logo

Proof of the Rotation Number Theorem 📂Geometry

Proof of the Rotation Number Theorem

Theorem 1

The winding number of a plane simple closed curve is iα=±1i_{\alpha} = \pm 1.

Explanation

It’s a brief but very intuitive and important theorem. The proof is somewhat unique.

Proof

Let’s say α(s)\alpha (s) is a curve that satisfies the condition of the theorem with length LL. 0u<vL 0 \le u < v \le L Define two points u,vu, v appearing according to the reparameterization of arc length of the curve. Here, we aim to define the two-variable function a(u,v)a (u, v) as a unit vector in the same direction as the vector whose starting point is α(u)\alpha (u) and endpoint is α(v)\alpha (v) but with size 11. Rewriting it in formulaic terms yields the following. a(u,v):=α(v)α(u)α(v)α(u) a(u,v) := {{ \alpha (v) - \alpha (u) } \over { \left\| \alpha (v) - \alpha (u) \right\| }} If u=vu=v, then because the denominator becomes 00, consider the limit when vuv \to u as the tangent of a plane curve tt. In other words, let’s assume a(u,u)=t(u)a(u,u) = t(u). Especially, a(0,L)a(0,L) should be treated as having made one full rotation (in the same sense as the left and right limits being different). a(0,L)=t(0)=t(L) a (0,L) = - t(0) = -t(L) With this definition, α\alpha is a C2C^{2} function in the following region Δ\Delta.

20210428_170008.png

Meanwhile, let’s define the two-variable function α(u,v)\alpha (u,v) defined in Δ\Delta as the magnitude of the angle formed with the horizontal axis (xx-axis) and a(u,v)a(u,v). Care must be taken not to be confused with the given curve α(s)\alpha (s), but such a definition simplifies expressions in the upcoming calculations, necessitating the reuse of α\alpha. Remember that this definition implies α(u,u)=θ(u)\alpha (u,u) = \theta (u).


Part 1. 2πiα=ACdα\displaystyle 2 \pi i_{\alpha} = \int_{\overline{AC}} d \alpha

iα=θ(L)θ(0)2π i_{\alpha} = {{ \theta (L) - \theta (0) } \over { 2 \pi }} The winding number is the integer iαi_{\alpha} that satisfies the above.

20210428_170406.png

Because it’s α(u,u)=θ(u)\alpha (u,u) = \theta (u), integrating α\alpha along dθd \theta to αdθ\displaystyle \int_{\alpha} d \theta is the same as integrating along line segment AC\overline{AC} to dαd \alpha in ACdα\displaystyle \int_{\overline{AC}} d \alpha. Thus, we obtain the following. 2πiα=θ(L)θ(0)=0Ldθdsds=αdθ=ACdα \begin{align*} 2 \pi i_{\alpha} =& \theta (L) - \theta (0) \\ =& \int_{0}^{L} {{ d \theta } \over { d s }} ds \\ =& \int_{\alpha} d \theta \\ =& \int_{\overline{AC}} d \alpha \end{align*}


Part 2. ACdα=ABdα+BCdα\displaystyle \int_{\overline{AC}} d \alpha = \int_{\overline{AB}} d \alpha + \int_{\overline{BC}} d \alpha

Green’s theorem: Given a simple plane C2C^{2} closed curve C\mathcal{C} surrounding a bounded region R\mathcal{R} moving counterclockwise and being smooth piecewise.

If the two functions P,QP,Q defined in R\mathcal{R} are differentiable in R\mathcal{R} C(Pdx+Qdy)=R(QxPy)dxdy \int_{\mathcal{C}} (Pdx + Qdy) = \iint_{\mathcal{R}} (Q_{x} - P_{y}) dx dy

Since α\alpha is a C2C^{2} function (with its second derivative continuous), 2αuv=2αvu\displaystyle {{ \partial^{2} \alpha } \over { \partial u \partial v }} = {{ \partial^{2} \alpha } \over { \partial v \partial u }} holds, and by Green’s theorem, Δdα=Δ(αudu+αvdv)=(2αvu2αuv)dudv=0dudv=0 \begin{align*} \int_{\Delta} d \alpha =& \int_{\Delta} \left( {{ \partial \alpha } \over { \partial u }} du + {{ \partial \alpha } \over { \partial v }} dv \right) \\ =& \iint_{\blacktriangle} \left( {{ \partial^{2} \alpha } \over { \partial v \partial u }} - {{ \partial^{2} \alpha } \over { \partial u \partial v }} \right) dudv \\ =& \iint_{\blacktriangle} 0 dudv \\ =& 0 \end{align*} In other words, because AC+CB+BAdα=0\displaystyle \int_{\overline{AC} + \overline{CB} + \overline{BA}} d \alpha = 0, we obtain the following. ACdα=ABdα+BCdα \int_{\overline{AC}} d \alpha = \int_{\overline{AB}} d \alpha + \int_{\overline{BC}} d \alpha


Part 3. iα=±1i_{\alpha} = \pm 1

20210428_181826.png

That the given curve rotates counterclockwise means integrating from 00 to LL while fixing u=0u=0. ABdα=BCdα=π \int_{\overline{AB}} d \alpha = \int_{\overline{BC}} d \alpha = \pi

20210428_182747.png

That the given curve rotates clockwise means integrating from LL to 00 while fixing v=0v=0. Therefore, ABdα=BCdα=π \int_{\overline{AB}} d \alpha = \int_{\overline{BC}} d \alpha = - \pi According to Part 2, if it’s counterclockwise, then ACdα=+2π\displaystyle \int_{\overline{AC}} d \alpha = + 2 \pi, if clockwise, then ACdα=2π\displaystyle \int_{\overline{AC}} d \alpha = - 2 \pi. Summarizing, ACdα=ABdα+BCdα=±2π \int_{\overline{AC}} d \alpha = \int_{\overline{AB}} d \alpha + \int_{\overline{BC}} d \alpha = \pm 2 \pi and by Part 1, we obtain iα=±1i_{\alpha} = \pm 1.


  1. Millman. (1977). Elements of Differential Geometry: p56. ↩︎