Dominated convergence makes continuous, and . Smooth compactly supported functions are dense in . Their Fourier transforms decay at infinity by integration by parts, so uniform approximation proves the Riemann-Lebesgue lemma: .
For density, smooth compactly supported functions are also uniformly dense in by cutoff and mollification. Each such target is a Schwartz function; its inverse Fourier transform is another Schwartz function, hence belongs to , and Fourier inversion recovers . Therefore the image of a function of is dense in .
The Fourier inversion theorem says that if , then agrees almost everywhere with the inverse integral of . If , this applies and gives almost everywhere. Hence the Fourier transform is injective as a map of equivalence classes, with dense image in .
Integrating the exponential gives the Fourier transform of an interval indicatorThe value at zero is the removable limit, for .
For , the convolution equals the length of the overlap of two intervals:It is continuous and compactly supported, with . Inverse Fourier transform turns this convolution into a product, soThis function is bounded near zero and at infinity, hence lies in . Fourier inversion, or the convolution theorem applied first to the integrable interval indicators, gives .
If the bounded injective Fourier transform were surjective, the open mapping theorem for Banach spaces would make its inverse bounded, yielding . But the preceding have uniform norm , whereas diverges.
For a quantitative lower bound, on we have . On each interval , . For these intervals lie below for large , soThis contradicts boundedness of the inverse. Therefore the Fourier transform has proper dense image in .
Articles by others on the same topic
There are currently no matching articles.