Fourier series (based) multiscale method for computational analysis in science and engineering: II. Composite Fourier series method for simultaneous approximation to functions and their (partial) derivatives

Abstract: Fourier series multiscale method, a concise and efficient analytical approach for multiscale computation, will be developed out of this series of papers. The second paper is concerned with simultaneous approximation to functions and their (partial) derivatives. On the basis of sufficient conditions of 2r (r is a positive integer) times term-by-term differentiation of Fourier series, a one-dimensional or two-dimensional function with general boundary conditions is decomposed into the linear combination of a corner function that describes the discontinuities at corners of the domain (only for the two-dimensional function), boundary functions that describe the discontinuities on boundaries of the domain and an internal function that describes the smoothness within the domain. It leads to the methodology of simultaneous approximation of functions and their (partial) derivatives with composite Fourier series. And specifically, for the algebraical polynomial based composite Fourier series method, the reproducing property of complete algebraical polynomials is theoretically analyzed and the approximation accuracy is validated with numerical examples. This study generalizes the Fourier series method with supplementary terms to simultaneous approximation of functions and their (partial) derivatives up to 2rth order.