I am trying to self teach differential geometry and to that effect I am trying to do the Homework in the MIT open course.

The specific question I am struggling with is:

Let $c$ be a regular curve such that $|c(s)| ≤ 1$ for all $s$. Suppose

that there is a point $t$ where $|c(t)| = 1$. Prove that the curvature at that

point satisfies $|κ(t)| ≥ 1$

This is what I have so far: The curve being regular implies that it can be arc length parametrized, which means that it's curvature is just the norm of the second derivative, or in 2D $|x'y'' - x''y'|$.

The curve magnitude being upper bounded by 1 means the curve is fully contained within the unit disk.

The curve being one at $t$ means that point is ON the unit circle. Thus my intuition says this curve can only "bend inwards" at a rate equal or higher than the circle, otherwise it will be an epsilon outside of the circle at a point infinitesimally close to $t$. In other words it seems to me that a if a curve with a point on the unit circle has a smaller curvature than that of the unit circle (i.e 1) the curve will "pop out" of it for some infinitesimal amount. But I am not sure if A) this is correct B) how to formalize it.

I am looking mostly for a hint and advice, not for the full solution. Thank you lots in advance.


Let $t_0$ be the point satisfying $\| \alpha(t_0)\| = 1$. Then, at $t_0$, the function $$t \mapsto \|\alpha(t) \|^2$$

assumes a maximum. Thus, its first order derivative at $t_0$ vanishes, that is:

$$2 \langle \alpha(t_0), \alpha'(t_0)\rangle = 0$$

And by the second derivative test, its second order derivative at $t_0$ is at most $0$, that is:

$$2 \langle \alpha'(t_0), \alpha'(t_0)\rangle + 2 \langle \alpha''(t_0), \alpha'(t_0) \rangle \leq 0$$

Assuming without loss of generality that $\alpha$ has unit speed, we get:

$$1 \leq \langle \alpha''(t_0), \alpha'(t_0) \rangle $$

Taking the norm of both sides and using the Cauchy-Schwarz inequality, we have:

$$1 \leq | \langle \alpha''(t_0), \alpha'(t_0) \rangle | \leq \|\alpha''(t_0)\| \cdot \|\alpha'(t_0)\| = |\kappa(t_0)|$$

as desired.

  • $\begingroup$ Sorry to be nitpicky but the second derivative test states: $f''(x_0) < 0 \implies x_0$ is a maximum, however you are stating the converse, i.e $x_0$ is a maximum $\implies f''(x_0) \leq 0$ Is that claim true? Can you please link a source, that;s the only part I don't fully follow. $\endgroup$ – Makogan May 29 '20 at 6:51
  • $\begingroup$ You have it backwards. The second derivative test gives necessary conditions for a point to be a local maximum: namely, that the first order derivative is 0 at the point and that the second one is non positive. As for a source, I'm on mobile right now but you can see a recent question I asked on my profile here where I proved the most general case possible. $\endgroup$ – Matheus Andrade May 29 '20 at 6:54
  • $\begingroup$ I think there's an issue Imma edit your answer a little bit. $\endgroup$ – Makogan May 30 '20 at 3:35

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.