Machine Learned Entropy for Phase Transitions in Aromatic Compounds

Aromatic compounds are involved in many chemical processes, for instance, as solvents or combustion products, and have been increasingly used in organic photovoltaics [1]. Here we develop a machine learning (ML) model for entropy on the example of benzene, anthracene, phenanthrene and coronene. To this end, we generate a dataset for the training of the ML model by carrying out molecular dynamics simulations of the condensed phases for each compound. Building on results from prior work [2], we use all-toms force fields to model these molecules, and carry out an extensive sampling of the parameter (temperature, density) space to generate the dataset. The dataset is then used to train and validate a ML model for the rapid determination of entropy, and the exploration of phase transitions processes.

 

[1] X. S. Zhao and J. I. Siepmann, J. Phys. Chem. B 2005, 109, 11, 5368–5374.

[2] C. F. Fu and S. X. Tian, J. Chem. Theory Comput. 2011, 7, 2240–2252.