AVS 47th International Symposium
    Vacuum Technology Wednesday Sessions
       Session VT-WeA

Paper VT-WeA5
Modelling Micro-Channel Flows with DSMC and a Particle Continuum Method

Wednesday, October 4, 2000, 3:20 pm, Room 201

Session: Vacuum Gas Dynamics
Presenter: T.J. Bartel, Sandia National Laboratories
Authors: T.J. Bartel, Sandia National Laboratories
M.A. Gallis, Sandia National Laboratories
Correspondent: Click to Email
Advances in micromachining technology have enabled fabrication of MEMS devices. Even at atmospheric pressures, flow through these devices can range from rarefied to transitional due to the length scale and the mean free path. The Direct Simulation Monte Carlo (DSMC) technique, a direct particle based simulation of the Boltzmann equation, is an appropriate and accurate simulation tool for these devices; however, the computational expense of determining subsonic flows is very large. The major issue is one of statistical resolution: the computational particle velocities are at their thermal speeds while the mean gas velocity is at M ~ 0.01. We will present a overview of the issues of using DSMC and show results for micro-channel flows obtained with our massively parallel version. We have developed a new strategy which accelerates convergence to the steady-state. In this method, we have define the computational particle as a mass per time quantity (rho*V*A). Now the computational phase space is not stochastic, but is deterministic based on the resonance time of each computational particle in a cell. This formulation greatly reduces the statistical noise and computational expense experienced in pure particle simulations. We use a Langevin formulation for the viscous and pressure forces. We use the DSMC method to obtain a spatial varying model for the transport properties (similar to a turbulence model) to extend the validity of the continuum method to non-continuum flows. We will compare results from this new method with DSMC and discuss future work. This strategy is more robust and computationally efficient than split domain methods: DSMC is used for a portion of the domain and a continuum solver for the remainder.