Algorithm 720: An algorithm for adaptive cubature over a collection of 3-dimensional simplices

Abstract

An adaptive algorithm for computing an approximation to the integral of each element in a vector of functions over a 3-dimensional region covered by simplices is presented. The algorithm is encoded in FORTRAN 77. Locally, a cubature formula of degree 8 with 43 points is used to approximate an integral. The local error estimate is based on the same evaluation points. The error estimation procedure tries to decide whether the approximation for each function has asymptotic behavior, and different actions are taken depending on that decision. The simplex with the largest error is subdivided into 8 simplices. The local procedure is then applied to each new region. This procedure is repeated until convergence.