On the Gibbs phenomenon. I: Recovering exponential accuracy from the Fourier partial sum of a nonperiodic analytic function (Q1195723)

Let \(f(x)\) be an analytic but nonperiodic function defined in \([-1,1]\) with discontinuity at the boundary \(x=\pm 1\). It is well-known that the traditional Fourier series, truncated after \(2N+1\) terms, shows spurious oscillations near the boundary \(x=\pm 1\), that is the Gibbs phenomenon. There is no convergence in the maximum norm. In this paper the authors show that from the first \(2N+1\) Fourier coefficients of \(f(x)\) one can get the first \(m\sim N\) coefficients in the Gegenbauer polynomials \(C^ \lambda_ n(x)\), provided that the parameter \(\lambda\) grows with the number of Fourier modes \(N\). The expansion yields exponential accuracy in the maximum norm. Some interesting numerical examples are given.