A collocation method for the numerical Fourier analysis of quasi-periodic functions. II: Analytical error estimates (Q986640)

This paper, the second one in a series of two consecutive papers in the same journal (for Part I, see [Discrete Contin. Dyn. Syst., Ser. B 14, No. 1, 41--74 (2010; Zbl 1237.37057)]), does a complete error analysis of the methodology for the Fourier analysis implemented in the first one. Essentially under the hypothesis of analyticity and quasi-periodicity with the vector of basic frequencies satisfying a Diophantine condition for the input function, the authors provide error bounds for two cases. The first one where the amplitudes are computed from known frequencies, and a second one, where both amplitudes and frequencies are computed by the procedure. Apart from the results of the main theorem and lemmas used to obtain the bounds for the errors of the methodology, the authors deduce and include some suitable remarks for efficient implementation that moreover assure convergence of the procedure towards the frequencies and amplitudes of the input function. In addition, since the bounds are computable, they are tested using the examples that are Fourier analyzed in the first paper.