Modeling surfaces of crystalline materials
Paul Upton, Applied Maths, uses the cluster for modeling surfaces of crystalline materials such as metals and semiconductors.
Particular attention is paid to the kinetics of step-edge roughening of vicinal surfaces. To do this, Monte Carlo simulations of simplified lattice models are run on the cluster and results are compared against purely theoretical predictions coming from stochastic models.
It is necessary to run repeated simulations with different random number seeds to get an averaged time series. Also, simulations are carried out with lattices of different sizes (i.e., numbers of lattice sites) in order to get a "scaling collapse" of data so as to check scaling predictions coming from the theory. The Linux cluster was absolutely essential for this work since different time runs can be placed on different cluster nodes and many repeated simulations are required to get good statistics.