Co-authors: Sheng-Han Teng, Hsin-An Chen, Chun-Wei Pao
Research Center for Applied Sciences, Academia Sinica, Taipei, Taiwan
LAMMPS Implementation of Neural Network Potential and Exhaustive Sampling of Complex Materials
Chemically complex materials, namely, materials comprised of multiple constituent species and molecules, are playing increasingly important role in material applications such as organometallic-halide perovskite materials and high entropy alloys. However, these materials pose grand challenges in modeling and simulation of their microstructure properties, which are critical for their performance. In this presentation, we will present our recent LAMMPS implementation of the neural network potential based on Behler and Parrinello energy partitioning scheme. We will demonstrate that the lammps-implemented neural network potential can provide up to 10^5 computational boost with respect to DFT calculations for perovskite materials, thereby allowing exhaustive sampling such as Monte Carlo simulation to investigate the configurations of high entropy alloys or mixed ion perovskite materials.