Why is the geodesic with the shortest length to the boundary perpendicular to the boundary?

57 Views Asked by At

Why is the geodesic with the shortest length to the boundary perpendicular to the boundary, i.e.

Let $M$ be a Riemannian manifold with boundary $\partial M$, $x\in M$. Let $p\in\partial M$ such that ${\rm d}(x,p)={\rm d}(x,\partial M)$. Let $\gamma:[0,l]\to M$ be a shortest geodesic joining $x$ to $p$. Then $\gamma'(l)~\bot~T_p(\partial M)$.

I'm sure that this result comes from the first variation formula because I encountered this question while reading a paper and the paper states "by the first variation formula, we have $\gamma'(l)~\bot~T_p(\partial M)$". But so far I haven't come up with a rigorous and complete proof. Here is my attempt:

Let $\alpha:[0,l]\times(-\varepsilon,\varepsilon)\to M$ be a smooth variation of $\gamma$ with fixed endpoints. Set $\displaystyle V(t,s)=\frac{\partial\alpha}{\partial s}$, $\displaystyle T(t,s)=\frac{\partial\alpha}{\partial t}$. By the first variation formula, we have: $$\frac{d}{ds}\bigg|_{s=0} {\rm length}(\gamma_s) = -\int_0^l\left\langle V(t,0),\nabla_T\frac{T}{||T||}\right\rangle dt,$$ where $\gamma_s(t)$ is the variation of $\gamma$ with parameter $s$ and $t$.

Since $\gamma$ is a shortest geodesic, the length of $\gamma_s$ is minimized at $s=0$. Therefore, the derivative of the length with respect to $s$ is zero at $s=0$. This gives us: $$\int_0^l\left\langle V(t,0),\nabla_T\frac{T}{||T||}\right\rangle dt = 0.$$

How to resolve it next? Any help would be appreciated!