Perhaps this is basic knowledge in Riemannian geometry, but I can't seem to figure out the answer. Here is the precise statement of my question.

Let $M$ be a Riemannian manifold, $p$ a point in $M$. Let $R$ be small enough that $exp_p$ restricts to a diffeomorphism on the ball $B_R(0)$ of radius $R$ centered at the origin, and let $U_R$ be the intersection of $B_R(0)$ and any two dimensional plane through the origin in $T_p M$. Question: does there exist $R$ such that $exp_p(U_R)$ is geodesically convex, in the sense that for every two points of $exp_p(U_R)$ the unique geodesic segment connecting them lies entirely in $exp_p(U_R)$?

It would be really convenient for me if the answer is yes. If so, I am curious to know if the statement is still true if $P$ is replaced by a subspace of higher dimension, but I only need the result for planes.


  • $\begingroup$ Wouldn't you get a counterexample by taking $M=\mathbb{R}^2$ and $U$ any nonconvex neighborhood of the origin? $\endgroup$ – Jonas Meyer Mar 13 '10 at 22:05
  • $\begingroup$ Arg, you're right. I want $U$ to be convex as well - I'll fix the question. $\endgroup$ – Paul Siegel Mar 13 '10 at 22:16
  • $\begingroup$ Also, if it helps, I'm happy to assume that $U$ is a small ball. $\endgroup$ – Paul Siegel Mar 13 '10 at 22:18
  • $\begingroup$ Jonas is right, of course. See en.wikipedia.org/wiki/Gauss's_lemma_(Riemannian_geometry) The memory is a bit vague but this should extend to your result for very small and strictly convex $U$. If I find a reference I will post it, meanwhile also see en.wikipedia.org/wiki/Normal_coordinates and en.wikipedia.org/wiki/Exponential_map $\endgroup$ – Will Jagy Mar 13 '10 at 22:35
  • $\begingroup$ I have tried using the Gauss lemma to prove that the answer is 'yes', but I just don't see how to force a geodesic into a potentially high codimension surface. Can you elaborate? $\endgroup$ – Paul Siegel Mar 13 '10 at 23:01

No, a generic Riemannian metric does not have totally geodesic 2-dimensional submanifolds at all. The property that you ask for is very rare. For example, it implies that $R(X,Y)Y$ belongs to the linear span of $X$ and $Y$, for every $p\in M$ and every $X,Y\in T_pM$. This means point-wise constant sectional curvature if I remember correctly.

  • $\begingroup$ Ack. This ruins my evening :( Thanks, though! $\endgroup$ – Paul Siegel Mar 13 '10 at 23:27
  • $\begingroup$ Sergei -- your last sentence if a bit confusing since it seems to say that any surface (which is a totally geodesic submanifold of itself) has constant Gaussian curvature. $\endgroup$ – algori Mar 14 '10 at 3:07
  • $\begingroup$ Could not believe in your global condition (for every point in M). May be I misunderstand something. $\endgroup$ – Petya Mar 14 '10 at 3:33
  • 1
    $\begingroup$ @Petya. "Point-wise constant sectional curvature" means that sectional curvature at fixed point does not depend on direction. Schur's theorem says that if dimension is $\ge3$ then point-wise constant sectional curvature is equivalent to constant curvature. (The argument is local; it only use Bianchi identity.) $\endgroup$ – Anton Petrunin Mar 14 '10 at 3:57
  • $\begingroup$ I see know (misread)! $\endgroup$ – Petya Mar 14 '10 at 4:26

I don't think that's true. If the dimension of $M$ is bigger than that of $P$, then a necessary condition is that every tangent hyperplane $P\subset T_pM$ develops locally into a totally geodesic submanifold of $M$. This is not true for arbitrary manifolds. Some examples of manifolds in which this is true include those of constant sectional curvature. Symmetric spaces being a special subclass which verifies this conditions. Someone more awake can probably come up with a sufficient criterion.

I can't think of a very good example right now, but I am pretty sure that if take the 3-dimensional Riemannian Schwarzschild solution, start from a $r$-orthogonal plane outside of the apparent horizon, you'd get a counterexample.

If the dimension of $P$ is the same as dimension of $M$, however, then you should be okay as long as you make $U$ small enough.

Edit: Ah, for $P$ 2 dimensional and $M$ 3, by applying the Codazzi equations one sees that a necessary condition for locally developing the hyperplanes to totally geodesic submanifolds is that $Ric(X,Y) = 0$ whenever $g(X,Y) = 0$. This is obviously a very strong condition that is not satisfied by most manifolds.


No. Take the sphere with $p$ the north pole, and let $U$ be the neighborhood of $0$ in $T_p S^2$ on which $\exp_p$ takes to a diffeomorphism onto $S^2$ without the south pole. Let $\gamma$ be a geodesic through the north pole, and let $p_1$ and $p_2$ be two points on $\gamma$ in the southern hemisphere. Then the shortest geodesic segment passes through the south pole, hence leaves $U$.

The difficulty is that length minimization is a global phenomenon, whereas $\exp$ is only a local diffeomorphism. Do you have additional assumptions in your problem? e.g. non-negative curvature?

  • $\begingroup$ Yes, but your $U$ is not small. $\endgroup$ – Jonas Meyer Mar 13 '10 at 22:44
  • $\begingroup$ Small enough geodesic balls are geodesically convex, but the question is about staying in the image of the plane if $U$ is small and convex. $\endgroup$ – Jonas Meyer Mar 13 '10 at 22:46
  • $\begingroup$ True, Jonas. There should exist a small $U$ for which Paul's convexity conclusion holds, following Will's suggestion of the Gauss lemma. My point is that if you just choose a neighborhood $U$ at the beginning, the property may not hold. $\endgroup$ – Tom LaGatta Mar 13 '10 at 22:47
  • $\begingroup$ Good point. However for my purposes I only need the result for $U$ sufficiently small. I'll try to rephrase the question to make this more clear. $\endgroup$ – Paul Siegel Mar 13 '10 at 22:48
  • $\begingroup$ I just removed my temporary downvote (it was for sorting purposes). $\endgroup$ – Jonas Meyer Mar 13 '10 at 23:38

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.