The project is based on the results of the initial Marie Curie Individual Fellowship during which we have developed the simulation scheme consisting in intertwining the short interval of standard molecular dynamics simulation with the longer interval of li nearly predicted dynamics with the overall speed-up close to 6 in comparison to standard molecular dynamics simulations of a system consisting of a few hundred flexible oxygen molecules. During the reintegration phase we would like to understand better the open methodological questions of the approach. The first methodological question which will be studied is the (linear) predictability of the systems simulated with and without the cut-off, as we have empirically found that the mean square 1-step predictio n error of the system simulated without the cut-off is by two orders of magnitude lower. The other methodological question is the comparison of the linearly predicted molecular dynamics trajectory against its extrapolation based on a Taylor series. As we h ave found that the linear prediction approach provides the trajectory which is much closer to "real" trajectory, we will investigate the relation between the two. As the second part of the project we will implement the method into the gen eric simulation package and will test the efficiency of the method in simulations of the molecules of applied importance. The application of the method to more complex molecules and molecular fragments (towards the ultimate goal - proteins and DNA) will be tried and the limits of the approach will be outlined. During the second part of the project we will study the possibility to accelerate molecular dynamics simulations by using other non-traditional simulation techniques (playing the role of "vir tual dynamics"), such as autoregressive modelling and, cellular automata. We will apply the autoregressive modelling to simulate stochastic force in the simulation of polymer liquids.
Call for proposal
See other projects for this call