Fourier transform of polygons

talk code

Summary:

Fourier analysis is a powerful tool for the discrete geometry of polytopes. This interaction is born from a striking formula of Brion, an explicit expression of the Fourier transform of the indicator function of a polytope. First, we will see heuristics for the large scale structure of this Fourier transform, which forms a “starburst” (see below). Then we derive Brion’s formula, which describes the Fourier transform entirely from local data of the vertices. Finally, we will use Brion’s formula to derive Pick’s theorem, relating the volume of a lattice polygon to a count of lattice points inside the polygon.

Presented at:

  • UC Berkeley student harmonic analysis, Fall 2025

🔗 Link to file


Does not work on mobile! please play on desktop sorry

Theorem: Brion's formula

Let $P\subset \RR^n$ be a simple convex polyt`ope, meaning every vertex $v$ has $n$ incident $\vec{w}^v_1,\dots,\vec{w}^v_n$. Define the matrix $M_v$ with columns the edge vectors $\vec{w}^v_k$. If $\chi_P$ is the indicator function of $P$, then its Fourier transform is given by

\[\widehat{\chi_P}(\xi) = \frac{1}{i^n} \sum_{v \text{ vertex}} \frac{|\det(M_v)| e^{i \langle\xi, v\rangle}}{\prod \langle \vec{w}^v_k,\xi\rangle}\]

This formula applies whenever the denominator is nonzero, meaning $\xi$ is not perpendicular to any edge of $P$

The applet above computes the Fourier transform of a 2D polygon using Brion’s formula. Clicking on a vertex toggles whether we include it in the sum. With all vertices enabled, we retrieve the Fourier transform of the polytope

Some remarks:

For details and explanations of this phenomena, see my notes.

Sources

I learned all this from the lovely book A friendly introduction to Fourier analysis on polytopes by Sinai Robins.