Let $M$ be a Riemannian manifold, and let $x, y \in M$ be non-conjugate points.

Let $r, R>0$ be two numbers. I am looking for a bound on the number of geodesics between $x$ and $y$ of Length between $r$ and $R$, like "there are at most $N(r, R)$ geodesics between $x$ and $y$", depending on the curvature of $M$.

It seems that one needs assumptions on the curvature to obtain such bounds. For example, one might want to assume that that sectional curvature is bounded from below by some $\kappa < 0$.

\Edit: The post of Robert Bryant suggested that one should also assume that $M$ has positive injectivity radius, or that $M$ is compact.


I think that no such bounds exist. For example, you can take $M$ to be a (complete) surface of revolution in Euclidean $3$-space that has negative curvature bounded from below by $-1$ and that is asymptotic to the classic pseudospherical surface (i.e., a 'spike' going off to infinity). On such a surface, you can find a sequence of pairs of nonconjugate points $(x_i,y_i)$ that are very close together while the $x_i$ go off to infinity along the spike and are such that the number of geodesics from $x_i$ to $y_i$ of length between $r$ and $R$ is at least $i$.

  • $\begingroup$ Thank you for pointing this out. If one assumes $M$ to be compact, such a bound would exist, wouldn't it? $\endgroup$ – Matthias Ludewig Aug 23 '13 at 15:52

I think that for any manifold with negative curvature having positive injectivity radius $\delta>0$ you get an upper bound of the form $C\delta^{-n} e^{c(R+d(x,y))}$ where $n$ is the dimension $c,C$ depend on the curvature and the dimension. The argument I have in mind actually yields a bound depending only on the injectivity radius at $x$ (or $y$) and goes as follows: if you fix one geodesic arc from $x$ to $y$ (say one of the shortest ones), and fix a lift $\tilde x$ of $x$ to the universal cover, then any other arc of length less than $R$ gives you an element in the fundamental group which displaces $\tilde x$ of less than $R+d(x,y)$, and two distinct arcs yield distinct elements. Now there is a well-known bound for the number of translates of $\tilde x$ by elements of the fundamental group at a distance $\le d$ of $\tilde x$ which is of the form $C{\rm inj}(x)^{-n}e^{cd}$ (where $c,C$ come from the asymptotics for the volume of balls in the universal cover), which in turn yields the claimed bound for the number of arcs. (Note that the lower bound $r$ does not play a role if the growth is indeed exponential).

  • $\begingroup$ Do you have a reference for the well-known bound you mention? $\endgroup$ – Matthias Ludewig Aug 24 '13 at 9:10
  • $\begingroup$ The bounds for the volume of the balls is from a paper of Knieper (GAFA 1997), the bound follows by considering a small ball around $\tilde x$ which embeds in the quotient. $\endgroup$ – Jean Raimbault Aug 24 '13 at 10:12

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.