A linear operator that maps a function or distribution f to the convolution k * f with a fixed kernel or distribution k, often realized as an integral operator with a translation-invariant kernel; central in harmonic analysis, PDEs, and signal processing.

Principle

Convolution operators implement translation-invariant linear responses: they commute with translations and are diagonalized by the Fourier transform, so their action is multiplication by the Fourier transform of the kernel.

Demonstration

In signal processing an LTI system with impulse response h yields output y = h * x; in PDEs, the solution operator for constant-coefficient linear equations on R^n is given by convolution with the fundamental solution or Green's kernel when available.

Misapplication

Assuming that convolution with an arbitrary kernel is bounded on all function spaces (e.g., on every L^p) or that convolution always regularizes; kernels that are distributions or non-integrable may not define bounded operators or may require careful domain specification.

Consequence

When the kernel satisfies appropriate integrability or multiplier conditions, the convolution operator provides smoothing, filtering, frequency-selective multiplication, and a framework for pseudodifferential calculus via symbol calculus in Fourier space.

Reversal

The reversal is treating pointwise multiplication in physical space as a convolution operator; under Fourier transform these roles swap, so confusing them loses the duality between time/space localization and frequency multiplication.

Boundary

Defined when the kernel is an L^1 function, a tempered distribution with controlled growth, or when multiplier conditions are satisfied; excludes nontranslation-invariant integral operators and nonlinear operators not reducible to convolution.

Semantic Tension

Tension exists between convolution operators (translation invariant) and more general integral operators with nontranslation-invariant kernels; practitioners sometimes call any integral transform a convolution, blurring invariance properties.

Synthesis

Convolution operators encapsulate translation-invariant linear actions via a fixed kernel, are diagonal in Fourier variables, and serve as the bridge between time/space domain operations (filtering, impulse response) and frequency-domain multipliers in analysis and applications.