This question stems from the discussion in:

how to define the injectivity radius of manifolds with boundary?

Suppose $(M,g)$ is a compact Riemannian manifold with boundary. In this context, let the injectivity radius of a point $x$ be the minimum distance from $x$ at which there is a point $y$ with more than one length-minimizing geodesic connecting $x$ to $y$.

Is it true that the injectivity radius as defined this way is bounded below by some nonzero value? If so, is there a standard reference for this fact?

In the discussion linked above, the following theorem is referenced:

Corollary 2. If for a complete Riemannian manifold with boundary, M, the sectional curvatures of the interior and the outward sectional curvatures of the boundary are no greater than $K$, then $N(p,\frac{\pi}{2K})$ is open in M and the distance function from p is convex on $N(p,\frac{\pi}{2K})$.

Where $N(p,\frac{\pi}{2K})$ is the set of points connected to $p$ by a unique geodesic of length $\frac{\pi}{2K}$ or less.

(Reference Paper) https://www.ams.org/journals/tran/1993-339-02/S0002-9947-1993-1113693-1/S0002-9947-1993-1113693-1.pdf

This seems like it is close to the result I am looking for. Earlier in the paper, it is also stated that there are no conjugate points in $N(p,\frac{\pi}{K})$.

Is there a simple step from this result that proves that the injectivity radius is nonzero?


Yes, they show that any compact Riemannian manifold with boundary is locally $\mathrm{CAT}(\kappa)$ for some $\kappa\in\mathbb{R}$. In particular the injectivity radius is positive.

  • 1
    $\begingroup$ This definitely works but is it necessary to go the generality of $CAT$-spaces? This should be proveable purely in Riemannian geometry language. $\endgroup$ – quarague Aug 21 '19 at 6:37
  • 2
    $\begingroup$ @quarague I do not have an example of curvature-free estimate on injectivity radius. So if there is such a proof, then it might give something new in Riemannian geometry. $\endgroup$ – Anton Petrunin Aug 21 '19 at 16:00
  • 1
    $\begingroup$ You definitely need curvature, I just think you don't need $CAT$-spaces. A $CAT$-space is by definition a metric space that satisfies a triangle comparison inequality. Historically this is a generalisation from Riemannian manifolds with bounded curvature. One can show that a Riemannian manifold with curvature bounded from above is a $CAT$-space but this is why $CAT$-spaces are defined the way they are. The question of OP was known before $CAT$-spaces where invented/defined. So one should be able to prove it directly with Riemannian geometry machinery. $\endgroup$ – quarague Aug 22 '19 at 6:46

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.