Shock initiation in energetic crystals is governed by extreme, transient pressure-shear fields that couple mechanics to chemistry far from equilibrium. A first-principles neural network potential (NNP) is developed to enable nanometer-scale shock simulations of RDX with near electronic-structure fidelity, and is applied to resolve the earliest chemistry triggered by collapse of a 40 nm nanovoid. Relative to a widely used ReaxFF parametrization, the NNP predicts a thinner and more coherent peripheral reaction front stabilized by a strong counter-rotating vortex pair, which delays bulk conversion while intensifying rim-localized mechanochemistry. Increasing piston speed tightens the vortex pair, enhances shear localization, and accelerates downstream conversion. High-pressure potential-energy surfaces further rationalize pathway selection: HONO elimination becomes strongly disfavored under compression, whereas N-NO2 scission, ring opening, and an intermolecular O-transfer channel remain kinetically accessible. Together, these results establish a mechanistic link between vortex-controlled shear localization and pressure-reordered reaction pathways, highlighting first-principles NNPs as a reliable route toward predictive shock-chemistry modeling in energetic materials.
Zhang et al. (Fri,) studied this question.