Convolution quadrature and discretized operational calculus. I (Q1098553)

The aim of this paper is to derive quadrature rules for the approximation of convolution integrals whose kernels are singular, the numerical evaluation of expressions arising in classical operational calculus where the Laplace transform of the convolution kernel is known, and to find stable discretizations of integral equations with convolution kernels. These topics are to be the subject of part II. The basic tool which is used is the numerical approximation of the convolution integral \(f*g(x)=\int^{x}_{0}f(x-t)g(t)dt,\) \(x>0\), by convolution quadrature rules. Approximations to f*g(x) on the grid 0,h,2h,...,Nh are found by means of a discrete convolution with the values of g on the same grid. The quadrature weights are determined with the help of the Laplace transform of f and a linear multistep method. It is shown that the convolution quadrature method is convergent with the same order as the underlying multistep method. The author applies his results to the problem of approximating the inverse Laplace transform and obtains an extension of Widder's inversion formula. The paper closes with an examination of the problem of approximating expressions of the form \[ (d/dx)^ k\int^{x}_{0}f(t)g(x-t)dt,\quad x>0. \] There are no numerical results.