TY - JOUR
T1 - Data filtering for effective analysis of crystal-solution interface molecular dynamics simulations
AU - Elts, Ekaterina
AU - Greiner, Maximilian M.
AU - Briesen, Heiko
PY - 2014/4/8
Y1 - 2014/4/8
N2 - Analysis of processes occurring at the solid-solution interface during crystal growth and dissolution simulations requires an effective way to detect rare, uncorrelated transitions from the liquid to the solid state or vice versa. Because of the oscillatory behavior of molecules, this is not a trivial problem. Usually, to take the thermal vibration and rotational flexibility of the molecules into account, the data (e.g., orientation, center of mass position) needed to determine the molecular state are averaged over some time interval. Then they are evaluated using some order parameters to classify the individual molecules as being either crystalline or in solution. In this case, the results can be very sensitive to the time interval, which is mostly chosen in some heuristic way. To suppress the problem of fast non-Markovian dynamics and to make the identification of the molecular state more reliable and robust, the application of a Kalman filter, optionally combined with a hysteretic approach, is proposed in this contribution. A scheme to estimate the filter parameters is introduced. To demonstrate the approach, simple and widely used order parameters based on the structural features of molecules are taken. The obtained results are clearly superior to those based on the data averaging technique and are important for the effective transition rates calculation as well as for the general analysis of the time evolution of interfaces.
AB - Analysis of processes occurring at the solid-solution interface during crystal growth and dissolution simulations requires an effective way to detect rare, uncorrelated transitions from the liquid to the solid state or vice versa. Because of the oscillatory behavior of molecules, this is not a trivial problem. Usually, to take the thermal vibration and rotational flexibility of the molecules into account, the data (e.g., orientation, center of mass position) needed to determine the molecular state are averaged over some time interval. Then they are evaluated using some order parameters to classify the individual molecules as being either crystalline or in solution. In this case, the results can be very sensitive to the time interval, which is mostly chosen in some heuristic way. To suppress the problem of fast non-Markovian dynamics and to make the identification of the molecular state more reliable and robust, the application of a Kalman filter, optionally combined with a hysteretic approach, is proposed in this contribution. A scheme to estimate the filter parameters is introduced. To demonstrate the approach, simple and widely used order parameters based on the structural features of molecules are taken. The obtained results are clearly superior to those based on the data averaging technique and are important for the effective transition rates calculation as well as for the general analysis of the time evolution of interfaces.
UR - http://www.scopus.com/inward/record.url?scp=84898487649&partnerID=8YFLogxK
U2 - 10.1021/ct400808d
DO - 10.1021/ct400808d
M3 - Article
AN - SCOPUS:84898487649
SN - 1549-9618
VL - 10
SP - 1686
EP - 1697
JO - Journal of Chemical Theory and Computation
JF - Journal of Chemical Theory and Computation
IS - 4
ER -