Parallel numeric solution of differential equations
Differential equations have been studied for over 300 years. Partial differential equations were first used by the Swiss mathematician and lawyer Nicolaus Bernoulli in the 18th century. Second-order partial differential equations are used to model a wide range of phenomena in science, engineering, and mathematics, such as the propagation of light and sound waves, the motion of fluids, and the diffusion of heat. The thesis deals with the parallel numerical solution of partial differential equations. Second-order partial differential equations are transformed into large systems of ordinary differential equations using the method of lines. The spatial derivatives in the partial differential equation are replaced by various types of finite differences. The resulting large systems of ordinary differential equations (initial value problem) are solved in parallel using Runge-Kutta methods and the newly proposed higher-order method based on Taylor series. The numerical experiments of the selected problems are calculated using a supercomputer with different numbers of compute nodes. The results show that the Taylor-series-based numerical method significantly over-performs state-of-the-art Runge-Kutta methods.
partial differential equations, ordinary differential equations, initial value problems, Taylor series, Runge-Kutta, method of lines, finite differences, parallel computations, supercomputers