Molecular Dynamics Simulation of Fluid Systems.

In this dissertation, we use molecular dynamics simulations to study several interesting phenomena in fluid systems. Some novel computational techniques are also introduced, including the use of dynamic linked lists, a method for simulating cellular automata, and a numerical diffraction method for visualizing crystal structures. Firstly, we study compressible fluid flow in narrow two-dimensional channels. In the simulation area, an upstream source is maintained at constant density and temperature while a downstream reservoir is kept at vacuum. A novel wall model, incorporating the effects of inelastic collisions between fluid and wall molecules, molecular structure of the wall, and a long-ranged fluid-wall interaction, is used in these simulations. We find that the velocity distribution differs significantly from a parabolic profile close to the walls. Also, for a sufficiently strong fluid-wall attraction the fluid velocity goes smoothly to zero at the wall, indicating the absence of velocity slip. As the fluid becomes more rarefied, the flow velocity goes to zero at the wall increasingly sharply in an increasingly narrow region, behavior which could be interpreted as velocity slip. These results are different from those of the usual kinetic theory of fluid flow at walls and also from the predictions of the Chapman-Enskog method. They are, however, in qualitative agreement with the non-linear theory of Eu. In the second system we study, instead of inducing flow with a pressure or density gradient, we drive an external temperature gradient across the system to simulate a directional solidification process. A solid is formed as the externally imposed temperature gradient is pulled across a sample which is initially in the liquid phase. If the solid-liquid interface moves more slowly than the pulling speed, the initially flat interface develops large sinusoidal irregularities when the tail of the external temperature gradient passes the moving interface. We identify the instability as a Mullins-Sekerka type, arising from the density gradient in the liquid phase which is in turn a consequence of the density difference between the solid and liquid phases.