Accelerated Molecular Dynamics through stochastic iterations to strengthen yield of path hopping over upper states (SISYPHUS)

We present a new method, called SISYPHUS (Stochastic Iterations to Strengthen Yield of Path Hopping over Upper States), for extending accessible time-scales in atomistic simulations. The method proceeds by separating phase space into basins, and transition regions between the basins based on a general collective variable (CV) criterion. The transition regions are treated via traditional molecular dynamics (MD) while Monte Carlo (MC) methods are used to (i) estimate the expected time spent in each basin and (ii) thermalize the system between two MD episodes. In particular, an efficient adiabatic switching based scheme is used to estimate the time spent inside the basins. The method offers various advantages over existing approaches in terms of (i) providing an accurate real time scale, (ii) avoiding reliance on harmonic transition state theory and (iii) avoiding the need to enumerate all possible transition events. Applications of SISYPHUS to low temperature vacancy diffusion in BCC Ta and adatom island ripening in FCC Al are presented. A new CV appropriate for such condensed phases, especially for transitions involving collective motions of several atoms, is also introduced.

Realistic time-scale fully atomistic simulations of surface nucleation of dislocations in pristine nanopillars

We use our recently proposed accelerated dynamics algorithm (Tiwary & van de Walle, 2011) to calculate temperature and stress dependence of activation free energy for surface nucleation of dislocations in pristine Gold nanopillars under realistic loads. While maintaining fully atomistic resolution, we achieve the fraction of a second time-scale regime. We find that the activation free energy depends significantly on the driving force (stress or strain) and temperature, leading to very high activation entropies. We also perform compression tests on Gold nanopillars for strain rates varying between 7 orders of magnitudes, reaching as low as 10^3/s. Our calculations show the quantitative effects on the yield point of unrealistic strain-rate Molecular Dynamics calculations: we find that while the failure mechanism for <001> compression of Gold nanopillars remains the same across the entire strain-rate range, the elastic limit (defined as stress for nucleation of the first dislocation) depends significantly on the strain-rate. We also propose a new methodology that overcomes some of the limits in our original accelerated dynamics scheme (and accelerated dynamics methods in general). We lay out our methods in sufficient details so as to be used for understanding and predicting deformation mechanism under realistic driving forces for various problems.