TATB under Dynamics Compression from Machine Leaning Simulations

We present the development and application of machine-learning interatomic potential based on Chebyshev polynomials to study organic molecular crystal (2,4,6-triamino-1,3,5-trinitrobenzene or TATB) under dynamic compression. We discuss the strategy to generate a diverge training dataset required for complicated chemistry of TATB. Our potential accurately predicts the structure properties and chemistry of TATB for a wide range of thermodynamic conditions. Equation of states of TATB under detonation from our simulations show excellent agreement with available experimental data. Our simulation can provide insights into the slow chemistry of TATB under dynamic compression. We also discuss the transferability of our model to other organic materials.