A DeepMD-GNN Potential Reveals the Initiation Mechanism of a Trinitromethyl-Based Green Primary Explosive

Abstract

Trinitromethyl-based green explosives have garnered extensive research interest by virtue of their high mass density, superior energy density, favorable oxygen balance, and eco-friendly properties. However, the development of new energetic materials remains challenging because experimental studies are costly, hazardous, time-consuming, and often inefficient, making it difficult to directly resolve microscopic initiation processes. Consequently, theoretical simulations are of great significance for guiding explosive design. Taking the representative trinitromethyl-based explosive 4-diazo-5-nitro-2-(trinitromethyl)-2,4-dihydro-3H-pyrazol-3-one as a model system, we developed a deep potential-based graph neural network potential for trinitromethyl energetic materials and established a framework for investigating their energetics, structural evolution, and hotspot-induced reaction mechanisms. The resulting model achieves near-density-functional-theory accuracy in energies and forces while accurately reproducing equation-of-state behavior and structural correlations. Moreover, it exhibits strong transferability to structurally related trinitromethyl-containing molecules. Reactive molecular dynamics simulations reveal a rearrangement-dominated initiation mechanism, in which intramolecular rearrangements and reversible isomerization precede extensive bond cleavage, followed by nitro-group elimination and ring-opening reactions. By combining first-principles accuracy with large-scale simulation capability, this work provides an efficient framework for studying trinitromethyl-based energetic materials and guiding the design of next-generation green explosives.