Let $C_{1}$ and $C_{2}$ be two circles, respectively with radius $1$ and $2$ endowed with the riemannian metric induced by the euclidean metric. I need to show that they are not isometric. This seems simple but I cannot get a proof.

Here is what I have tried. I work with a generic circle $C_{r}$ of radius $r>0$. First, we have the local coordinate $\theta$ on the circle such that, locally:

$$\phi(\theta)=(r\cos(\theta),r\sin(\theta)),\quad \theta\in[0,2\pi)$$

Denote by $g_{0}=(dx)^{2}+(dy)^{2}$ the euclidean metric on $\mathbb{R}^{2}$. The induced metric is given by


If $\gamma:I\to M$ is a differentiable curve (in the sequel, a curve) from an interval $I\subset\mathbb{R}$ to a riemannian manifold $M,g$, the length of $\gamma$ is defined as


The distance between two points $x,y\in M$ relatively to the metric $g$ is defined as

$$d_{g}(x,y) = \inf\{\ell(\gamma)\mid\gamma:[0,1]\to M,\gamma\text{ is a curve},\gamma(0)=x,\gamma(1)=y\}$$

It is easy to see that if

$$i:M,g\to N,h$$

is an isometry (i.e. a diffeomorphism such that $i^{\ast}h=g$), then for any curve $\gamma:I\to M$ we have


Hence, we have

$$d_{g}(x,y) = d_{h}(i(x),i(y))$$

Therefore, let

$$\gamma:[0,1]\to C_{r}:t\mapsto \gamma(t) = (r\cos(s(t)),r\sin(s(t)))$$

A quick computation shows that

$$\ell(\gamma)=r\int_{0}^{1}\vert \dot s(t)\vert_{1} dt$$

where $\vert v\vert_{k}$ denotes the euclidean norm of $v\in\mathbb{R}^{k}$.

As we want to prove that $C^{1}$ and $C^{2}$ are not isometric, we need to show that there exists no isometry $i:C^{1}\to C^{2}$. If we suppose there exists such an isometry, we must have

$$\ell(\gamma)=\ell(i\circ \gamma)$$

for any curve $\gamma:[0,1]\to C_{1}$.

But I am stuck. How to conclude from this?


The point is that the lenght of a curve does not depends on the parametrization. So $i \circ \gamma_1$ should be a reparametrization of $\gamma_2$, in particular they have the same lenght. So if $C_1, C_2$ are isometric this would implies as @Henning Makholm said that $2\pi = 4\pi$ which is absurd.

  • $\begingroup$ Thanks. To be sure to understand it correctly : what you say is that there is a curve, say, $\gamma_{1}(t)=(\cos(t),\sin(t))$ which is bijective from $[0,2\pi)$ to the small circle and that $(i\circ\gamma_{1})(t)$ is also bijective from $[0,2\pi)$ to the big circle. As $(i\circ\gamma_{1})$ is a reparametrization of $t\mapsto (r\cos(t),r\sin(t))$, they have the same length $2\pi r$ and as length is invariant under isometries, we should have $2\pi = 2\pi r$. $\endgroup$ – MoebiusCorzer Jun 16 '17 at 14:57
  • $\begingroup$ Exactly. Notice that the square root is very important for being invariant under parametrization, so if you take $e(\gamma) := \int g(\overset{\cdot}{\gamma},\overset{\cdot}{\gamma})dt$ (the "energy") you won't obtain something independant of the parametrization. (But it is still useful to physicists since it contains some "dynamical" informations.) $\endgroup$ – user171326 Jun 16 '17 at 15:04
  • $\begingroup$ There still is something I am not sure about. In order to have a reparametrization, one needs $\gamma_{1}^{-1}\circ i^{-1}\circ\gamma_{2}:[0,2\pi)\to [0,2\pi)$ to be of positive derivative if I am not mistaken. This is ensured by the fact that they are all $C^{1}$-diffeomorphisms? $\endgroup$ – MoebiusCorzer Jun 16 '17 at 15:24
  • 1
    $\begingroup$ You are right, but reversing the direction doesn't affect the lenght, so it doesn't matter if the derivative is negative (just change it by e.g $\tilde \gamma_1(t) := \gamma_1(2\pi - t)$.) $\endgroup$ – user171326 Jun 16 '17 at 15:29
  • 1
    $\begingroup$ You're welcome, glad I could help ! $\endgroup$ – user171326 Jun 16 '17 at 21:20

My approach would be to prove that the metric induces a measure on the manifold which is preserved by isometries.

Then the two circles have total measure $2\pi$ and $4\pi$, respectively, and therefore cannot be isometric.

  • $\begingroup$ Thank you. That's an interesting approach, but I am beginning with riemannian geometry and this exercise is given to "beginners". Introducing a (useful) measure on a riemannian manifold can be done naturally, I guess, by using Riesz representation theorem for Borel measures on locally compact spaces and by mimicking the construction of the Lebesgue measure. However, all this seems overly complicated for this problem. $\endgroup$ – MoebiusCorzer Jun 16 '17 at 14:25

Hint One can save some effort and conceptual difficulty here by specializing the general parameterized curve $\gamma$ to a curve that suitably respects the Riemannian structure, i.e., a unit-speed geodesic. Since the isometry $i$ preserves (parameterized) geodesics, $i \circ \gamma$ is also unit-speed geodesic.

Computing $\gamma$ explicitly using the usual embeddings into $\Bbb R^2$ shows that our unit-speed geodesic $\gamma$ satisfies $\gamma(0) = \gamma(2 \pi)$. Now, apply $i$ to both sides.


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.