Parallel Störmer-Cowell methods for high-precision orbit computations (Q1807761)

Second-order ordinary differential equations of the form \(y''=f(y)\) are often solved numerically with Runge-Kutta-Nyström methods. These are, however, designed for sequential computers. The authors present a class of high-order parallel methods which can be understood in the framework of general linear methods. At each step a block of \(k\) stage values is computed using the whole previous block. As these \(k\) values are independent of each other, they can be computed in parallel on a system with \(k\) processors. The methods resemble Störmer-Cowell methods. For \(k=4,5\), the authors construct explicit methods with stage order \(q=k\) and step-point order \(p=k+1\) and implicit methods with \(q=k+1\) and \(p=k+2\). For \(k\geq 6\) they present explicit methods with \(q=k\) and \(p=k+2\) and implicit methods with \(q=k+1\) and \(p=k+3\). Finally, they demonstrate the effectiveness of their methods on a number of classical test examples and compare with other codes like DOPRIN and PIRKN.