Let $M$ and $N$ be manifolds with Riemannian metrics $g$ and $h$ respectively. A diffeomorphism $F: M\to N$ is an isometry if \begin{equation*} h_{F(x)}(T_x F(u), T_x F(v))=g_x(u,v) \end{equation*} for all $x\in M$ and $u,v\in T_x M$. A submanifold $\Sigma\subset M$ is called totally geodesic if for every $x\in\Sigma$ and every $v\in T_x\Sigma\subset T_x M$, the geodesic in $M$ with initial point $x$ and initial velocity is contained in $\Sigma$.

I want to show

i) The intersection of $S^n$ with any vector subspace (at least 2D) is a connected complete totally geodesic sub manifold.

Then deduce:

ii) Every connected, complete, totally geodesic submanifold of $S^n$ is the intersection of $S^n$ with vector subspace of $\mathbb{R}^{n+1}$

Here is my thought for (1): The first of my claim is that the closed totally geodesic subspaces of the sphere $S^n\subset\mathbb{R}^{n+1}$ are precisely the intersection of $S^n$ with linear subspace of $\mathbb{R}^{n+1}$. This follows from the description of the geodesics on $S^n$, according to Jurgen Jost's 4th Edn. So one would naturally look at how to determine the geodesics of $S^n$, in this case, the idea is to show the intersection of $S^n$ with any linear subspaces of $\mathbb{R}^{n+1}$ determined a great circle The orthogonal group $O(n+1)$ operates isometrically on $\mathbb{R}^{n+1}$, and since it maps $S^n$ into $S^n$, it also operate isometrically on $S^n$. Now let take a point $x\in S^n$ and a vector $v\in T_x S^n$. Let $E$ be the 2D plane through the origin of $\mathbb{R}^{n+1}$ containing $v$. We claim that the geodesic $\gamma_v$ through $x$ with tangent vector $v$ is the great circle (intersection of 2-plane with $S^n$ through the origin) through $x$ with tangent vector $v$ (parametrized by arc length). For this, let $S\in O(n+1)$ be the reflection across that 2-plane $E$. Together with $\gamma_v$, $S\gamma_v$ is also a geodesic through $x$ with tangent vector $v$. Now the uniqueness of the solution to geodesic equations implies $\gamma_v=S\gamma_v$, and thus the image of $\gamma_v$ is the great circle we want.

Hence we've shown that every tangent vector on the plane uniquely determines a connected complete totally geodesic sub manifold. Then what do I need to do for (2), i.e. to conclude every connected complete totally geodesic complete totally geodesic submanifold of $S^n$ is the intersection of $S^n$. I am lost, any correction on what I've written is appreciated.

  • $\begingroup$ A vector subspace of one dimension cuts the sphere in two distinct points????? $\endgroup$ – Semsem Feb 26 '14 at 5:49
  • $\begingroup$ sorry i don't understand what you mean…could you point out where in what i've written is not clear to you? I guess you mean the fact that $S$ is the reflection across the 2D plane means cutting the sphere in 2 distinct points. $\endgroup$ – math101 Feb 26 '14 at 5:55
  • $\begingroup$ in point i: the intersection of $S^n$ and a vector subspace of dimension one is not connected $\endgroup$ – Semsem Feb 26 '14 at 5:58

First of all, the comment by Semsem is valid. If the vector space is one dimensional, then $V\cap \mathbb S^n$ is not connected. That is, statement (i) is correct only if you consider vector subspaces that are not one dimensional.

Let $M$ be a totally geodesic submanifold of $\mathbb S^n$. After some $O(n+1)$ on $\mathbb S^n$, we assume that the north pole $N=(0,\cdots, 0, 1)$ is in $M$. Let $L = T_NM$ and treat this space as a vector subspace in $\mathbb R^{n+1}$. Let $V = N \oplus L$. It is a vector subspace in $\mathbb R^{n+1}$.

Now we show $M = G := V \cap \mathbb S^n$ (Note that $V$ is at least two dimensional so $G$ is connected). First of all, let $x\in G$. Then there is a geodesic $\gamma$ such that $\gamma(0) = N$ and $\gamma(1) =x$. Then this is a portion of a great circle and thus $\dot \gamma (0) \in L$. As $M$ is totally geodesic, $x=\gamma(1)\in M$. Thus $G\subset M$.

On the other hand, let $q\in M$. Then as $M$ is connected and complete, there is a geodesic in $M$ joining $N$ and $q$. But this geodesic is also a geodesic in $\mathbb S^n$, as $M$ is totally geodesic. Thus this geodesic is a great circle and so are inside $G$ (As its tangent vector at $N$ is by definition in $L$). Thus $M\subset G$.

  • $\begingroup$ Well, even if the subspace is $1$ dimensional, it is true that every geodesic of the sphere whose tangent vector is tangent to the intersction with the subspace is totally contained in the intersection: there are no such geodesics! $\endgroup$ – Mariano Suárez-Álvarez Feb 26 '14 at 7:11
  • $\begingroup$ @MarianoSuárez-Alvarez: You are right. What I want to say is that the space is not connected (While it said it's connected in (i)) $\endgroup$ – user99914 Feb 26 '14 at 7:13
  • $\begingroup$ can you (or anyone) remind me what we mean by "connected", and the difference between a "connected" Riemanian manifold and a non-connected one? $\endgroup$ – math101 Feb 26 '14 at 10:40
  • $\begingroup$ @algebra_geometry101: There is in general a definition of connectedness of a topological space. But in the case of a manifold, connectedness is the same as "for all $p$ and $q$ in $M$, there is a continuous paths joining $p$ and $q$". So for example, if $M$ is just a set of two points, then it is not connected in this sense. $\endgroup$ – user99914 Feb 27 '14 at 7:10

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.