DeePMD ML Potential Model for Molecular Simulation

Molecular Dynamics with DeePMD [DeepModeling Tutorial!]

Required Libraries : numpy, pandas, deepmd,dpdata, ase, matplotlib

System : 1,3-Butadiene Cyclization

This is a tutorial for DeePMDpackage, a Machine Learning Interatomic Potential (MLP) model for interatomic potential energy and force field and to perform molecular dynamics (MD).

In-general, steps to build MLPs usually require three main steps:

1. Data generation [00.data]
2. Training [01.train]
3. Molecular dynamics [02.lmp]

For data generation, we will be using the [VASP](https://www.vasp.at/") package. \

Here, I will not go into the details for generating the initial configuration. However, there are several ways to do that. A straightforward approach involves performing a classical MD simulation to obtain random configurations. Subsequently, ab-initio calculations are performed to compute the potential energy and corresponding forces (in this case, we will be using VASP). \

The output configurations are stored in the folder inside the 00.data directory.

Note: Energies and forces for these configurations are obtained using PBE functional.

Author:

Rahul Verma
Dept. of Chemical and Biomolecular Engineering
North Carolina State University, Raleigh, NC, USA
Email: [email protected]