Harnessing machine learning enabled quickly predicting density of CHON molecules for discovering new energetic materials

The application of machine learning in the research and development of energetic materials is becoming increasingly widespread for performance prediction and inverse design. Many advances have been achieved, especially in the discovery of various new energetic materials. However, the research of main properties such as data acquisition, molecular characterization, and limitations of research objects is insufficient. Density, as a critical factor influencing the detonation performance of energetic materials, is difficult to predict with high precision and speed at a large scale. In this study, machine learning techniques are employed to predict the density of CHNO materials and as a result to explore new energetic materials simultaneously possessing high performance and stability. By screening the dataset of 16 548 candidate molecules, 175 potential high-performance energetic molecules were identified. Among the candidates, it is noted that the molecule with a detonation velocity of 7.328 Km/s and a detonation pressure of 24.48 GPa was achieved, which is comparable to TNT. The study shows that the transformative potential of machine learning in accelerating the discovery of novel energetic materials vital for diverse applications and the optimized molecules are expected to accelerate the development of next-generation energetic materials.