Density Effects on the Thermal Decomposition of LLM-105 Explored by Neural Network Potential

2,6-Diamino-3,5-dinitropyrazine-1-oxide (LLM-105), a relatively new insensitive explosive, has attracted considerable attention. The initiation of explosives is generally associated with localized “hot spots” formed in voids or pores, making density effects crucial for understanding their decomposition behavior. Here, we developed an ab initio neural network potential (NNP) for LLM-105 and employed deep potential molecular dynamics (DPMD) simulations to investigate its thermal decomposition at different densities. The results reveal pronounced density dependence throughout the decomposition process. At higher densities, intermolecular reactions are enhanced while unimolecular C–NO2 cleavage is suppressed, leading to reduced NO2 release and increased H2O formation. Elevated densities also accelerate product formation and overall decomposition rates. The initial decomposition temperature remains nearly constant across different densities. Meanwhile, lower-density regions exhibit enhanced energy release, characteristic of pore-collapse hotspots, which implies a greater tendency for detonation initiation. Furthermore, the reaction network of LLM-105 was established, with the primary pathways dominated by intramolecular hydrogen transfer and C–NO2 cleavage. The reversible hydrogen transfer acts as a buffering step that may contribute to the structural stability and low sensitivity of LLM-105. Two low-barrier pathways leading to the major products N2 and CO2 were identified: unimolecular dissociation of intermediate IM3 and bimolecular reactions between IM3 and NO2.