3
$\begingroup$

Let $p:M\to \mathbb{R}^{n+1}$ be the closed immersed hypersurface. Is the following thing right? If there exists a point $x_0$ in $\mathbb{R}^{n+1}$ such that $\langle x(p)-x_0,\nu(p)\rangle>0$ for all $p\in M$, then $M$ is diffeomorphic to a sphere. Here $\nu(p)$ is the outward normal vector.

$\endgroup$
2
  • 1
    $\begingroup$ A cylinder in $\mathbb{R}^3$ satisfies your condition (with $x_0=0$ for the cylinder $\{(x,y,z)|x^2+y^2=1\}$). Maybe you want to assume compactness of $M$ ? $\endgroup$ Jul 22, 2014 at 12:38
  • 8
    $\begingroup$ Please use $\langle a,b\rangle$ (with \langle and \rangle) rather than $<a,b>$ for the inner product. Besides making the formulas more readable, it makes fewer people furious. (I took the liberty of making this change.) $\endgroup$ Jul 22, 2014 at 12:41

2 Answers 2

1
$\begingroup$

Assuming that $M$ is connected compact without boundary the answer is YES.

  1. Consider a map $\pi: p \mapsto \frac{x(p)-x_0}{||x(p)-x_0||}$ from $M$ to $S^n$.

  2. $\pi$ is an immersion since $\langle x(p)-x_0,\nu(p)\rangle >0$.

  3. $\pi$ is a covering since $\pi$ is an immersion, $M$ is a closed $n-$manifold, and $S^n$ is connected.

  4. $\pi$ is a homeomorphism since it is a covering over the simply connected $S^n$ . So $M$ is diffeomorphic to $S^n$.

$\endgroup$
2
  • $\begingroup$ Can you tell me the details of step 2? Why $\pi$ is an immersion? $\endgroup$
    – user129602
    Jul 23, 2014 at 20:36
  • $\begingroup$ Consider a map from $\mathbb{R}^{n+1}\setminus x_0$ to $S^n$, $\pi':x\mapsto\frac{x-x_0}{\|x-x_0\|}$. It is clear that $D\pi':T_x\mathbb{R}^{n+1}\to T_{\pi'(x)}S^n$ is the orthogonal projection along the vector $x-x_0$ which generates the kernel of $D\pi'$. We have $\pi=\pi'\circ p$ where $p$ is the embedding of $M$. So we have $D\pi=D\pi'\circ Dp$. Condition $\langle x(p)-x_0,\nu(p)\rangle>0$ implies that the image of $Dp$ does not contain $x-x_0$ - the kernel of $D\pi'$. Hence $D\pi$ is an injection, and more over, an isomorphism. $\endgroup$ Jul 24, 2014 at 8:34
2
$\begingroup$

Let me suppose that $M$ is in addition compact, and for simplicity that $x_0=0$. Let $\phi : M \to S^n$ be the radial projection, which is a smooth map.

If $\phi$ is not an immersion, there is a tangent vector $v \in T_x M$ such that $Dp(v)$ is parallel to the vector $p(x) \in \mathbb{R}^{n+1}$. But $$\langle v, \nu(x)\rangle=0$$ as the tangent and normal spaces are orthogonal, so $\langle p(x), \nu(x)\rangle=0$, a contradiction. Thus $\phi$ is an immersion.

But $\phi$ is also proper, and a proper codimension 0 immersion is a covering map. As any compact covering space of $S^n$ is a finite union of copies of $S^n$ (this is phrased strangely so that it still holds for $n=1$), it follows that $M$ is diffeomorphic to a finite union of copies of $S^n$.

One clearly can't insist that it is a single copy of $S^n$ in general, just by taking e.g. two spheres of different radii. So assume $M$ is connected if this is what you want.

$\endgroup$
1
  • $\begingroup$ Here you use $Dp(v)$ is parallel to $p(x)$ and $\langle v,\nu(x)\rangle=0$ to conclude that $\langle p(x),\nu(x)\rangle =0$. But I don't see how to prove this unless you have $Dp(v)$ is parallel to $v$. $\endgroup$
    – user129602
    Jul 23, 2014 at 20:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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