3
$\begingroup$

I am asked to calculate the winding number of an ellipse (it's clearly 1 but I need to calculate it)

I tried two different aproaches but none seems to work.

I would like to know why none of them work (I believe it is because these formulas only work if I have a curve parametrized by arc lenght).

Approach 1:

A valid parametrization : $\gamma=(a\cos t,b\sin t)$, with $t \in [0,2\pi], \, a,b \in \mathbb{R}$

$\dot{\gamma}(t)=(-a\sin t,b\cos t)$, with $t \in [0,2\pi], \, a,b \in \mathbb{R}$

$\ddot{\gamma}(t)=(-a\cos t,-b\sin t)$, with $t \in [0,2\pi], \, a,b \in \mathbb{R}$

$\det(\dot{\gamma}(t)|\ddot{\gamma}(t)) = \renewcommand\arraystretch{1.2}\begin{vmatrix} -a\sin t & -a\cos t \\ b\cos t & -b\sin t \end{vmatrix}=ab \sin^2 t+ab \cos^2 t=ab$

$||\dot{\gamma}(t)||^3=(\displaystyle\sqrt{(-a\sin t)^2+(b\cos t)^2})^3=(\displaystyle\sqrt{a^2\sin^2 t+b^2\cos^2 t})^3=a^3b^3$

$\kappa(t)=\displaystyle\frac{ab}{a^3b^3}=\displaystyle\frac{1}{a^2b^2}$

$\mathcal{K}_\gamma = \displaystyle\int_{0}^{2\pi} \displaystyle\frac{1}{a^2b^2} \ dt= \displaystyle\frac{2\pi}{a^2b^2}$, $\mathcal{K}_\gamma$ is the total curvature of the curve.

$i_\gamma=\displaystyle\frac{\displaystyle\frac{2\pi}{a^2b^2}}{2\pi}=\displaystyle\frac{1}{a^2b^2}$...which is not necessarily 1.

Approach 2:

Winding # = $\displaystyle\frac{1}{2\pi}\displaystyle\int_{\gamma}\displaystyle\frac{-y}{x^2+y^2}\>dx+\displaystyle\frac{x}{x^2+y^2}\>dy$

That gives us $\displaystyle\frac{1}{2\pi}\displaystyle\int_{0}^{2\pi}\left( \displaystyle\frac{-b\sin t}{a^2\cos^2 t+b^2\sin^2 t}(-a\sin t)+\displaystyle\frac{a\cos t}{a^2\cos^2 t+b^2\sin^2 t}(b\cos t) \right)\>dt$

$\displaystyle\frac{1}{2\pi}\displaystyle\int_{0}^{2\pi}\left( \displaystyle\frac{ab}{a^2\cos^2 t+b^2\sin^2 t }\right)\>dt$, which I computed and cannot be calculated.

Clearly the second approach is valid if we are dealing with a circumference of radius 1. We can generalize for the elipsee using Green's Theorem. I would also like if someone could show me this way as well.

Thank you

$\endgroup$
2
  • $\begingroup$ If you use a polar coordinate system, and parametrise the ellipse using $\theta(t)$ and $r(t)$ for $t_0 \le t \le t_1$, then the winding number is by definition $$\frac{\theta(t_1) - \theta(t_0)}{2 \pi}$$ $\endgroup$ – Glärbo Feb 20 at 11:55
  • $\begingroup$ Thank you for your help. I had indeed read that as well, but I really can't seem to know what to do in order to end up with $\theta (t)$ and $r (t)$. Maybe this is a bit obvious, but I'm really not being able to see any further. Any hints? $\endgroup$ – hugh_maths Feb 20 at 13:01
1
$\begingroup$

The last integral indeed has a closed form to evaluate you can exploit the symmetry about $\pi$ to get the integral as $$ \begin{align} \pi I &= ab \int_{0}^{\pi} \dfrac{1}{a^2 \cos^2(t) + b^2 \sin^2(t)}dt\\ &= 2ab\int_{0}^{\pi/2} \dfrac{\sec^2(t)}{a^2+b^2 \tan^2(t)} dt \end{align}$$ Now substituite $\tan(t) = u$ $$ \dfrac{\pi}{2ab} I = \int_{0}^{\infty}\dfrac{1}{a^2+b^2x^2}dx = \dfrac{\pi}{2ab} $$ thus giving $I =1$

$\endgroup$
3
  • $\begingroup$ Thank you for your kind answer. Shouldn't this be equal to 1 since this is the winding number and its been known to be 1? $\endgroup$ – hugh_maths Feb 20 at 12:57
  • $\begingroup$ it is 1 the lhs is a constant times the integral, does this answer your question @hugh_maths $\endgroup$ – Aditya Dwivedi Feb 20 at 12:58
  • $\begingroup$ oh yes, of course, I'm sorry! Oh so we should be done ! 1 is indeed the winding number! Thank you so much! Any thoughts on why my first approach is wrong (or seems to be) ? $\endgroup$ – hugh_maths Feb 20 at 13:03
1
$\begingroup$

Green's Theorem will tell you that the winding number (given by approach #2 — your total curvature in #1 is incorrect because you need an arclength integral) of the ellipse is the same as the winding number of any circle (which is easy to compute).

Let $P=-\dfrac y{x^2+y^2}$ and $Q=\dfrac x{x^2+y^2}$. Note that on $\Bbb R^2-\{0,0\}$, we have $\dfrac{\partial P}{\partial y} = \dfrac{\partial Q}{\partial x}$. Thus, if $R$ is the region between the ellipse $E$ and a circle $C$ centered at the origin lying inside it, we have $$\int_E P\,dx+Q\,dy - \int_C P\,dx + Q\,dy = \iint_R \left(\frac{\partial Q}{\partial x} - \frac{\partial P}{\partial y}\right)\,dA = 0.$$ But you can easily calculate that $\int_C P\,dx + Q\,dy = 2\pi$.

$\endgroup$
3
  • $\begingroup$ Thank you so much!@ Aditya Dwivedi helped me and I was able to evaluate the integral. As for the approach # one, I believe the total curvature divided by $2\pi$ gives us the index, which is the same as the winding number. Or they differ ? Cause I do not understand why the result comes up different than 1. $\endgroup$ – hugh_maths Feb 20 at 19:02
  • $\begingroup$ Yeah, for a convex curve $C$, it is the case that $\int_C \kappa\,ds = \pm 2\pi$ (hard theorem). But you have to integrate with respect to arclength, not $dt$. $\endgroup$ – Ted Shifrin Feb 20 at 19:25
  • $\begingroup$ Additional comment: The curvature of the ellipse is not constant, so you'd better double-check your computations! You made a glaring error there. $\endgroup$ – Ted Shifrin Feb 20 at 19:34

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.