Applied Math Seminar
Spring Quarter 2006
3:15 p.m.
Sloan Mathematics Corner
Building 380, Room 380-C


Friday, May 19, 2006

Matt West
Aeronautics and Astronautics, Stanford

Multiscale Markov Chain Simulation for Particulate Processes


Abstract:

This talk outlines accelerated methods for forward-time simulations of continuous-time Markov Chains arising from models of particulate processes. The problem of interest is the formation of rain-droplets from condensation nuclei, and determination of the final chemical composition. This simulation is challenging because of the large problem size (10^9 particles or more), the high dimensionality (5-10 variables per particle) and the continuum of active scales that prevents the direct application of scale-separated methods. We present a numerical scheme based on multi-scale accept-reject and clustered-particle dynamics that achieves significant acceleration of the simulation.