GPUMD: A package for constructing accurate machine-learned potentials and performing highly efficient atomistic simulations

DOI PDF 被引用文献3件
  • Zheyong Fan
    College of Physical Science and Technology, Bohai University 1 , Jinzhou 121013, People’s Republic of China
  • Yanzhou Wang
    MSP Group, QTF Centre of Excellence, Department of Applied Physics, Aalto University 2 , FI-00076 Aalto, Espoo, Finland
  • Penghua Ying
    School of Science, Harbin Institute of Technology 4 , Shenzhen 518055, People’s Republic of China
  • Keke Song
    Beijing Advanced Innovation Center for Materials Genome Engineering, University of Science and Technology Beijing 3 , Beijing 100083, China
  • Junjie Wang
    National Laboratory of Solid State Microstructures, School of Physics and Collaborative Innovation Center of Advanced Microstructures, Nanjing University 5 , Nanjing 210093, China
  • Yong Wang
    National Laboratory of Solid State Microstructures, School of Physics and Collaborative Innovation Center of Advanced Microstructures, Nanjing University 5 , Nanjing 210093, China
  • Zezhu Zeng
    Department of Mechanical Engineering, The University of Hong Kong 6 , Pokfulam Road, Hong Kong SAR, China
  • Ke Xu
    Department of Physics, Research Institute for Biomimetics and Soft Matter, Jiujiang Research Institute and Fujian Provincial Key Laboratory for Soft Functional Materials Research, Xiamen University 7 , Xiamen 361005, People’s Republic of China
  • Eric Lindgren
    Chalmers University of Technology, Department of Physics 8 , 41926 Gothenburg, Sweden
  • J. Magnus Rahm
    Chalmers University of Technology, Department of Physics 8 , 41926 Gothenburg, Sweden
  • Alexander J. Gabourie
    Department of Electrical Engineering, Stanford University 9 , Stanford, California 94305, USA
  • Jiahui Liu
    Beijing Advanced Innovation Center for Materials Genome Engineering, University of Science and Technology Beijing 3 , Beijing 100083, China
  • Haikuan Dong
    College of Physical Science and Technology, Bohai University 1 , Jinzhou 121013, People’s Republic of China
  • Jianyang Wu
    Department of Physics, Research Institute for Biomimetics and Soft Matter, Jiujiang Research Institute and Fujian Provincial Key Laboratory for Soft Functional Materials Research, Xiamen University 7 , Xiamen 361005, People’s Republic of China
  • Yue Chen
    Department of Mechanical Engineering, The University of Hong Kong 6 , Pokfulam Road, Hong Kong SAR, China
  • Zheng Zhong
    School of Science, Harbin Institute of Technology 4 , Shenzhen 518055, People’s Republic of China
  • Jian Sun
    National Laboratory of Solid State Microstructures, School of Physics and Collaborative Innovation Center of Advanced Microstructures, Nanjing University 5 , Nanjing 210093, China
  • Paul Erhart
    Chalmers University of Technology, Department of Physics 8 , 41926 Gothenburg, Sweden
  • Yanjing Su
    Beijing Advanced Innovation Center for Materials Genome Engineering, University of Science and Technology Beijing 3 , Beijing 100083, China
  • Tapio Ala-Nissila
    College of Physical Science and Technology, Bohai University 1 , Jinzhou 121013, People’s Republic of China

書誌事項

公開日
2022-09-20
DOI
  • 10.1063/5.0106617
公開者
AIP Publishing

この論文をさがす

説明

<jats:p>We present our latest advancements of machine-learned potentials (MLPs) based on the neuroevolution potential (NEP) framework introduced in Fan et al. [Phys. Rev. B 104, 104309 (2021)] and their implementation in the open-source package gpumd. We increase the accuracy of NEP models both by improving the radial functions in the atomic-environment descriptor using a linear combination of Chebyshev basis functions and by extending the angular descriptor with some four-body and five-body contributions as in the atomic cluster expansion approach. We also detail our efficient implementation of the NEP approach in graphics processing units as well as our workflow for the construction of NEP models and demonstrate their application in large-scale atomistic simulations. By comparing to state-of-the-art MLPs, we show that the NEP approach not only achieves above-average accuracy but also is far more computationally efficient. These results demonstrate that the gpumd package is a promising tool for solving challenging problems requiring highly accurate, large-scale atomistic simulations. To enable the construction of MLPs using a minimal training set, we propose an active-learning scheme based on the latent space of a pre-trained NEP model. Finally, we introduce three separate Python packages, viz., gpyumd, calorine, and pynep, that enable the integration of gpumd into Python workflows.</jats:p>

収録刊行物

被引用文献 (3)*注記

もっと見る

詳細情報 詳細情報について

問題の指摘

ページトップへ