MateriApps
A Portal Site of Materials Science Simulation

now 340 Apps

Inquiry / Application Request
Detailed search

DeePMD-kit

  • Openness:3 ★★★
  • Document quality:3 ★★★

A software package that generates high-accuracy interatomic potentials using deep learning trained on first-principles molecular dynamics data. The DeePMD model enables molecular dynamics simulations with density functional theory (DFT) accuracy at greatly reduced computational cost. It can be coupled with molecular dynamics codes such as LAMMPS, and is widely applicable to large-scale systems, high-temperature and high-pressure conditions, and the exploration of novel materials.

Information
Last Update:2026/01/07

Official site

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

License

GNU Lesser General Public License v3.0 (LGPL-3.0)

Availability

An environment where Python is available
Core Developers
  • Han Wang（Institute of Applied Physics and Computational Mathematics, China）
  • Linfeng Zhang（Center for Machine Learning Research, Peking University, China）
  • Jiequn Han（Department of Mathematics, Princeton University, USA）
  • Weinan E（Center for Machine Learning Research, Peking University / Princeton University, USA）
Target substance/model

Solid
Physical quantities that can be computed

pressure/stress tensor, energy distribution function / radial distribution function, elastic constants / bulk modulus, diffusion coefficient, thermal conductivity, phase transition analysis, interface structure, reaction path
Methodology

Deep neural network
Parallelization

Supports hybrid parallelization combining MPI parallelism and GPU acceleration (CUDA)
Related App
Document

Han Wang, Linfeng Zhang, Jiequn Han, E. Weinan, Comput. Phys. Comm. 228, 178 (2018).
Jinzhe Zeng, et al., J. Chem. Phys. 159, 054801 (2023).
Jinzhe Zeng, et al., J. Chem. Theory Comput. 21, 4375 (2025).
TOP