Slope in polar coordinates connected to rotation matrix

Here is how you can the tangent line slope, given a function in polar coordinates:

Let's see it from a matrix algebra point of view:

\begin{align*} \begin{pmatrix} dx/d\theta \\ dy/d\theta \\ \end{pmatrix} &= \begin{pmatrix} \cos \theta & -\sin \theta \\ \sin \theta & \cos \theta \\ \end{pmatrix} \begin{pmatrix} dr/d\theta \\ r\\ \end{pmatrix} \\ &= R_{\theta} \begin{pmatrix} dr/d\theta \\ r\\ \end{pmatrix} \end{align*}or, reversely:

\begin{align*} \begin{pmatrix} dr/d\theta \\ r\\ \end{pmatrix} &= R_{-\theta} \begin{pmatrix} dx/d\theta \\ dy/d\theta \\ \end{pmatrix} \end{align*}The matrix on the left looks familiar, huh? It's just a rotation matrix!

$ \begin{pmatrix} dr/d\theta \\ r\\ \end{pmatrix}$ deserves a closer look on its own right. $r$ is the distance from the origin, $dr/d\theta$ is the rate of deviation from the origin as $\theta$ varies. Take the circle $r = 1$ as an example, $r$ is constant here, so $dr/d\theta$ is always zero, and $ \begin{pmatrix} dr/d\theta \\ r\\ \end{pmatrix}$ is a constant vector:

ezplot('x^2 + y^2 - 1', [-1.2 1.2 -1.2 2.2]) axis equal hold on x1 = -1:.2:1 x2 = x1(2:(end-1)) x = horzcat(x1, x2) y = horzcat(sqrt(1- x1.^2), -sqrt(1 - x2.^2)) scatter(x, y) u = repmat(0, size(x)) v = repmat(1, size(x)) quiver(x, y, u, v)

Take the parabola $y = x^2$ as another example. Its polar form is $r = \tan\theta \sec\theta$.

\begin{align*} dr/d\theta &= (1 + \sin^2 \theta) / \cos^3 \theta \\ &= (r^3 + ry^2) / x^3 \\ \begin{pmatrix} dr/d\theta \\ r \\ \end{pmatrix} &= \begin{pmatrix} (r^3 + ry^2) / x^3 \\ r \\ \end{pmatrix} \end{align*}where $r = \sqrt{x^2 + y^2}$. Both $dr/d\theta$ and $r$ goes to infinity when $\theta$ goes to $\pi / 2$, the ratio $\frac{dr/d\theta}{r}$ also goes to infinity, which means the vector \begin{pmatrix} dr/d\theta \\ r \\ \end{pmatrix} becomes more and more flat.

which in a vector field likes like this:

x = -2:.01:2 y = x.^2 plot(x, y) hold on xvec = -1.8:0.2:1.8 yvec = xvec.^2 scatter(xvec, yvec) vvec = sqrt(xvec.^2 + yvec.^2) uvec = (vvec.^3 + vvec .* yvec.^2) ./ xvec.^3 quiver(xvec, yvec, uvec, vvec)

A quick look confirms that when you rotate \begin{pmatrix} dr/d\theta \\ r \\ \end{pmatrix} by $\theta$, you get the tagent line at $\theta$, and as $\theta$ increases, the line gets more and more vertical, which is indeed true: $y' = 2x$.

## 0 comments:

Post a Comment