Analysis of nonadiabatic molecular dynamics trajectories

The trajectory-based nonadiabatic dynamics simulation of complex systems generally produces a large number of output data containing all details of dynamics evolution. It is not possible to simply view or analyze such a huge amount of high-dimensional data by direct human observation, thus it is necessary to extract or identify a few key coordinates governing the geometric evolution for the simplified understanding of the reaction mechanism that is relevant to chemical intuition. Unsupervised machine learning approaches can help us to analyze the trajectory evolution in the nonadiabatic molecular dynamics. Several dimensionality reduction approaches are briefly discussed and a few examples are given to show the step-by-step analysis protocol. The overall purpose of this chapter is to provide a tutorial review on how to perform the analysis of trajectory evolution.