DeePMD-kit | Computational Chemical Sciences Open-Source Software Development Group

DeePMD-kit is an open-source package for constructing deep-neural network (DNN) representations of the interatomic potential energy surface (PES) derived from ab-initio data. The DNN-PES reproduces the ab-initio PES with high accuracy and is extensive and differentiable. DeePMD-kit is interfaced with the popular open-source tool LAMMPS to perform large-scale molecular dynamics (MD) simulations. MD simulations based on the DNN-PES, called deep potential molecular dynamics (DPMD), have computational cost that scales linearly with system size and is many orders of magnitude lower than that of AIMD for the same system.

Tiger @ Princeton

Software Application

Machine Learning

Supercomputers

Mira

Cori

Other

Download

https://github.com/deepmodeling/deepmd-kit

Documentation

http://chemlabs.princeton.edu/ccsc/tutorials/#dpmd

Publications

L. Zhang, J. Han, H. Wang, R. Car, and W. E, Phys. Rev. Lett. 120, 143001 (2018).

L. Zhang, J. Han, H. Wang, W. Saidi, R. Car, and W. E, in Advances in Neural Information Processing Systems 31, edited by S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett (Curran Associates, Red Hook, 2018) pp. 4436–4446.

CCS Contributor(s)

Computational Chemical Science Center: Chemistry in Solution and at Interfaces

Software Contributions Supported by CCS award

This center is the sole developer.

Summary

DeePMD-kit is an open-source package for constructing deep-neural network (DNN) representations of the interatomic potential energy surface (PES) derived from ab-initio data.