A linear operator between Banach (or normed) spaces that maps bounded sets to relatively compact sets — equivalently, the image of the unit ball has compact closure.

Principle

Compactness of a linear map in infinite dimensions generalizes finite-dimensional approximability: compact operators behave like limits of finite-rank operators and so inherit discrete spectral features absent for general bounded operators.

Demonstration

On L^2([0,1]) an integral operator (Tf)(x)=∫_0^1 K(x,y) f(y) dy with K continuous is a compact operator because it maps the unit ball to an equicontinuous, pointwise-bounded family whose closure is compact by Arzelà–Ascoli.

Misapplication

Assuming every bounded operator is compact (false in infinite dimensions) or treating compactness as preserved under arbitrary limits in weaker topologies; for example, the unilateral shift on ℓ^2 is bounded but not compact.

Consequence

Compact operators have spectral properties analogous to matrices: their nonzero spectrum is at most countable with 0 as only possible accumulation point, and eigenvectors associated with distinct nonzero eigenvalues are finite-dimensional, enabling Fredholm theory.

Reversal

A bounded operator that fails to send bounded sets to relatively compact sets (e.g., the bilateral shift or identity on infinite-dimensional space) illustrates the opposite behavior: continuous but without finite-dimensional approximability in norm.

Boundary

Defined for linear maps between normed spaces; notion excludes nonlinear maps (unless linearized) and depends on the topology induced by the norm — different topologies alter compactness; not every Banach space admits nontrivial compact operators.

Semantic Tension

Often conflated with finite-rank operators: every finite-rank operator is compact, but compact operators may be infinite-dimensional limits of finite-rank ones; the tension concerns approximability in operator norm versus algebraic rank.

Synthesis

A compact operator is a linear map whose action compresses bounded sets into relatively compact images, providing an infinite-dimensional analogue of finite-dimensional operators with discrete-like spectral structure and powerful solvability results.