Shock compression pathways to pyrite silica from machine learning simulations

Silica is a fundamental component of planetary interiors, and the knowledge of its high-pressure polymorphs is essential for constructing planetary models. The pyrite silica has been synthesized in static compression with diamond anvil cells, however, it has never been observed in dynamic compression experiments, despite enormous efforts. Here, based on a machine-learning interatomic potential fitted from first-principles calculations, we perform multimillion-atom molecular dynamic simulations to study the dynamics compression pathways to obtaining the pyrite phase starting from stishovite. We find that the pyrite phase can be achieved through shock compressions along three directions of stishovite: [100], [110], and [001]. Our study reveals the orientation-dependent phase transition of stishovite during shock compression, with displacive phase transition occurring along the [100] direction and diffusive phase transformation along [110] and [001] directions. Our research provides insight into obtaining pyrite-type silica from stishovite through dynamic compression and elucidates the microscopic transformation mechanisms involved.