Computational methods to simulate large classical particle systems with applications to fluids and microporous materials.

摘要

A completely discrete—discrete space, time and energy—computational model and its implementation is presented here as an alternative to continuous type simulations such as Molecular Dynamics and continuum Monte Carlo methods. The final goal is to reduce the amount of the calculation per particle in order to increase the timescale of the simulation and the size of the particle system simulated. The computation is performed on a lattice with discrete particles in discrete states.; The fundamental particle simulation algorithms are studied. The well known neighbor-list or Verlet-list method is speeded up from quadratic to linear size dependence. This algorithm is applicable to continuous simulations also, it is lattice independent. Then the cell-list method is adapted to discrete lattices, using efficient bit manipulations. A field-representation method is developed as well as a generalization of the Lattice-Gas method.; The discrete computational techniques and other programming methods for today's advanced computer architectures are discussed in detail with their implementations in Fortran. The high performance of the whole method is mainly due to these techniques. Performance data are presented for the different methods.; The pair distribution function and density profiles inside parallel slits are calculated for different fluid densities. The lattice effects are successfully removed from the pair distribution function. The results agree well with continuum results for large slit widths. New results on layering for very narrow slit widths are presented.; Finally a lattice-based geometrical method is presented to characterize the structure of void space inside zeolites. This method is generic, it can be applied to arbitrary atomic and geometrical systems as well.