Numerical Analysis Seminars: Computing averages from stochastic molecular dynamics
9 November 2012 14:00 in CM105
I will discuss properties of numerical algorithms for stochastic differential equations in molecular dynamics. Methods such as Langevin dynamics are reliable and well-studied performers for the purpose. In many cases, the purpose of molecular dynamics is the computation of a configurational average. I will show that in this setting it is possible to obtain a superconvergence result (an unexpected increase in order of accuracy) for stationary averages when a particular method is used. Our approach relies on a backward error analysis technique (similar to that pioneered in the setting of Hamiltonian systems). I will demonstrate performance with numerical experiments involving realistic model problems for molecular simulation.
Contact David.Bourne@durham.ac.uk for more information