5

Define $$f_n(\mathbf{r})=\frac{1}{n}\sum_{k=1}^n\exp\left(2\pi i\binom{\cos\left(2\pi k/n\right)}{\sin\left(2\pi k/n\right)}\cdot\mathbf{r} \right)$$ as the sum of $n$ counterpropagating plane waves. Then, let $I_n^R(z)\geq0$ be the "density of states" image in $\mathbb{C}$ of the ball $\left\{\mathbf{r}\mid|\mathbf{r}|<R\right\}\subset\mathbb{R}^2$ under the map $f_n$, normalized in some "sane" way.

  • Is $I_n^\infty$ known?

$I_1^\infty$ is just a uniform-density ring of radius 1, ie

$$I_1^\infty(re^{i\theta})\propto\delta(r-1)$$

where $\delta$ is the Dirac delta.

For $I_2$, since $f_n(x,y)=\cos(2\pi x)$ and so we get a "horizontal line" of length two in $\mathbb{C}$ with density somewhat like this:

$$I_2(x+iy)\propto\frac{\delta(y)}{\sqrt{1-x^2}}$$

where $-1\leq x \leq 1$ (for $x$ outside that it's clearly zero). In general, $I_n$ for even $n$ is just going to be a "horizontal line" of length two with some density function, since $f_n$ is purely real-valued and oscillates somewhere between -1 and 1.

For odd $n$ the situation gets more interesting, and I'm not sure how to approach it. In such situations I usually resort to brute-force computation and have fun (while making no symbolic progress at all!). For example, here is a plot of $f_5$ on the ball of radius 4 (magnitude is encoded by brightness, and phase encoded by hue):

enter image description here

Here are pictures of $I_5^1,I_5^2,I_5^3$ and $I_5^4$:

enter image description here

enter image description here

enter image description here

enter image description here

As we increase the computation radius, a pattern begins to emerge:

$I_5^{10}$:

enter image description here

$I_5^{20}$:

enter image description here

$I_5^{50}$:

enter image description here

$I_5^{300}$:

enter image description here

Does anyone recognize what function this is? And are there general formulas for $I_n^\infty$ for odd $n$?

At the very least, I hope people enjoy the images! Right-click and open in a new tab to see higher resolution versions.

DumpsterDoofus
  • 1,600
  • 1
  • 9
  • 19
  • Neat question! Going by the picture of $I_5^1$, it looks like the real axis is pointing vertically downward in these plots -- is that on purpose? Also, have you checked whether the support of the plots matches a hypocycloid of $n$ cusps? –  May 13 '15 at 03:17
  • @Rahul: Yes, the real axis is pointing downwards (I probably should have rotated the images 90 degrees counterclockwise, but never got around to doing it, so apologies). I also haven't done any checking on what the support is, but I wouldn't be surprised if it turned out to be something simple. – DumpsterDoofus May 13 '15 at 03:21
  • Hypocycloids sure seem to match pretty well: http://i.stack.imgur.com/C3cPe.png –  May 13 '15 at 04:12
  • @Rahul: That's pretty cool! I'm guessing knowing the complete density function is hard, but at least knowing (or rather having a good conjecture about) the shape of the support is enough for me. Thanks. – DumpsterDoofus May 13 '15 at 15:01

0 Answers0