non-elementary integral

Every continuous function has an antiderivative — that is guaranteed by the Fundamental Theorem of Calculus. But it is a separate and surprising fact that the antiderivative cannot always be written using the familiar elementary functions: powers, roots, exponentials, logarithms, trig and inverse-trig functions, and finite combinations of them. A non-elementary integral is one whose antiderivative provably falls outside that toolkit. The standard examples are gentle-looking and famous: e^(-x^2), (sin x)/x, e^x/x, 1/ln(x), and sqrt(1 + cos^2 x).

This is not a matter of insufficient cleverness; it is a theorem. Liouville's theorem, sharpened by the Risch algorithm, gives a precise criterion and a decision procedure for whether a given elementary integrand has an elementary antiderivative. When the answer is no, it is a hard mathematical fact, not a temporary gap in our skill. The usual response is to name the new function defined by the integral — the error function erf for e^(-x^2), the sine integral Si for (sin x)/x, the exponential integral Ei, the logarithmic integral li — and then study and tabulate it exactly as the logarithm was once a newly named integral.

The single most important thing to internalize is that non-elementary is a statement about the alphabet of elementary functions, not about whether the integral has a value. The definite integral can be a perfectly definite number — sometimes a famous closed form like sqrt(pi) for the full Gaussian — and the antiderivative function can be computed, plotted, and used to any precision. These functions are the gateway from elementary calculus into the world of special functions, where the toolkit is simply enlarged to include them.