Use of Interval Arithmetic and GRID Computing in Computational Molecular Science: Bounding Errors and Locating Global Minima

Does 64-bit floating point arithmetic provide sufficient precision on new generation supercomputers? For quantum calculations on large systems there are strong arguments to suggest not. We will resolve this issue and improve the numerical stability of these methods by using interval arithmetic a fundamentally different floating point paradigm that has just been implemented in a form that is usable in large-scale scientific computer codes. Intervals also provide some unique algorithms that, e.g. guarantee the location of global minima, and are ideally suited to loosely-coupled parallel computing environments (GRIDs). The utility of these approaches for locating minimum energy structures on potential energy surfaces will be explored.