Accelerating Atomistic Modeling with RANN Potentials

We study the implementation of Rapid Artificial Neural Network (RANN) potentials in the LAMMPS simulation software. Due to complicated atomic descriptors and lengthy neighbor-list computations, traditional machine learning potentials frequently achieve excellent accuracy but have large computational costs. 

In order to overcome these difficulties, the RANN framework makes use of angular screening approaches, which lower computing complexity without sacrificing prediction accuracy. The lab uses energy conservation tests, force and pressure validation, and large-scale molecular dynamics simulations to assess the numerical stability and physical dependability of these cutting-edge machine learning force fields.

Through this project, students gain highly sought-after practical experience in high-performance computing, computational materials engineering, molecular dynamics simulations, LAMMPS, interatomic potential creation, and machine learning for materials science.

If you are eager to push the boundaries of scientific computing, computational physics, materials science, machine learning, and artificial intelligence, we are looking for driven undergraduate students like you to join our expanding research team.