A function K(x,y) that appears under an integral sign to couple a source variable y to a field or evaluation variable x, thereby defining an integral transform or an integral operator acting on functions of y to produce functions of x.

Principle

An integral kernel encodes how values at one point influence values at another through integration; properties of K (symmetry, smoothness, separability, singularities) determine mapping, continuity, and spectral features of the associated operator.

Demonstration

The convolution kernel K(x,y)=k(x-y) defines a shift-invariant integral transform (f * k)(x)=∫ k(x-y)f(y) dy; the Green's function G(x,y) of a differential operator serves as an integral kernel mapping source distributions to solutions.

Misapplication

Treating any function of two variables as a well-behaved kernel without checking integrability, boundedness, or distributional singularities (for example assuming a singular Green's function defines a bounded operator on L2 without qualification).

Consequence

Given a suitably regular kernel, one obtains a bounded or compact linear operator whose mapping properties (smoothing, rank, spectrum) follow from kernel structure; manipulations like eigenexpansions and numerical quadrature then become meaningful.

Reversal

Pointwise multiplication by a function a(x) is not an integral kernel: it is a local operator represented by a delta-type kernel K(x,y)=a(x)δ(x-y), contrasting nonlocal integral coupling with purely local action.

Boundary

The term refers to kernels appearing in integral transforms and operators on measurable or functional spaces; it excludes unrelated uses of the word 'kernel' such as null spaces of linear maps unless explicitly described as a function-of-two-variables kernel.

Semantic Tension

Confusion often arises between 'kernel' meaning an integral kernel (function of two variables), a reproducing kernel in RKHS (with positive-definiteness constraints), and the algebraic kernel/nullspace; these notions overlap but impose different structures and assumptions.

Synthesis

An integral kernel is the two-variable rule K(x,y) that defines how an integral operator assembles contributions from y to produce values at x; its analytic and algebraic properties determine the operator's regularity, invertibility, and numerical behavior.