##  [Spectral Pollution](/spectral-pollution-0) 

 Definition

The contamination of computed spectral approximations by spurious eigenvalues or spectral values that do not converge to the true spectrum of the underlying operator, often introduced by discretization, truncation, or domain approximation.

 

 

 

 

 

 





## Principle

Principle

Arises when the discrete approximation space admits sequences of approximate eigenfunctions whose Rayleigh quotients converge to values outside the true spectrum, or when boundary/truncation artefacts create false spectral accumulation points.

 

 

 

 

 





## Demonstration

Demonstration

In a Galerkin approximation of an unbounded elliptic operator truncated to a finite domain, computed eigenvalues may appear that do not approach any eigenvalue of the full operator as mesh is refined — these are spectral pollutants introduced by the truncation and boundary treatment.

 

 

 

 

## Misapplication

Misapplication

Treating all computed isolated eigenvalues as physically meaningful without verifying convergence or stability can lead to acceptance of spurious modes and incorrect modal decompositions or stability predictions.

 

 

 

 

 





## Consequence

Consequence

Spectral pollution corrupts modal analysis, reduces reliability of reduced-order models, mispredicts resonant frequencies, and can mask true spectral features especially near continuous spectrum thresholds.

 

 

 

 

## Reversal

Reversal

A faithful spectral approximation produces computed eigenvalues and eigenfunctions that converge (in appropriate norms) to elements of the true spectrum; absence of pollution is characterized by stability of approximate spectral values under refinement and domain extension.

 

 

 

 

 





## Boundary

Boundary

Pertains to spectral approximations of operators (differential, integral, or nonselfadjoint) under discretization or domain truncation; excludes deliberate spectral regularization techniques that modify the operator spectrum by design.

 

 

 

 

 





## Semantic Tension

Semantic Tension

Overlaps conceptually with numerical artifacts like pseudospectrum effects or discretization-induced eigenvalue drift; spectral pollution specifically denotes spurious limit points of approximate spectra that have no counterpart in the true operator spectrum.

 

 

 

 

 





## Synthesis

Synthesis

Spectral pollution is the appearance of nonconvergent, spurious spectral values in numerical approximations due to discretization or truncation; avoiding it requires careful choice of approximation spaces, boundary treatments, and stability analyses to ensure computed spectra reflect the true operator.