Let $M$ be a smooth Riemannian manifold and let $p \in M$. Suppose $r \ll \text{inj}(p)$ (the injectivity radius at $p$) and fix $t \in (0,r)$ then define the map $$ T_pM \ni v \mapsto \exp_p(tv) \in M$$ I know this map is a diffeomorphism. How can I estimate its Lipschitz constant?

If I take $v_1, v_2 \in T_pM$ and I consider the geodesics $\gamma_1$ and $\gamma_2$ starting at $p$ with initial speed $v_1$ and $v_2$ respectively, it would be enough to prove $$ \frac{d}{dt} d \left(\gamma_1(t), \gamma_2(t)\right) \le |v_2-v_1| $$ to conclude that the Lipschitz constant is bounded by $t$. But even if I consider the differential of the distance given by $$ \frac{d}{dt} d (\gamma_1(t), \gamma_2(t))= \langle \dot{\alpha_t}(d_0), \dot{\gamma_2}(t) \rangle - \langle \dot{\alpha_t}(0), \dot{\gamma_1}(t) \rangle $$ where $d_0 = (\gamma_1(t), \gamma_2(t))$ and $\alpha_t$ is a unit speed geodesic connecting $\gamma_1(t)$ to $\gamma_2(t)$, I only obtain \begin{align*} \frac{d}{dt} d (\gamma_1(t), \gamma_2(t)) & = \langle \mathcal{P}^{\gamma_2}_{t \to 0}[\dot{\alpha_t}(d_0)], v_2 \rangle - \langle \mathcal{P}^{\gamma_1}_{t \to 0}[\dot{\alpha_t}(0)], v_1 \rangle \\ & = \langle \mathcal{P}^{\gamma_2}_{t \to 0}[\dot{\alpha_t}(d_0)] , v_2 \rangle -\langle \mathcal{P}^{\gamma_2}_{t \to 0}[\dot{\alpha_t}(d_0)], v_1 \rangle + \langle \mathcal{P}^{\gamma_2}_{t \to 0}[\dot{\alpha_t}(d_0)]-\mathcal{P}^{\gamma_1}_{t \to 0}[\dot{\alpha_t}(0)], v_1 \rangle \\ & \le |v_2-v_1| + \langle \mathcal{P}^{\gamma_2}_{t \to 0}[\dot{\alpha_t}(d_0)]-\mathcal{P}^{\gamma_1}_{t \to 0}[\dot{\alpha_t}(0)], v_1 \rangle \end{align*} where $\mathcal{P}^{\gamma}_{t_0 \to t_1}[v]$ is the parallel transport of $v \in T_{\gamma(t_0)}$from $\gamma(t_0)$ to $\gamma(t_1)$ along $\gamma$.

It is clear to me that the second summand must go to $0$ as $t \to 0^+$ and this would imply the result for small $t$. But I have not idea of how to prove that!

  • $\begingroup$ What is $\mathrm{inj}(x)$? Thanks $\endgroup$
    – Alex Ortiz
    Commented Mar 28, 2019 at 17:43
  • 1
    $\begingroup$ It is the infectivity radius at $x$. Basically I am only considering a neighborhood of $0_p \in T_pM$ where the exponential map is a diffeomorphism onto some open neighborhood of $p$. $\endgroup$
    – Bremen000
    Commented Mar 28, 2019 at 17:45
  • $\begingroup$ Yes, as long as you assume an upper sectional curvature bound. This essentially is the content of Rauch Comparison Theorem. $\endgroup$ Commented Mar 28, 2019 at 21:24
  • $\begingroup$ I have never heard about this result, could you please expand your comment? $\endgroup$
    – Bremen000
    Commented Mar 28, 2019 at 22:06
  • $\begingroup$ Did you ever try to read do Carmo's "Riemannian Geometry"? He covers Rauch's theorem and his book is my favorite introduction to R.G. $\endgroup$ Commented Mar 29, 2019 at 23:09


You must log in to answer this question.

Browse other questions tagged .