Active Learning of Fewest-Switches Surface Hopping Dynamics of Pyrene

This repository contains data and analysis tools on the Active Learning (AL) procedure for pyrene, as described in the publication "MELTS: Fully Automated Active Learning for Fewest-Switches Surface Hopping". A description of the archives is provided as follows: pyrene-al-results.tar.gz contains the output AL data from MELTS, including the final saved dataset and information on intermediary steps. pyrene-ml-dynamics.tar.gz contains the raw Newton-X trajectories of cumulative nanosecond-long machine learning (ML) dynamics propagated with MELTS. nanosec-dyn-csv.tar.gz contains the ML dynamics data as output by ULaMDyn, in csv format. NEA.tar.gz contains the nuclear ensemble approach (NEA) emission spectrum data for the two excited states. Additionally, there are two Jupyter notebooks with all procedures used to analyze the dataset and produce the figures used in the paper: al-analysis.ipynb analyzes the AL-FSSH procedure with the data from fulvene-al-results.tar.gz. dyn_spectrum_analysis.ipynb analyzes the ML-FSSH dynamics executed after AL, with the data from nanosec-dyn-csv.tar.gz, and also the emission spectrum, with the data from NEA.tar.gz.