EM-HyChem: Bridging molecular simulations and chemical reaction neural network-enabled approach to modelling energetic material chemistry

This study introduced a physics-inspired, top-down approach for modelling the reaction kinetics of energetic materials, based on observations of the time scale separation between pyrolysis and oxidation reactions. This modelling approach, named EM-HyChem, was developed with the inspiration of the original hybrid chemistry (HyChem) model, in which the reaction mechanism is divided into two submodels: pyrolysis and oxidation. In EM-HyChem, the key pyrolysis products and reaction mechanism are identified from the perspective of molecular fragments via geometry analysis, which is validated via neural network potential-enabled molecular dynamic simulations. A chemical reaction neural network (CRNN) model is applied to extract the rate parameters for the pyrolysis step from the reproduction of thermogravimetric experiments. An EM-HyChem model is later constructed by combining the pyrolysis step together with the oxidation models for the pyrolysis products. Two representative EMs, i.e., 1,3,5-trinitroperhydro-1,3,5-triazine (RDX) and 1,3,5,7-tetranitro-1,3,5,7-tetrazocane (HMX), are considered here to evaluate the performance of the EM-HyChem model. The predicted burning rates across a wide range of pressure conditions (1–100 atm) are in good agreement with the experimental measurements and the results of other models. Further agreement among the temperature profile, melt layer thickness and surface temperatures support the EM-HyChem model.