Abstract:We describe the first principle‐based electron force field (eFF) methodology for modeling the simultaneous dynamics of electrons and nuclei (eMD) evolving nonadiabatically under transient extreme conditions. We introduce the parallel implementation of eFF (pEFF) that makes it practical to perform simulations of the nonadiabatic dynamics of materials in extreme environments involving millions of nuclei and electrons, over multi‐picoseconds time scales, and demonstrate its application to: (i) accurately determine density and predict percent ionization of hydrogen at high pressure (∼61 GPa) and temperatures up to 15,300 K and (ii) determine, the single shock Hugoniot for lithium metal directly from the shock wave kinematics, i.e., mass velocities (Up) and shock wave velocities (Us), and shock density data. For (i), the density and ionization fractions of hydrogen atoms were calculated using the isobaric–isothermal ensemble at an isotropic pressure of 61.4 GPa and for temperatures between 300 K and 15,300 K. The results at 15,300 K describe a molecular fluid with density ρ = 0.36 g/cm3, in close agreement with existing experiments and theory, and ∼0.5% ionization. This result provides no indication of the existence of a critical plasma phase‐transition point at this particular temperature and pressure, as previously predicted by others. For (ii), the relationship between Up and Us was characterized to be linear and plastic in the range 1–20 km/s, and the single shock Hugoniot was determined in close agreement with published results for experimentally reported Ups. In addition to this, we provide a description of the materials' behavior for large Ups in terms of the appearance of a weak metallic plasma phase by Up = 10 km/s, with ≃ 8% ionization, gradually transitioning to a denser plasma with an estimated ≃ 35% ionization by Up = 15 km/s. Last but not least, we confirm the computational efficiency and scalability of pEFF by comparing its single processor performance against the fastest existing serial code, which results in a linear speedup ∼10× faster for every 16,000 particles in favor of pEFF, and by evaluating its parallel performance in terms of its strong and weak scaling capabilities. Our results, on Los Alamos's Lobo supercomputer (a 38TFLOPSs Linux HPC with Quad‐core AMD Opteron nodes interconnected with an Infiniband), show strong scaling with near ideal speedups for loads >62 particles per processor. Weak scaling is shown to be close to linear under the same per‐processor load range. As an absolute reference, an NVT run with 2 million particle lithium bulk system (0.5 M nuclei and 1.5 M electrons) on Lobo takes ∼0.44 s/timestep on 1024 processors (∼1 day/ps using an integration timestep of 0.005 fs). © 2010 Wiley Periodicals, Inc. J Comput Chem, 2011

Shock Hugoniot calculations using on-the-fly machine learned force fields with ab initio accuracy

An ab initio approach to the Hugoniot

Efficient Calculation of the Lattice Thermal Conductivity by Atomistic Simulations with Ab Initio Accuracy

Adaptive Local Basis Set for Kohn-Sham Density Functional Theory in a Discontinuous Galerkin Framework II: Force, Vibration, and Molecular Dynamics Calculations

On-the-fly machine learned force fields for the study of warm dense matter: application to diffusion and viscosity of CH

Path Integral Calculations of the Hydrogen Hugoniot Using Augmented Nodes

Accurate Hellmann-Feynman forces from density functional calculations with augmented Gaussian basis sets

By-passing the Kohn-Sham equations with machine learning

Accurate interatomic force field for molecular dynamics simulation by hybridizing classical and machine learning potentials

A Machine Learning Force Field for Bio-Macromolecular Modeling Based on Quantum Chemistry-Calculated Interaction Energy Datasets

On-the-fly machine learning force field generation: Application to melting points

Large-scale, Long-Term Nonadiabatic Electron Molecular Dynamics for Describing Material Properties and Phenomena in Extreme Environments.

Non-adiabatic Dynamics Modeling Framework for Materials in Extreme Conditions

Machine learned Force-Fields for an ab-initio Quality Description of Metal-Organic Frameworks

High-accuracy thermodynamic properties to the melting point from ab initio calculations aided by machine-learning potentials

Atomic cluster expansion potential for large scale simulations of hydrocarbons under shock compression

Scalable Ab Initio Electronic Structure Methods with Near Chemical Accuracy for Main Group Chemistry

Machine Learning Inter-Atomic Potentials Generation Driven by Active Learning: A Case Study for Amorphous and Liquid Hafnium dioxide

Accelerating finite-temperature Kohn-Sham density functional theory with deep neural networks

Large-scale Molecular Simulations of Hypervelocity Impact of Materials