Developing machine learning interatomic potentials for classical molecular dynamics simulations of complex perovskites

Project Information

machine-learning, molecular-dynamics
Project Status: In Progress
Project Region: CAREERS
Submitted By: Gaurav Khanna
Project Email: ashgiri@uri.edu
Project Institution: University of Rhode Island
Anchor Institution: CR-University of Rhode Island

Mentors: Ashutosh Giri, Michael Strickler
Students: Jaymes Dionne

Project Description

Our research group is developing machine learned (ML)-interatomic potentials for molecular dynamics simulations geared towards understanding the thermal properties of complex perovskites structures. The perovskite materials that will be modeled under this project will include metal halide perovskites and oxide-based perovskites. The ML-based potential development process will include gathering training data via density functional theory calculations followed by the utilization of deep learning framework to construct deep potential neural model. Ultimately, the potentials will be utilized for molecular dynamics simulations of the perovskites that will be performed with the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) package. We will utilize URI’s HPC or the UNITY cluster to perform the tasks.

The student will obtain extensive experience working on an HPC cluster (command-line Linux, LAMMPS package, SLURM job scheduler, optimal submission parameters etc.) and will also learn to use the generated data-sets to train a ML/DL model.

Project Information

machine-learning, molecular-dynamics
Project Status: In Progress
Project Region: CAREERS
Submitted By: Gaurav Khanna
Project Email: ashgiri@uri.edu
Project Institution: University of Rhode Island
Anchor Institution: CR-University of Rhode Island

Mentors: Ashutosh Giri, Michael Strickler
Students: Jaymes Dionne

Project Description

Our research group is developing machine learned (ML)-interatomic potentials for molecular dynamics simulations geared towards understanding the thermal properties of complex perovskites structures. The perovskite materials that will be modeled under this project will include metal halide perovskites and oxide-based perovskites. The ML-based potential development process will include gathering training data via density functional theory calculations followed by the utilization of deep learning framework to construct deep potential neural model. Ultimately, the potentials will be utilized for molecular dynamics simulations of the perovskites that will be performed with the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) package. We will utilize URI’s HPC or the UNITY cluster to perform the tasks.

The student will obtain extensive experience working on an HPC cluster (command-line Linux, LAMMPS package, SLURM job scheduler, optimal submission parameters etc.) and will also learn to use the generated data-sets to train a ML/DL model.