Thermal stability ranking of energetic crystals via a neural network potential-enabled MD simulation protocol

Abstract

Assessing the thermal stability of energetic materials (EMs) remains challenging due to the limitations of traditional experimental and computational methods. This study develops an optimized molecular dynamics (MD) protocol based on a neural network potential (NNP) to enable reliable quantitative prediction of EM thermal stability. Key improvements include the use of nanoparticle models and reduced heating rates.​ Systematic investigations on RDX show that nanoparticle structures mitigate decomposition temperature (T_d) overestimation compared to periodic models, with surface effects dominating over particle size. Lower heating rates (e.g., 0.001 K/ps) further reduce deviation, bringing RDX T_d within 80 K of experimental values (vs. > 400 K in conventional simulations). Kissinger analysis of the heating rate-T_d relationship supports the feasibility of optimizing heating rates to align with experimental T_d.​ Applied to eight representative EMs, the optimized protocol yields thermal stability rankings in excellent agreement with experiments (R² = 0.96), outperforming traditional periodic models (R² = 0.85). This work establishes a robust computational framework for EM thermal stability evaluation, particularly valuable in data-limited scenarios.