 20:06
up vote 5 down vote accepted

By choosing good local coordinates, you can assume that $M = \mathbb{R}^n\subset\mathbb{R}^{n+1} = \overline{M}$. Specifically, assume that $M = \{x\in \mathbb{R}^{n+1} : x_{n+1} = 0\}$. Then $M = f^{-1}(0)$, where $f\colon \mathbb{R}^{n+1}\to \mathbb{R}$ is the map $f(x) = x_{n+1}$. Since $0$ is a regular value for $f$, this is exactly what you want.

share|cite|improve this answer
This proof also generalizes to show that codimension doesn't matter: every embedded $k$-submanifold is locally the inverse image of a regular value. – Jesse Madnick May 8 '12 at 20:31

Seems like you should be able to do this globally. Define $f: \overline M \to \mathbb R$ by $$f(x) = \textrm{dist}(x, M)$$ Then the function $\phi:\mathbb R \times M \to \overline M$ given by $$(t,p) \mapsto \exp_p(t\nabla f)$$ is, for some small $t$, a diffeomorphism (this follows from existence and uniqueness results for ODE since by definition $exp_p$ gives the geodesic from $p$ in the normal direction $\nabla f$) and therefore $0$ is a regular value for $f$.

share|cite|improve this answer
Dear treble, you have to be a little bit careful using the distance function. E.g. if $M = \{0\} \subset \mathbb R = \overline{M}$, then your function $f(x)$ is just $|x|$, which is not differentiable at $0$, making it a little tricky to talk about $0$ being a "regular value for $f$". Regards, – Matt E May 8 '12 at 21:42
This difficulty is superficial. You merely need to decide "what side" of $M$ you are on. This is equivalent to choosing a normal vector (i.e. how to define $\nabla f$) in some region of $M$, or in your example, a choice of $f = x$ or $f = -x$. The above argument then applies to either choice. But you are right, I could have clarified this in the original answer. – treble May 8 '12 at 22:00
I will further remark that if $\overline M = \mathbb R^3$ and if $(t, x, y)$ are coordinates as in my answer, then $\partial_t, \partial_x, \partial_y$ form what is known as a "Darboux Frame." Moreover, in higher dimensions the well-known Fermi coordinates are an example of the construction in my answer. I am just remarking that, in particular, these useful coordinates also answer affirmatively the OP's question. – treble May 8 '12 at 22:15
Dear treble, Sounds good! (Although for this to work globally I guess you need a hypothesis to the effect that $M$ separates $\overline{M}$.) Cheers, – Matt E May 8 '12 at 23:24
Thanks for the comments :) You are right, I was really assuming that $M$ was oriented when I was thinking that the argument should work globally. – treble May 8 '12 at 23:35

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.