数据与计算发展前沿 ›› 2023, Vol. 5 ›› Issue (4): 27-37.

CSTR: 32002.14.jfdc.CN10-1649/TP.2023.04.003

doi: 10.11871/jfdc.issn.2096-742X.2023.04.003

• 专刊: “基础研究”联合专刊 • 上一篇    下一篇

基于机器学习的力场模型研究综述

陈美霖1,2(),刘端阳3,徐黎明1,2,汪洋1,*()   

  1. 1.中国科学院计算机网络信息中心，北京 100083
    2.中国科学院大学，北京 100049
    3.中国科学院半导体研究所，北京 100083
  • 收稿日期:2023-06-01 出版日期:2023-08-20 发布日期:2023-08-23
  • 通讯作者: *汪洋（E-mail: wangyang@cnic.cn）
  • 作者简介:陈美霖，中国科学院计算机网络信息中心，硕士研究生，主要研究方向为机器学习力场等。
    本文承担工作为文献的搜集整理及整体内容的撰写。
    CHEN Meilin is a master student of Computer Network Information Center, Chinese Academy of Sciences. Her major research field is machine learning force fields.
    In this paper, she is responsible for collective literature reviews and thesis writing.
    E-mail: mlchen@cnic.cn|汪洋，中国科学院计算机网络信息中心，战略中心主任，中国科学院大学硕士研究生导师，博士，高级工程师，主要研究方向为大数据分析、态势感知系统。
    本文承担的工作为整体规划和论文指导。
    WANG Yang is the director of the Strategic Center of the Computer Network Information Center of the Computer Network Information Center, the master's supervisor, doctor, and senior engineer of the University of the Chinese Academy of Sciences. His main research interests are big data analysis and situation awareness systems.
    In this paper, he is responsible for the overall planning and paper guidance.
    E-mail: wangyang@cnic.cn
  • 基金资助:
    中国科学院网络安全和信息化专项应用示范培育项目“集成电路用单晶硅加工工艺的人工智能辅助软件与平台”(CAS-WX2023PY-0101)

A Review of Force Field Models Based on Machine Learning

CHEN Meilin1,2(),LIU Duanyang3,XU Liming1,2,WANG Yang1,*()   

  1. 1. Computer Network Information Center, Chinese Academy of Sciences, Beijing 100083, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
    3. Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China
  • Received:2023-06-01 Online:2023-08-20 Published:2023-08-23

RichHTML

29

PDF

679

摘要：

【应用背景】在过去的几十年里，由于原子结构以及计算的复杂性，传统力场方法在解决某些问题时较为吃力。【目的】而机器学习方法的引入，有望解决许多曾经无法攻克的难题，平衡计算效率和计算精度之间的制约关系。【方法】该方法不依赖于先入为主的知识，通过从小规模高精度分子动力学模拟数据中学习来对力场进行建模，同时对原子核和核外电子的运动做了近似假设，从而很大程度上简化了力场的生成过程。【结果】机器学习力场旨在达到与传统力场几乎同样的精度并大幅度地提高计算效率。本文概述了机器学习力场的发展以及其相关理论知识，介绍了几种比较常见的机器学习力场方法，最后探讨了机器学习力场的不足以及未来需要克服的挑战。

关键词: 半导体领域机器学习, 机器学习力场, sGDML, 神经网络

Abstract:

[Background] In the past few decades, due to complexity of the atomic structure and the computation for investigating the structure, traditional force field methods have been struggling in solving certain problems. [Objective] The introduction of machine learning methods is expected to solve many previously intractable problems and balance the constraints between computational efficiency and accuracy. [Methods] Machine learning force fields methods do not rely on preconceived knowledge and model the force field by learning from small-scale high-precision molecular dynamics simulation data. At the same time, approximate assumptions are made for the motion of atomic nuclei and extranuclear electrons, greatly simplifying the generation process of the force field. [Results] Machine learning force fields methods aim to achieve almost the same accuracy as traditional force fields methods while significantly improving computational efficiency. This article provides an overview of the development and related theoretical knowledge of the machine learning force fields methods, introduces several common methods, and finally explores the shortcomings of machine learning force fields methods and the challenges that need to be tackled in the future.

Key words: machine learning in semiconductor field, machine learning force field, sGDML, neural network