ON THE GIBBS PHENOMENON .1. RECOVERING EXPONENTIAL ACCURACY FROM THE FOURIER PARTIAL SUM OF A NONPERIODIC ANALYTIC-FUNCTION

It is well known that the Fourier series of an analytic and periodic function, truncated after 2N + 1 terms, converges exponentially with N, even in the maximum norm. It is also known that if the function is not periodic, the rate of convergence deteriorates; in particular, there is no convergence in the maximum norm, although the function is still analytic. This is known as the Gibbs phenomenon. In this paper we show that the first 2N + 1 Fourier coefficients contain enough information about the function, so that an exponentially convergent approximation (in the maximum norm) can be constructed. The proof is a constructive one and makes use of the Gegenbauer polynomials C(n)lambda(x). It consists of two steps. In the first step we show that the first m coefficients of the Gegenbauer expansion (based on C(n)lambda(x), for 0 less-than-or-equal-to n less-than-or-equal-to m) of any L2 function can be obtained, within exponential accuracy, provided that both A and m are proportional to (but smaller than) N. In the second step we construct the Gegenbauer expansion based on C(n)lambda, 0 less-than-or-equal-to n less-than-or-equal-to m, from the coefficients found in the first step. We show that this series converges exponentially with N, provided that the original function is analytic (though nonperiodic). Thus we prove that the Gibbs phenomenon can be completely overcome.