Title:Pseudo-Spectral Collocation Methods for Hyperbolic Equations with Arbitrary Precision: Applications to Relativistic Gravitation

Abstract:Numerical precision is a fundamental aspect of problems in computational physics with high accuracy requirements. The implementation of arbitrary-precision floating-point arithmetics, that can go beyond the usual 64 floating-point arithmetic, is useful whenever it can be applied to numerical algorithms that perform satisfactorily within the significant range of the number of digits of precision. In this paper, we advocate the use of Pseudo-Spectral Collocation methods for implementing high-order bit precision algorithms because of their exponential convergence (for smooth problems) and capacity for high-compression storage of the information. We illustrate the potential of these methods by applying them to very demanding evolution problems in the context of relativistic gravitation. In particular, to problems involving gravitational collapse, whose dynamics exhibits self-similarity properties.