An FFT-based algorithm for \(2\)D power series expansions (Q1963035): Difference between revisions

An algorithm for computing the power series coefficients \(f_{m,n}\) of a bivariate analytic function \(f(z_1,z_2)\) is pesented. Let \(g(t_1, t_2)\) with \(g(t_1,t_2)= 0\) for \(t_1< 0\) or \(t_2<0\) be the inverse Laplace transform of \[ G(s_1, s_2)= {1- e^{-s_1}\over s^2_1} {1- e^{-s_2}\over s^2_2} f(e^{-s_1}, e^{-s_2}). \] Computing the values \(g(m,n)\) for positive integers \(m\), \(n\) by generalized Riemann zeta function evaluation and two-dimensional fast Fourier transform, then \[ f_{m,n}= g(m+ 1,n+1)- g(m+ 1,n)- g(m,n+ 1)+ g(m,n). \] This approach is an extension of an algorithm for univariate power series which was proposed by \textit{J. Abata} and \textit{H. Dubner} [SIAM J. Numer. Anal. 5, 102-112 (1968; Zbl 0155.22101)].