A numerical technique that represents the solution of a differential problem as a global expansion in basis functions (Fourier, Chebyshev, etc.) and enforces the governing equations in spectral coefficient space or at collocation points, achieving high accuracy for smooth solutions.

Principle

The organizing idea is to approximate the solution by a truncated series in globally supported basis functions so that differential operators act simply on coefficients; for smooth problems this yields exponential or spectral convergence as the number of modes increases.

Demonstration

Solve the periodic Poisson problem on an interval by expanding u(x) = Σ_k û_k e^{ikx}, transform the PDE to algebraic relations for û_k via multiplication by (−k^2) in Fourier space, and invert modewise to obtain spectral-accurate solutions when f is smooth.

Misapplication

Using spectral methods on problems with discontinuities or insufficient regularity causes severe Gibbs oscillations and loss of convergence; applying global bases on complex geometries without appropriate domain decomposition or mapping leads to poor conditioning and implementation difficulty.

Consequence

For smooth solutions on simple geometries, spectral methods deliver extremely high accuracy per degree of freedom and fast convergence; they are widely used in fluid dynamics, wave propagation, and stability computations where smoothness and periodicity or mapped domains apply.

Reversal

The opposite approach is local polynomial or piecewise bases (finite element or finite volume methods) that sacrifice global spectral accuracy for local adaptability, robustness on complex domains, and handling of discontinuities.

Boundary

Effective when the solution and coefficients are smooth and the geometry admits global bases or well-designed mappings; not suitable as-is for problems with shocks, steep gradients, or highly irregular domains without hybridization or filtering techniques.

Semantic Tension

Competes with high-order finite element and discontinuous Galerkin methods: spectral methods emphasize global bases and spectral convergence for very smooth problems, while high-order local methods trade some asymptotic accuracy for geometric flexibility and local conservation.

Synthesis

Spectral methods approximate solutions with global basis expansions so that differential operators reduce to simple algebraic operations on coefficients; when smoothness and geometry permit, they achieve spectral (often exponential) convergence, but they require careful treatment of nonsmooth features and complex domains.