Is there a 2 dimensional Riemannian manifold $M$ whose curvature is not negative but its geodesic flow is an ergodic flow?

  • $\begingroup$ @thedude but am I mistaken to think that it is not ergodic because the unit tangent bundle possess two disjoint copy of the torus $T^2$ which separate the unit tangent bundle and is invariant under the geodesic flow? $\endgroup$ Jul 8 '18 at 14:28
  • $\begingroup$ @RW What is the argument for ergodicity? May I ask you to read my previous comment? $\endgroup$ Jul 8 '18 at 14:29
  • 1
    $\begingroup$ @Ali Taghavi - Oups - sorry - of course the geodesic flow on the torus is not ergodic - since the slope is preserved $\endgroup$
    – R W
    Jul 8 '18 at 14:45
  • $\begingroup$ I believe that if $M$ is a closed hyperbolic surface, and if you then alter the metric to be negatively curved except at a single point of zero curvature, the resulting geodesic flow will still be ergodic. I do not have a proof at hand. $\endgroup$
    – Lee Mosher
    Jul 8 '18 at 15:03
  • 1
    $\begingroup$ @AliTaghavi I thought the geodesic flow was defined on the manifold, but now I see it is defined on the tangent bundle. So I learned something from your question. Thank you. $\endgroup$
    – thedude
    Jul 8 '18 at 19:19

It was proved by Donnay that any compact orientable surface can be given a Riemannian metric for which the geodesic flow is ergodic. See Theorem 1 of this article. On the other hand, there are no negatively curved metrics on a sphere or a torus.


Anosov flows are ergodic and a geodesic flow can be Anosov even if the curvature is not strictly negative. This was studied by Eberlein in the seventies, in an article from 1973 entitled when is a geodesic flow of Anosov type? and other authors (Klingenberg etc).

A byproduct of this study is the following: on a nonpositively curved surface, the geodesic flow is of Anosov type as soon as there is a point of negative curvature on all geodesics. So you can start with a surface with negative curvature, make a small perturbation so that the curvature vanishes on a single point $p$, and the geodesic flow is still Anosov and ergodic. Now the Anosov condition is an open one so you can make again a small perturbation so that the curvature becomes positive at the point $p$ and you still have an ergodic Anosov geodesic flow. There are many variations around these ideas.

  • $\begingroup$ Thank you very much for this very great answer. I am sorry that I can not accept the two answers simultaneously. meta.mathoverflow.net/questions/1491/… $\endgroup$ Jul 8 '18 at 19:49
  • $\begingroup$ But a slight perturbation of a compact surface with strictly negative curvature is stile a surface with strictly negative curvature. Am I right? But i think that there is modification of your argument. I try to find the valuable reference you mentioned in your answer and learn them. thanks again for your attention to my question and these very helpful references $\endgroup$ Jul 9 '18 at 7:54
  • 1
    $\begingroup$ The curvature is zero at a point, the flow is Anosov, you make a perturbation such that the curvature becomes positive around the point and the flow stays Anosov. $\endgroup$
    – coudy
    Jul 9 '18 at 16:26

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.