Harmonic analysis breaks functions along their hidden symmetries
Harmonic analysis grew out of studying harmonic functions and how they behave at a boundary. It splits functions and measures into pieces by symmetry, scale, or frequency, then proves estimates for the operators those splittings produce, reaching past Fourier series into settings where orthogonality alone falls short.
Fourier series and the Fourier transform are its basic examples, but today’s specialists also work with Fourier multipliers, spectral decompositions, and oscillatory and singular integrals. A separate abstract tradition deals with topological groups, where the Peter–Weyl theorem, Pontryagin duality, and results of Plancherel type are central, and the subject reaches into number theory, ergodic theory, and representation theory.
One root is classical potential theory. The Poisson integral formula rebuilds a harmonic function inside a disk or half-space from its values on the edge, but whether those integrals converge and whether boundary values exist proved subtler than Fourier methods could settle. Such questions led to maximal estimates and singular integral operators, with the Hilbert transform, tied to conjugate harmonic functions, as the prototype; in higher dimensions the Riesz transforms play that role.
The Hardy–Littlewood maximal function controls pointwise convergence, differentiation of integrals, and limits at the boundary, and it set the pattern for weak-type inequalities, interpolation, and weighted estimates. Calderón–Zygmund theory states when singular integrals are bounded on Lp spaces. Its signature trick splits an integrable function into a bounded good part and bad pieces confined to small cubes, each averaging to zero, so a function is cut up locally by size before being broken into frequencies. Littlewood–Paley theory sorts functions by scale and swaps strict orthogonality for almost-orthogonality.
Restriction problems show how far the field has moved from classical Fourier analysis. They ask whether a transform still makes sense when looked at only on a curved surface of lower dimension, like a paraboloid, a cone, or a sphere. For integrable functions the transform is continuous, so restriction is harmless; for larger classes it becomes a hard estimate, studied equivalently through an extension operator that superposes plane waves with frequencies on the surface.
Source: Harmonic analysis
