摘要
本文介绍了笔者课题组开发的针对复杂体系精确计算的软件包LSASP,包括核心算法,软件架构以及开发进展。与传统经验力场的程序不同,LSASP软件整合现有的通用电子结构计算软件,提供统一数据接口,以通用型全局势能面神经网络势函数作为主要势能面计算工具,涵盖全局结构搜索,自动反应路径取样,过渡态寻找,分子动力学,蒙特卡洛模拟等势能面采样功能的大尺度计算平台。LSASP软件的核心算法,随机势能面行走(SSW),具有自有知识产权,在势能面全局搜索,过渡态反应路径搜索,神经网络势函数模拟等方面具有国际竞争力。
Here we introduce a new software, large-scale atomic simulation package （LSASP）, developed at Fudan University targeting for exploring the potential energy surface （PES） of complex systems. The key algorithm, the architecture and the latest progress will be described in detail. Different from the tradi- tional force field programs, LSASP integrates common electronic structure calculation packages and novel neural network potential for force evaluation and can carry out a wide range of simulation tasks on potential energy surface exploration, including global structure optimization, automated reaction pathway sampling, the transition state searching, molecular dynamics and Monte Carlo simulation. The stochastic surface walking （SSW） method developed in Liu group play key roles in LSASP PES exploration and is competitive internationally, which can be utilized to achieve a range of functionalities, such as global optimization, transition state search and neural network potential simulation.
出处
《中国科学基金》
CSSCI
CSCD
北大核心
2018年第1期95-100,共6页
Bulletin of National Natural Science Foundation of China
关键词
势能面
全局优化
反应路径采样
LSASP软件
potential energy surface
global optimization
reaction pathway sampling
LSASP software
