Let $(M,g)$ be the (underlying) manifold, $(S,g|)$ be a submanifold. Let $a,b,c \in S$. It's not in general true that $d_M(a,b)\leq d_M(a,c) \implies d_S(a,b)\leq d_S(a,c)$.


The above seems to be always locally true if $a,b,c$ are sufficiently close to each other, but not globally. Is it really?If it isn't, could you give a counterexample?


1) But is there a sufficient condition one can put on $S$ so that $d_M(a,b)\leq d_M(a,c) \implies d_S(a,b)\leq d_S(a,c)?$

2) How about a necessary and sufficient condition?


1 Answer 1


Your condition implies that there is a nondecreasing function, $\phi_a\colon\mathbb R_{\ge0}\to \mathbb R_{\ge0}$ such that $$|a-x|_S=\phi_a(|a-x|_M).$$

One can reformulate it the following way, if you fix a point $a\in S$ then the angle between chord $[ax]_M$, $x\in S$ and the tangent space $T_xS$ depends only on the distance $|a-x|_M$.

This is quite strong global condition.

In particular if $S$ is a hypersurface then any pont is umbilical in the strongest sense ― all its principle curvatures are equal. In the higher codimensions, at each point, the absolute value of the normal curvature vector in all directions has to be the same.

  • $\begingroup$ Thanks for your answer, but I'm not sure I could agree. Your symbol $|a-x|$ is mine $d(a,x)$ etc. Please see below. I'm not assuming $d_M(a,x_1)=d_M(a,x_2) \implies d_S(a,x_1)=d_S(a,x_2)$. I'm only assuming $d_M(a,x_1)=d_M(a,x_2) \implies d_S(a,x_1)\leq d_S(a,x_2)$, so strict inequality could hold, and hence there may not exist such a map $\phi_a$, as you wrote. $\endgroup$ Apr 4, 2016 at 19:45
  • $\begingroup$ I edited the question, and added a question I. Could you give a counterexample in question I? Thanks! $\endgroup$ Apr 4, 2016 at 22:56
  • 2
    $\begingroup$ @Let'stalkmath Consider the hyperboloid $S=\{z=xy\}$ in $\mathbb R^3$ and $a=0$. Distance from $0$ is preserved along the $x$- and $y$-axis, but distance in $S$ is larger than in $\mathbb R^3$ along all other geodesics emanating from $0$. Use this to find $b$, $c\in S$ yourself. $\endgroup$ Apr 5, 2016 at 6:10
  • 1
    $\begingroup$ @SebastianGoette From the first condition it follows that norm of the restriction of differential $|d_x\mathrm{dist}_a|_{T_xS}|$ depends only on $|a-x|_M$ so the condition follows. (Sory, I did not understand what do you mean by "some power of the distance".) $\endgroup$ Apr 5, 2016 at 8:33
  • $\begingroup$ @Anton Of course, sorry ... $\endgroup$ Apr 5, 2016 at 13:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.