explicit error bound for modified numerical iterated integration by means of sinc methods精品.pdfVIP

Explicit Error Bound for Modified Numerical Iterated Integration by Means of Sinc Methods B Tomoaki Okayama( ) Hiroshima City University, 3-4-1, Ozuka-higashi, Asaminami-ku, Hiroshima, Japan okayama@hiroshima-cu.ac.jp Abstract. This paper reinforces numerical iterated integration devel- oped by Muhammad–Mori in the following two points: (1) the approxi- mation formula is modified so that it can achieve a better convergence rate in more general cases, and (2) an explicit error bound is given in a computable form for the modified formula. The formula works quite efficiently, especially if the integrand is of a product type. Numerical examples that confirm it are also presented. Keywords: Sinc quadrature · Sinc indefinite integration · Repeated integral · Verified numerical integration · Double-exponential transformation 1 Introduction The concern of this paper is efficient approximation of a two-dimensional iterated integral I = b q (x) f (x, y) dy dx, (1) a A with an a priori rigorous error bound. Here, q (x) is a monotone function that may have derivative singularity at the endpoints of [a, b], and the integrand f (x, y) also may have singularity on the boundary of the square region [a, b] × [A, B] (see also Figs. 1 and 2). In this case, a Cartesian product rule of a well known one-dimensional quadrature formula (such as the Gaussian formula and the Clenshaw–Curtis formula) does not work properly, or at least its mathematically- rigoro