5

I'm trying to learn differential geometry through one of MIT's online courses (lecture notes found here: http://ocw.mit.edu/courses/mathematics/18-950-differential-geometry-fall-2008/lecture-notes/ch1_revised.pdf) and am stuck with what should be an easy question. The question asks to show the following: Suppose $c(s)$ is a regular curve ($c'(s) \ne 0$) in the plane with $|c(s)| \le 1$ and suppose there is a point t with $|c(t)| = 1$. Then $|\kappa(t)| \ge 1$. Here, $\kappa$ is defined as

$\frac{det(c', c'')}{||c'||^3}.$

It's should be straightforward if we write $c$ as the graph of a function and rotate so that $c(t) = (0,1)$. However, I feel as if there should be a more elegant solution using only what's found in the first day's notes (found in the link). I just don't see it.

2 Answers2

3

Geometrically, you should understand that if the curve $c$ stays inside the unit circle and touches at a point, it must be at least as curved as the circle at that point.

Here's a hint: Use the fact that $t$ is a local maximum of the function $|c(s)|^2$, and put your calculus to work :)

Ted Shifrin
  • 115,160
1

Let me assume, without loss of generality, that $c$ is parametrized by arclength, so that $|c'|\equiv 1$, and let me also assume $t_0=0$ (where $c(t_0)$ has length one). Furthermore, upon reparametrizing $c$ backwards, we may assume that $\mathcal{B}=(c(0),c'(0))$ is a direct orthonormal basis of $\Bbb R^2$ (this frame is indeed orthonormal, as follows from the fact that $|c|^2$ attains a local maximum at $0$, and the the preceding hypothesis).


The curvature then takes on the form $$\kappa(0)=\det(c'(0),c''(0))=\det_{\mathcal B}(c'(0),c''(0))=-c(0)\cdot c''(0)$$ Also, the Taylor expansion of $c$ near $0$ is $c(t)=c(0)+tc'(0)+\frac12t^2c''(0)+o(t^2)$, and thus $$\begin{array}{ccl} |c(t)|^2&=&\underbrace{c(0)\cdot c(0)}_{=1}+2t\underbrace{c(0)\cdot c'(0)}_{=0}+(\underbrace{2\times\frac12 c''(0)\cdot c(0)}_{=-\kappa(0)}+\underbrace{c'(0)\cdot c'(0)}_{=1})t^2+o(t^2)\\ &=&1+(1-\kappa(0))t^2+o(t^2) \end{array}$$ This expression has to stay $\leq 1$ near $0$, which in turn forces $1-\kappa(0)\leq 0$

  • Thanks, guys. I was having trouble finding a way to get my hands on the curvature. Just using the fact that the derivative is perpendicular to the curve wasn't helping. My trouble with this and other courses heavy on analysis is that I get lost in the computational machinery. I just have no intuition for it. – user71216 May 12 '13 at 22:26
  • @user71216: If you want an alternative reference with perhaps more concrete examples, I'll toot my own horn and suggest you look at my course notes which have been used rather widely: http://math.uga.edu/~shifrin/ShifrinDiffGeo.pdf – Ted Shifrin May 12 '13 at 23:06
  • @user71216 has this been helpful to you? – Olivier Bégassat May 13 '13 at 01:33
  • Ted, I'll check out your course notes. I studied algebraic geometry so I like that you quickly introduce the twisted cubic :)

    Oliver, yes, it was very helpful. I had the picture in mind, but I could never find a way to relate it to the curvature.

    – user71216 May 13 '13 at 09:53