a generalized function

Sometimes the objects we need do not fit the ordinary idea of a function. A perfect point charge, an instantaneous hammer blow, the density of a mass concentrated at a single point — each tries to put a finite amount of something into a place of zero width, which would require an infinite value there. No honest function can do that. A generalized function is a way to make such idealized objects precise and legal, by changing what we mean by 'function' just enough to let them in.

The trick is to stop asking what value the object has at each point and instead ask only what it does to a smooth, well-behaved probe. You never weigh the point charge directly; you only ever measure its total effect against a smooth test function, by a pairing that behaves like an integral. A generalized function is defined entirely by the collection of numbers it produces when paired against every smooth probe. Ordinary functions still count: each one gives the rule 'multiply by me and integrate', and two ordinary functions give the same answers against every probe exactly when they are equal. So nothing familiar is lost, and genuinely new objects like point sources are gained.

This idea matters because the equations of physics keep producing such idealizations: a point source for the heat or Laplace equation, a sudden impulse for a vibrating string, the charge density of an electron treated as a point. Generalized functions give all of these a rigorous home and, remarkably, let us differentiate them as many times as we like. In the modern theory the precise name for a generalized function is a distribution, and the smooth probes are the test functions.