MathOverflow is a question and answer site for professional mathematicians. 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

Let $\Sigma$ be a smooth convex surface in Euclidean 3-space and $\gamma$ be a unit speed minimizing geodesic in $\Sigma$. Assume that for some $a < b < c$, we have $$\gamma'(a)=\gamma'(b)=\gamma'(c).$$

Is it true that $\gamma'$ is constant on one of two intervals $[a,b]$ or $[b,c]$?


  • This is a simplification-variation of an other question I heard from D. Burago.
  • It is not hard t construct an example of a minimizing geodesic such that $\gamma'(a)=\gamma'(b)$, but $\gamma'$ is not a constant on $[a,b]$. (See the example below.)
  • I would be also interested in the analog for $n$ points.
  • This paper: Total curvature and spiralling shortest paths by Barany, Kuperberg, Zamfirescu is relevant.

Example. I will construct a convex polyhedron, but it is easy to smooth. Consider polyhedron defined by 5 inequlaities: $$z\ge 0,\ \ |x+\alpha{\cdot}y|\le \alpha\ \ \text{and}\ \ z\pm\beta{\cdot}y\le \beta$$ and look at the minimizing geodesic between points $(0,1-\epsilon, \beta{\cdot}\epsilon)$ and $(0,-(1-\epsilon), \beta{\cdot}\epsilon)$. For appropriately chousen $\alpha$, $\beta$ and $\epsilon$ this minimizing geodesic will pass through the faces in this order $$\{z+\beta{\cdot}y= \beta\},\ \ \{x+\alpha{\cdot}y= \alpha\},\ \ \{z=0\},\ \ \{x+\alpha{\cdot}y= -\alpha\},\ \ \{z-\beta{\cdot}y= \beta\}$$ and it will have the same velocity vector $(1,0,0)$ on both faces $\{z+\beta{\cdot}y= \beta\}$ and $\{z-\beta{\cdot}y= \beta\}$.

share|cite|improve this question
My reaction to this is that this doesn't sound that hard but if you and Dmitri can't answer it, it's not likely I can either. Nice question, though. I will have to give it a try. – Deane Yang May 2 '11 at 16:32
I realize that it should be just homework for me, but could you provide the details of the example with two points with the same velocity? – Deane Yang May 2 '11 at 19:16
Presumably gluing your $[a,b]$ example to its reverse can only be realized on a nonconvex surface? – Joseph O'Rourke May 2 '11 at 19:53
@Deane: I add the example. – Anton Petrunin May 2 '11 at 22:48
Anton, thanks for the example. – Deane Yang May 3 '11 at 3:48

Just to help visualize the example, if I have interpreted the description correctly, this is one view, for $\alpha=\beta=1$, and $\epsilon=\frac{1}{16}$.
In this view, the $x$-axis is horizontal, the $z$-axis vertical.

Addendum. The shortest path (yellow) from $p_1$ to $p_2$ follows the face sequence (Bk, R, Bt, L, F).
I tried to indicate (in the nearly invisible short dashed lines) the initial direction of the path on the (mauve) Bk and F faces, in their planar unfoldings. I should note that these initial velocities are not exactly $(1,0,0)$, which presumably is only achieved by "appropriately chosen" $\alpha, \beta, \epsilon$.

share|cite|improve this answer
Thanks for the picture. It's really helpful. – Deane Yang May 3 '11 at 13:03
On the second diagram, the angles of rotations at $p_1$ and $p_2$ should be bit more than π/2. – Anton Petrunin May 4 '11 at 17:00
Awesome picture, as always. – Thomas Richard Apr 19 '12 at 20:22

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.