The summation of series (Q1082771): Difference between revisions

The power series (1) \(s(x)=\mu_ 1-\mu_ 2x+\mu_ 3x^ 2-..\). can be regarded as a formal expansion in powers of x of the Stieltjes integral \((2)\quad I(x)=\int^{\infty}_{0}f(u)/(1+xu)du,\) provided we identify the \(\mu_ k\) as moments \((3)\quad \mu_ k=\int^{\infty}_{0}u^{k- 1} f(u)du\) of the function f(u). It will thus theoretically be a condition for the applicability of the methods of this paper that the \(\mu_ k\) be in fact moments, and this turns out to be the case in many series that arise in Applied Mathematics. In other cases, the coefficients in (1) may be asymptotically close to moments, tending rapidly to them as k tends to infinity, and then (1) may conveniently be summed as the sum of two series, one of which has moments as coefficients, and the other rapidly convergent. In order to be able to make use of (2) for summation of (1), we need to be able to determine f(u) from its moments \(\mu_ k\), and this essentially involves the inversion of the Mellin transform (3). The previous results by the author on the summation of power and Fourier series are extended, and the summation of other types of series is now treated. Furthermore, the summation formula is now developed in such a way as to decrease the influence of the integral contained therein, and which has to be computed by numerical quadrature. In some cases, this development can be done to such an extent that the integral itself can be neglected, and this gives rise to efficient convergence acceleration formula, in which one has a very close control on the error.