Numerical Integration as an Initial Value Problem (1806.01638v1)

Abstract: Numerical integration (NI) packages commonly used in scientific research are limited to returning the value of a definite integral at the upper integration limit, also commonly referred to as numerical quadrature. These quadrature algorithms are typically of a fixed accuracy and have only limited ability to adapt to the application. In this article, we will present a highly adaptive algorithm that not only can efficiently compute definite integrals encountered in physical problems but also can be applied to other problems such as indefinite integrals, integral equations and linear and non-linear eigenvalue problems. More specifically, a finite element based algorithm is presented that numerically solves first order ordinary differential equations (ODE) by propagating the solution function from a given initial value (lower integration value). The algorithm incorporates powerful techniques including, adaptive step size choice of elements, local error checking and enforces continuity of both the integral and the integrand across consecutive elements.