Fredholm alternative

For a square linear system A x = b you learned a clean either/or: either A is invertible and there is exactly one solution for every b, or A is singular and then b must lie in a special subspace for any solution to exist, in which case there are infinitely many. The Fredholm alternative is this same dichotomy lifted to differential and integral operators — the structural law of when a linear boundary-value problem can be solved.

Stated for a boundary-value problem L u = f: exactly one of two cases holds. Either the homogeneous problem L u = 0 has only the trivial solution, and then L u = f has a unique solution for every source f (and a genuine Green's function exists). Or the homogeneous problem has nontrivial solutions, and then L u = f is solvable only for sources f satisfying the solvability condition — orthogonality to every solution of the homogeneous adjoint problem L* v = 0 — and when solvable the solution is non-unique, determined only up to adding homogeneous solutions. The number of independent homogeneous solutions of L equals the number for L*, so the dimensions match exactly as in finite linear algebra.

This is the organizing principle behind everything else in this field. It tells you in advance whether a Green's function exists, when resonance will defeat a steady solution, and what compatibility a source must satisfy. It holds for compact integral operators and for the elliptic boundary-value problems of physics, which is why it underlies the rigorous theory of the Laplace, Helmholtz, and steady-state equations. The honest caveat: the clean alternative requires the operator to be of the right type (Fredholm, with finite-dimensional kernel) — it can fail for operators with continuous spectrum, where existence is subtler.