Fourier transform

History

French mathematician Joseph Fourier proposed this idea in 1807 while studying how heat propagates in solids. He realized that any temperature distribution could be represented as a sum of simple waves. Later, the method began to be applied everywhere — from music to quantum physics.

How it works

Imagine that from a footprint in the sand you need to guess which waves left it. The Fourier transform acts like a detective: it alternately tries on waves of different lengths and strengths to the footprint. Where the match is good, a big 'clue' appears. In a computer, this is done by multiplying the signal by test sinusoids and summing the results. The result is a spectrum — a graph showing which frequencies are present.

History

French mathematician Joseph Fourier proposed this idea in 1807 while studying how heat propagates in solids. He realized that any temperature distribution could be represented as a sum of simple waves. Later, the method began to be applied everywhere — from music to quantum physics.

How it works

Imagine that from a footprint in the sand you need to guess which waves left it. The Fourier transform acts like a detective: it alternately tries on waves of different lengths and strengths to the footprint. Where the match is good, a big 'clue' appears. In a computer, this is done by multiplying the signal by test sinusoids and summing the results. The result is a spectrum — a graph showing which frequencies are present.

History

Jean-Baptiste Joseph Fourier in 1807 (published in 1822) in a memoir on heat propagation showed that an arbitrary function can be expanded into a trigonometric series. The idea met resistance from Lagrange and Laplace, who doubted the convergence of the series. Later, Dirichlet (1829) clarified the conditions. In the 20th century, Laurent Schwartz extended the transform to generalized functions (Dirac delta function). Today it is the basis of harmonic analysis.

How it works

Mathematically, the kernel e^{-iωt} acts as a 'probe': for each frequency ω, we compute the correlation of the signal with a harmonic oscillation of that frequency. The integral measures the 'similarity'. Thanks to the orthogonality of sinusoids, the decomposition is invertible without loss. The Fast Fourier Transform uses the symmetry of complex roots of unity, reducing the computation of a large matrix to a series of small ones. In quantum mechanics, the wave function in momentum representation is the Fourier transform of the coordinate wave function, and the relation Δx Δp ≥ ħ/2 follows directly from the properties of the FT.

Open problems & fun facts

Not all functions have a Fourier transform in the classical sense — a generalization to distributions is required.; Gibbs phenomenon: at points of discontinuity, the Fourier series produces overshoots whose amplitude does not decrease with increasing number of terms (about 9% of the jump).; Interpretation of negative frequencies in physical applications is not always intuitive.; Non-stationary signals require windowed FT or wavelets for adequate analysis.

💡 The Fourier transform of a finite function (equal to zero outside a finite interval) is an analytic function on the entire complex plane — this is the Paley–Wiener theorem.