0
$\begingroup$

Let $(M^2,g)$ be a complete, two-dimensional Riemannian manifold be given; also given is $\gamma: [0,\infty) \to M$, an injective geodesic in $M$.

Suppose there are two geodesic segments $\gamma_i : [a_i,b_i] \to M$, $i = 1,2$ with endpoints on $\gamma$ and \begin{equation} \gamma_1(b_1) = \gamma_2(a_2). \end{equation} Moreover, the points $\gamma_1(a_1),\gamma_1(b_1) = \gamma_2(a_2), \gamma_2(b_2)$ appear in chronological order along $\gamma$.

You may also assume that $\gamma_1,\gamma_2$ 'lie on either side of $\gamma$', by which I mean that \begin{equation} \langle \gamma_1'(b_1), n \rangle \text{ and } \langle \gamma_2'(a_2) , n \rangle \end{equation} have the same sign, writing $n \in T_{\gamma_1(b_1)}M$ for the unit normal to $\gamma$.

The concatenated, piecewise defined geodesic $\gamma_2 \! \cdot \! \gamma_1$ intersects $\gamma$ three times. Is there a genuine geodesic $\Gamma \neq \gamma$ in $M$ that intersects $\gamma$ three or more times?

Edit. I've specialized the original question (which was phrased in terms of $N$ geodesic segments), to the special case where $N = 2$.

$\endgroup$
7
  • 1
    $\begingroup$ Can you clarify what you are asking a bit more? It seems like it should be easy to produce such $\gamma$ and $\Gamma$ on the cylinder (e.g. with a $\gamma$ a straight line and $\Gamma$ a helix). $\endgroup$
    – RBega2
    Commented Nov 4, 2023 at 13:42
  • 1
    $\begingroup$ @RBega2 I was thinking to a situation on a cylinder as a counterexample, no? $\endgroup$ Commented Nov 4, 2023 at 14:50
  • 1
    $\begingroup$ @RBega2 I've clarified the question; the manifold $M$ and the geodesic $\gamma$ are meant to be given. $\endgroup$
    – Leo Moos
    Commented Nov 4, 2023 at 15:56
  • 1
    $\begingroup$ @PietroMajer Would you like to expand on your comment? On a standard cylinder, $\gamma$ would be a spiral or a straight line; both intersect other spirals infinitely often, no? $\endgroup$
    – Leo Moos
    Commented Nov 4, 2023 at 15:57
  • 1
    $\begingroup$ @LeoMoos For a moment I thought there could be a configuration of $\gamma_i$ that may skip the other geodesics so as to get less than $N+1$ intersections, but yes there are too many possibilities $\endgroup$ Commented Nov 4, 2023 at 18:03

1 Answer 1

2
$\begingroup$

Here's a counterexample.

Take a cone of unrolled angle strictly between $\frac\pi2$ and $\pi$. Smooth the vertex and take $\gamma$ to be a radial ray far from the vertex. Then you can make a piecewise geodesic which loops around and intersects $\gamma$ as many times as you like, but no genuine geodesic intersects it more than twice.

$\endgroup$
2
  • $\begingroup$ Thanks for the answer! Are you sure the example works, though? It seems to fail the first condition, as the points are not in chronological order along $\gamma$, no? Maybe you could make it work by choosing a non-rotationally symmetric smoothing of the vertex. $\endgroup$
    – Leo Moos
    Commented Nov 6, 2023 at 11:12
  • $\begingroup$ In the unrolled sector, there are geodesics from $1$ to $2e^{i\phi}$, $2$ to $3e^{i\phi}$, etc $\endgroup$
    – Zack
    Commented Nov 6, 2023 at 12:31

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .