Consider the Riemannian manifold $\mathbb{R}^n$ and a smooth Riemannian metric $G:\mathbb{R}^n\rightarrow\mathbb{R}^{n\times{n}}$. What is the minimum assumption on $G$ such that the manifold $\mathbb{R}^n$ is complete as a metric space with respect to the Riemannian distance determined by $G$. The Riemannian distance $d_G(x,y)$ between points $x,y\in\mathbb{R}^n$ is defined as follows: $d_G(x,y)=\inf_{\chi\in\Omega(x,y)}\int_0^1{\sqrt{\left(\frac{d\chi(s)}{ds}\right)^TG(\chi(s))\frac{d\chi(s)}{ds}}ds}$, where $\Omega(x,y)$ is the set of all the piecewise smooth paths connecting $x$ to $y$. It can be easily checked that if $G$ satisfies: $\omega_1\Vert{y}\Vert_2^2\leq y^TG(x)y\leq\omega_2\Vert{y}\Vert_2^2$ for any $x,y\in\mathbb{R}^n$ and for some positive constants $\omega_1$ and $\omega_2$, then $\mathbb{R}^n$ is a complete metric space with respect to $d_G$. Moreover, if we can find a change of coordinate such that the metric in the new coordinate satisfies the preceding assumption, again the result is true. I am just wondering if there exist weaker assumptions on $G$ to show the result or there exist counterexamples to show that these assumptions are necessary.

  • 3
    $\begingroup$ These conditions are certainly not necessary. The bounds on $G$ imply that the volume of a geodesic ball grows like $r^n$ where $r$ is the radius. But there are plenty of complete Riemannian manifolds whose geodesic balls grow at either a slower or faster rate. In fact, these conditions are extremely strong and far stronger than needed for completeness. $\endgroup$ – Deane Yang Sep 23 '10 at 0:21
  • 1
    $\begingroup$ Assuming that $G \ge c g$, where $c$ is a positive constant and $g$ is the standard flat metric, suffices. $\endgroup$ – Deane Yang Sep 23 '10 at 0:47
  • $\begingroup$ How can I show the result using this condition? Do I need to construct a Cauchy sequence on the original metric and show that it has a limit? $\endgroup$ – Majid Sep 23 '10 at 1:27
  • $\begingroup$ Same proof as before should work. I don't see why $\omega_2$ is needed. $\endgroup$ – Deane Yang Sep 23 '10 at 1:34
  • $\begingroup$ By having that lower bound, I can show that any Cauchy sequence in $\mathbb{R}^n$ with respect to $d_G$ is a Cauchy sequence with respect to $d_g$, where $g$ is Euclidean metric or the flat metric. As we know, $\mathbb{R}^n$ with respect to $d_g$ is a complete metric space. Hence, the sequence converge to a point in $\mathbb{R}^n$ with respect to $d_g$. Now, how can I show that the sequence converges to the same point with respect to $d_G$? $\endgroup$ – Majid Sep 23 '10 at 2:17

You are assuming that the identity map between the Riemannian metric and the Euclidean metric is bi-Lipschitz, but this is certainly far too strong. Here are two counter-examples:

  1. Take some self-diffeomorphism $f$ of $\mathbb{R^n}$ with derivative unbounded below and above and pull back the Euclidean metric.

  2. Take hyperbolic $n$-space, which is complete and diffeomorphic to $\mathbb{R}^n$ but not bi-Lipschitz to it.

| cite | improve this answer | |
  • $\begingroup$ Is this condition also necessary? At the moment I don't see why that sould be the case, but I cannot construct a counter-example. An intuitive approach would be: Let $p_n=(n,0,0,...)$. Choose the metric such that $G$ equals the usual Euclidian metric everywhere except $\bigcup K_n$ where $K_n$ is the ball with radius $1/n$ around $p_n$. Now choose $G$ such that $G\leq 1/n g$ somewhere in $K_n$. Then certainly $G$ does not fulfill the condition stated, but probably $R^n$ will be complete wrt this metric as well. $\endgroup$ – user34711 Jun 6 '13 at 11:01

A standard necessary and sufficient condition is that every bounded set for your metric is bounded in $\mathbb{R}^n$ with the standard Euclidean norm.

| cite | improve this answer | |
  • $\begingroup$ I'm sure this is well known to people experienced with manifolds, but do you know of a text book mention that this fact? $\endgroup$ – ttb Jun 30 '17 at 22:08

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.