Fourier Analysis - Key Properties

The mathematical properties of Fourier transforms reveal deep connections between time and frequency domains, providing powerful tools for analyzing linear systems and wave phenomena.

Convolution Theorem and Linear Response

TheoremConvolution Theorem

The Fourier transform converts convolution to multiplication:

$\mathcal{F}[f * g](k) = \mathcal{F}[f](k) \cdot \mathcal{F}[g](k)$

where the convolution is defined as:

$(f * g)(x) = \int_{-\infty}^{\infty} f(x')g(x - x')dx'$

Similarly, multiplication in real space becomes convolution in Fourier space:

$\mathcal{F}[f \cdot g](k) = \frac{1}{2\pi}[\mathcal{F}[f] * \mathcal{F}[g]](k)$

This theorem is fundamental in signal processing and physics. For a linear time-invariant system with impulse response $h(t)$ , the output for input $f(t)$ is:

$y(t) = (h * f)(t) \implies \tilde{y}(\omega) = \tilde{h}(\omega)\tilde{f}(\omega)$

The transfer function $\tilde{h}(\omega)$ completely characterizes the system's frequency response.

ExampleRC Circuit as a Low-Pass Filter

For an RC circuit with input voltage $V_{in}(t)$ and output $V_{out}(t)$ across the capacitor:

$\tau\frac{dV_{out}}{dt} + V_{out} = V_{in}, \quad \tau = RC$

Fourier transforming with $\mathcal{F}[dV/dt] = i\omega\tilde{V}$ :

$\tilde{V}_{out}(\omega) = \frac{1}{1 + i\omega\tau}\tilde{V}_{in}(\omega) = H(\omega)\tilde{V}_{in}(\omega)$

The transfer function $H(\omega) = 1/(1 + i\omega\tau)$ has magnitude:

$|H(\omega)| = \frac{1}{\sqrt{1 + \omega^2\tau^2}}$

which falls off for $\omega > 1/\tau$ , making this a low-pass filter with cutoff frequency $\omega_c = 1/\tau$ .

Uncertainty Relations

TheoremFourier Uncertainty Principle

For a function $f(x)$ with Fourier transform $\tilde{f}(k)$ , define the variances:

$(\Delta x)^2 = \frac{\int x^2|f(x)|^2dx}{\int|f(x)|^2dx} - \left(\frac{\int x|f(x)|^2dx}{\int|f(x)|^2dx}\right)^2$

$(\Delta k)^2 = \frac{\int k^2|\tilde{f}(k)|^2dk}{\int|\tilde{f}(k)|^2dk} - \left(\frac{\int k|\tilde{f}(k)|^2dk}{\int|\tilde{f}(k)|^2dk}\right)^2$

Then the uncertainty relation holds:

$\Delta x \cdot \Delta k \geq \frac{1}{2}$

with equality for Gaussian wave packets.

In quantum mechanics, setting $k = p/\hbar$ gives Heisenberg's uncertainty principle:

$\Delta x \cdot \Delta p \geq \frac{\hbar}{2}$

This is not a statement about measurement but a fundamental property of wave functions: narrow localization in position requires broad spread in momentum, and vice versa.

Sampling Theorem

TheoremNyquist-Shannon Sampling Theorem

If a function $f(t)$ has Fourier transform $\tilde{f}(\omega)$ that vanishes for $|\omega| > \omega_{\max}$ (band-limited), then $f(t)$ is completely determined by its samples at intervals $\Delta t = \pi/\omega_{\max}$ :

$f(t) = \sum_{n=-\infty}^{\infty} f(n\Delta t)\,\text{sinc}\left(\frac{t - n\Delta t}{\Delta t}\right)$

where $\text{sinc}(x) = \sin(\pi x)/(\pi x)$ . The Nyquist frequency is $\omega_N = \pi/\Delta t = 2\omega_{\max}$ .

This theorem is crucial for digital signal processing: signals must be sampled at least twice per period of the highest frequency component to avoid aliasing.

ExampleAliasing in Digital Audio

Human hearing ranges to about 20 kHz. CD audio samples at 44.1 kHz, giving a Nyquist frequency of 22.05 kHz. Frequencies above this fold back into the audible range (aliasing), so an anti-aliasing filter removes frequencies above 20 kHz before sampling.

Multi-dimensional Fourier Transforms

DefinitionFourier Transform in $\mathbb{R}^n$

For a function $f(\mathbf{x})$ on $\mathbb{R}^n$ :

$\tilde{f}(\mathbf{k}) = \int_{\mathbb{R}^n} f(\mathbf{x})e^{-i\mathbf{k} \cdot \mathbf{x}}d^nx$

$f(\mathbf{x}) = \frac{1}{(2\pi)^n}\int_{\mathbb{R}^n} \tilde{f}(\mathbf{k})e^{i\mathbf{k} \cdot \mathbf{x}}d^nk$

The Laplacian becomes multiplication: $\mathcal{F}[\nabla^2 f](\mathbf{k}) = -k^2\tilde{f}(\mathbf{k})$ where $k = |\mathbf{k}|$ .

ExampleGreen's Function for Poisson Equation

To solve $\nabla^2\phi(\mathbf{x}) = -\rho(\mathbf{x})/\epsilon_0$ , Fourier transform both sides:

$-k^2\tilde{\phi}(\mathbf{k}) = -\tilde{\rho}(\mathbf{k})/\epsilon_0$

Thus $\tilde{\phi}(\mathbf{k}) = \tilde{\rho}(\mathbf{k})/(k^2\epsilon_0)$ . For a point charge $\rho(\mathbf{x}) = q\delta^3(\mathbf{x})$ , $\tilde{\rho}(\mathbf{k}) = q$ :

$\phi(\mathbf{x}) = \frac{q}{\epsilon_0(2\pi)^3}\int \frac{e^{i\mathbf{k} \cdot \mathbf{x}}}{k^2}d^3k = \frac{q}{4\pi\epsilon_0|\mathbf{x}|}$

recovering Coulomb's law.

RemarkFast Fourier Transform (FFT)

The discrete Fourier transform of $N$ points naively requires $O(N^2)$ operations. The FFT algorithm reduces this to $O(N\log N)$ by exploiting symmetries, revolutionizing digital signal processing, crystallography, and numerical PDE solvers.

These properties make Fourier analysis an indispensable tool for understanding oscillatory and wave phenomena across all areas of physics.