Supercritcal to Superionic: ACE'ing the response of water, hydrocarbons, and ammonia under shock conditions

We present three newly developed machine learning interatomic potentials (MLIPs) for water, hydrocarbons, and ammonia, designed to provide quantum-accurate descriptions of atomic behavior under shock conditions. Built upon the atomic cluster expansion (ACE) framework, these potentials are specifically parameterized to enable high-fidelity molecular dynamics (MD) simulations at extreme conditions—up to 100 GPa and temperatures ranging from 500 K to 6000 K. Validation demonstrates that our ACE potentials accurately reproduce experimental data and density functional theory (DFT) predictions for various thermodynamic properties, such as shock Hugoniots, transport properties, and the complex phase diagrams of water and ammonia, including supercritical fluids, superionic ice, and dissociated mixtures. This work offers three robust interatomic potentials, enabling large-scale, precise simulations of water, hydrocarbons, and ammonia under shock conditions.