Skip to content

Theory Overview

SpectralDiffX implements three families of pseudospectral methods. Each exploits the idea that derivatives become algebraic operations in the appropriate transform space, enabling spectral (exponential) accuracy for smooth functions.


Methods at a Glance

Method Domain Basis Functions Transform
Fourier Periodic \([0, L)\) \(e^{ikx}\) FFT
Chebyshev Non-periodic \([-1, 1]\) \(T_n(\cos\theta)\) DCT
Spherical Harmonics Sphere \(S^2\) \(Y_\ell^m(\theta,\phi)\) SHT

Spectral Accuracy

The hallmark of spectral methods is spectral (exponential) convergence for smooth functions:

\[\|u - u_N\|_\infty \sim C \, e^{-\alpha N}\]

compared to the algebraic convergence of finite difference methods (\(O(h^p)\) for order \(p\)).

This means that doubling the resolution squares the accuracy — rather than merely doubling it.


Explore the Theory