3
$\begingroup$

Let $M$ be a complete, connected Riemannian manifold without boundary. Given a point $p\in M$ and a subset $K$ of $S_p M$, the unit sphere in $T_p M$, define the $K$-cone of directions $C(K)$ around by $C(K) := \{w \in T_p M| \ w/|w| \in K\}$.

Given $p$ and $K$ as above, define $\text{inj}(p, K)$ to be the supremum of values $r \geq 0$ for which the exponential map at $p$ is injective on $C(K) \cap B_r (0)$.

Suppose that

$\sup_{p \in M} \sup_{K \subset S_p M, K \text{ non-null}} \text{inj}(p, K) < \infty$.

Does it follow that the manifold $M$ is compact?

Note: By non-null, I mean not of null measure as a subset of the smooth manifold $S_p M$.

$\endgroup$
2
  • $\begingroup$ Your condition implies all geodesics are closed - let $K = \{ v \}$ for some $v \in T_p M \setminus \{0\}$ - so the distance function on $M$ is bounded and hence, $M$ must be compact. $\endgroup$ Commented May 13, 2021 at 9:15
  • $\begingroup$ Wouldn’t $\{v\}$ have null measure though? $\endgroup$
    – Nate River
    Commented May 13, 2021 at 9:16

1 Answer 1

3
$\begingroup$

Such a manifold has bounded diameter, and is therefore compact. To see this, let $0 < T < \infty$ be the quantity you define. I claim that $\mathrm{diam} M \leq T$.

Additionally for each point $p \in M$ and unit tangent vector $v \in T_p M$ let $\tau(v)$ be the first conjugate time along the geodesic $\gamma: t \mapsto \exp_p(tv)$. Explicitly, $\tau(v) > 0$ is the first time $t > 0$ so that $\gamma(t)$ is a conjugate point of $p$ along the geodesic.

Then $\tau(v) \leq T$ for all unit vectors, which can be obtained as follows. Let $\rho_j \downarrow 0$ and $K_j = \{ w \in T_p M \mid \lvert w \rvert = 1, \lvert w - v \rvert \leq \rho_j \}$. By definition of $T$ there are vectors $v_j \neq w_j \to v$ so that the corresponding geodesics intersect before this time. Therefore $\gamma$ has a conjugate point before or at $\gamma(T)$.

It is well-known that a geodesic is not length-minimising past its first conjugate point. Therefore every length-minimising geodesic has length at most $T$, and the claim is proved.

Edit. To see why $\gamma$ must reach a conjugate point before time $T$, one can argue as follows. Let $q_j \neq p$ be the first intersection point of the geodesics directed by $v_j,w_j$. Because $v_j,w_j \to v$ we may extract a subsequence so that $q_j \to q$ as $j \to \infty$, where $q = \gamma(t)$ for some $t > 0$. This point $q$ is conjugate to $p$. If $D \exp_p(t v)$ were non-singular then there would exist a neighbourhood $U \subset T_p M$ of $tv$ so that $\exp_p \mid U$ is a diffeomorphism onto its image. But eventually $t v_j \neq t w_j$ both lie in $U$, and $q_j$ in its image. In particular $\exp_p$ is not injective in $U$.

$\endgroup$
3
  • $\begingroup$ Sorry, how do you get the conjugate point for $\gamma$ from the $v_j$ and $w_j$? I assume you take some convergent subsequence of the intersection points, converging to, say $p_0$, then define a Jacobi field that vanishes at $p_0$ using the geodesics associated to the $v_j$ and $w_j$? However I'm not sure how to contruct this explicitly. $\endgroup$
    – Nate River
    Commented May 13, 2021 at 14:41
  • 1
    $\begingroup$ I've included additional details, arguing along the lines you suggested - does that answer your question? $\endgroup$
    – Leo Moos
    Commented May 13, 2021 at 15:20
  • $\begingroup$ Yes, thank you! I’ve marked the answer as accepted. $\endgroup$
    – Nate River
    Commented May 13, 2021 at 15:33

You must log in to answer this question.

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