3

I have the following question:

Assume that $α$ is a regular curve in $R^2$ and all the normal lines of the curve pass though the origin. Prove that $α$ is contained in a circle around the origin. (Recall the normal line at $α(t)$ is the line through $α(t)$ pointing in the direction of the normal vector $N(t)$.)

My attempt:

I know that if $α(t)$ is contained in a circle, then the curvature is constant. So I have to prove that the curvature at $α(t)$ is constant. I also know that the equations of the normal lines through the origin that passes through arbitrary points on the curve $α(t)$ denoted as $(t, α(t))$ have the equations:

$$α(t) = \frac{t}{α'(t)}$$

I'm confused on how to proceed from here.

Varun Iyer
  • 6,074

5 Answers5

5

More than a year has passed. I hope my solution still helps.

I suppose you know the Frenet frame and the Frenet-Serret formulas. Let me denote the unit tangent vector and unit normal vector at $\alpha(t)$ by $\mathbf{T}(t)$ and $\mathbf{N}(t)$. Finally, denote $\mathbf{B}(t)=\mathbf{T}(t)\times\mathbf{N}(t)$.

My strategy is to prove that

  1. $||\alpha||$ is constant, so that (the trace of) $\alpha$ is contained in some sphere centered around $\mathbf{0}=(0,0,0)$.
  2. $\mathbf{B}(t)$ is constant, so that the curve is a plane curve, i.e., it lies on some plane.

For 1, we differentiate $||\alpha|| ^2$ with respect to $t$: \begin{align*} \frac{d||\alpha(t)||^2}{dt}=\frac{d}{dt}\alpha(t)\cdot\alpha(t)=2\alpha(t)\cdot\alpha'(t) \end{align*} Since the normal line at $\alpha(t)$ passes through $\alpha(t)$ and $\mathbf{0}$, thus the vector $\alpha(t)-\mathbf{0}=\alpha(t)$ is parallel to the normal vector $\mathbf{N}(t)$, and so it is orthogonal to the tangent vector $\mathbf{T}(t)=\alpha'(t)$. Therefore $\alpha(t)\cdot\alpha'(t)=0$. Hence, $||\alpha||^2$ is constant, and so is $||\alpha||$.

For 2, we differentiate $\mathbf{B}(t)$: \begin{align*} \frac{d\mathbf{B}(t)}{dt}&=\mathbf{T}'(t)\times\mathbf{N}(t)+\mathbf{T}(t)\times\mathbf{N}'(t) \\ &=\alpha''(t)\times\frac{\alpha''(t)}{|\alpha''(t)|}+\alpha'(t)\times(\lambda\alpha'(t)) \\ &=0+\lambda(\alpha'(t)\times\alpha'(t)) \\ &=0 \end{align*} Here I write $\mathbf{N}(t)=\lambda\alpha(t)$ because $\alpha(t)$ is parallel to $\mathbf{N}(t)$ as aforementioned. This proves 2.

Since the intersection of a sphere and a plane is precisely a circle, we conclude that (the trace of) $\alpha$ is contained in a circle. Moreover, since the normal lines are contained in the plane, thus the centre of the circle is the origin. $\qquad\square$

Tengu
  • 4,072
  • The factorization with $\lambda$ seems off? Also, can you elaborate why "since the normal lines are contained in the plane, thus the centre of the circle is the origin". Thanks for the great answer! – IntegrateThis Aug 15 '21 at 07:09
  • 1
    @IntegrateThis Sorry I'm not sure about the 1st question. I am essentially just using the basic fact that the cross-product is bilinear over $\mathbb{R}$. For the 2nd question, having a second thought about it I think my claim here is wrong. The correct reasoning should be "since the normal vector $\mathbf{N}(t)$ is parallel to the vector $\alpha(t)$, thus the centre is the origin". Thanks for the question. – Hopf eccentric Aug 21 '21 at 12:11
  • Right, I was mistaken with the factorization of $\lambda$ comment. – IntegrateThis Aug 25 '21 at 05:08
4

We have $\alpha'(t) \perp N(t) \parallel \alpha(t)$ for all $t$, so $\alpha' \cdot \alpha = 0$,

therefore

$$\frac{d\|\alpha\|}{dt} = \frac{\alpha\cdot\alpha'}{\|\alpha\|} = 0,$$

so $\|\alpha\|$ is constant. $~~~~\square$

Anon
  • 5,489
2

If all normal lines pass through a fixed point, then for all $s \in I$, there is a $\lambda = \lambda(s)$ such that

$$\alpha(s) + \lambda(s)n(s) = p,$$ where $p$ is our supposed fixed point. So, deriving we are left with:

$$\alpha'(s) + \lambda'(s)n(s) + \lambda(s)(-k(s)t(s) - \tau(s)b(s)) = 0$$ $$t + \lambda'n - \lambda k t - \lambda \tau b = 0 \implies (1-\lambda k)t + \ldots$$

But $\{t,n,b\}$ form an orthogonal basis, and so it is true that $(1 - \lambda k) = 0, \lambda' = 0$ and $\lambda \tau = 0$.

So $\lambda' = 0 \implies \lambda = \text{constant}$. Also, $k = \frac{1}{\lambda}$, $\tau = 0$.

See the curvature is constant and equal to $\frac{1}{\lambda}$, which is the curvature for a circle when $\lambda = R$, $R$ being the circle's radius. And so we can contain $\alpha(s)$ in a circle centered on the origin - it need only have radius $R > \lambda$.

ViktorStein
  • 4,838
DrHAL
  • 866
  • the curve is in $\mathbb{R}^2$ so we need to prove that $\lambda$ is positive too, right? – R_Squared Oct 12 '23 at 13:11
  • Actually, planar curves having constant curvature are well-characterized: they are either a straight line or a circle (or their segments). Then since the curvature can't be zero in this case, we must be in the case of a circle. We can assume $\lambda$ will be positive as this is linked only to a choice of orientation. – DrHAL Jan 09 '24 at 11:10
0

Hint:

Assume WLOG that $\alpha$ is parametrized by arc-length. The normal line at the point $\alpha(s)$ is given by:

$$\alpha(s)+\lambda n(s),$$

where $n$ is the unit normal.

By your hypothesis, for each $s$ there exists a $\lambda(s)$ such that

$$\alpha(s)+\lambda(s)n(s)=0.$$

Hence

$$\|\alpha(s)\|=|\lambda(s)|,$$

so it suffices to see that $\lambda$ is constant.

Reveillark
  • 13,044
  • how does this prove that the curvature is constant? Is lambda the curvature? – Varun Iyer Sep 06 '16 at 01:11
  • @VarunIyer It's not a full proof, I gave you a hint. $\lambda$ is the parameter of the normal line at a point. $\lambda(s)$ is the value of this parameter at which the line meets the origin. – Reveillark Sep 06 '16 at 01:17
  • I understand that, how do you prove that $||\alpha(s)||=|\lambda(s)|$? I don't understand how you reached this statement from the statment above it. – Varun Iyer Sep 06 '16 at 01:29
  • $|\alpha(s)|=|-\lambda(s)n(s)|$ and $n$ is unitary. – Reveillark Sep 06 '16 at 01:31
  • Ok how does that prove that $\lambda(s)$ is constant? And I'm confused how to relate this to the curvature. – Varun Iyer Sep 06 '16 at 01:33
  • It doesn't. There are other ways of seeing that $\alpha$ is contained in a circle, besides checking the curvature is constant. Try differentiating the different symbols involved. For instance: $\frac{d}{ds}|\alpha|^2=2\langle \alpha,\alpha'\rangle$. – Reveillark Sep 06 '16 at 01:44
  • so as an alternative, could i assume that a(t) parameterizes to a circle and prove the curvature is constant? – Varun Iyer Sep 06 '16 at 01:48
  • A curve $\alpha$ is contained in a circle if and only if its norm $|\alpha|$ is constant (try drawing a picture). – Reveillark Sep 06 '16 at 01:49
  • I'm still not able to see that. Do you have a link where i could understand this more? I'm not that proficient with differential geo. – Varun Iyer Sep 06 '16 at 01:51
0

Let $\alpha=\beta \mathbb{n}$ where $\beta$ is a function of $s$. Now by definition $$\mathbb{t}=\frac{d\alpha}{ds}=\frac{d\beta}{ds}\mathbb{n}-\beta\kappa\mathbb{t}$$

This implies that $\frac{d\beta}{ds}=0$ and that $\kappa=-\frac{1}{\beta}$