Npj Comput. Mater.: “鱼与熊掌”可以兼得—兼具高精度、高速度的分子动力学！

2023-05-17 03:05

海归学者发起的公益学术平台

分享信息，整合资源

交流学术，偶尔风月

分子动力学（molecular dynamics, MD）广泛应用于物理、化学、生物、制药、地质、材料、纳米技术、半导体芯片等领域的原子尺度微观分析。然而，长期以来，在“速度”和“精度”这两个核心指标上，分子动力学存在“鱼与熊掌不可兼得”的问题——经典分子动力学（classical MD, CMD）速度快，但精度低，难以满足高精度计算要求；第一性原理分子动力学（ab-initio MD, AIMD）精度高，但速度慢，难以计算大系统。研发出兼具高精度、高速度的分子动力学方法，一直是各个领域的研究人员追求的共同目标。

Fig. 1 The comparison of predicted energy and force.

来自湖南大学电气与信息工程学院的刘杰教授团队基于自主研制的非冯·诺依曼（non von Neumann, NVN）芯片架构，提出了一种新型分子动力学方法NVNMD。NVNMD利用深度神经网络模型可以高精度拟合高维势能面的优势，保障了分子动力学的计算精度；并利用新型非冯·诺依曼架构的高效率优势，保障了分子动力学的计算速度。

Fig. 2 The molecular dynamics simulation results of GeTe and Li10GeP2S12 using the proposed NVNMD.

因此，NVNMD兼具AIMD级别的高精度、CMD级别的高速度——实现了“鱼与熊掌可以兼得”。通过使用小分子体系、“晶态-不定形态”相变、四元素体系内原子扩散等多种类别的分子动力学计算开展测试，证明了NVNMD兼具高速度与高精度的优势。由于NVNMD为通用性分子动力学方法，其在各个领域的微观分析和基础研究中有广泛应用前景。

Fig. 3 Schematic comparison of calculation step between vN and NvN architecture.

该文近期发表于npj Computational Materials 8，107(2022)，英文标题与摘要如下，点击左下角“阅读原文”可以自由获取论文PDF。

Accurate and Efficient Molecular Dynamics based on Machine Learning and Non Von Neumann Architecture

Pinghui Mo, Chang Li, Dan Zhao, Yujia Zhang, Mengchao Shi, Junhua Li & Jie Liu

Force field-based classical molecular dynamics (CMD) is efficient but its potential energy surface (PES) prediction error can be very large. Density functional theory (DFT)-based ab-initio molecular dynamics (AIMD) is accurate but computational cost limits its applications to small systems. Here, we propose a molecular dynamics (MD) methodology which can simultaneously achieve both AIMD-level high accuracy and CMD-level high efficiency. The high accuracy is achieved by exploiting deep neural network (DNN)’s arbitrarily-high precision to fit PES. The high efficiency is achieved by deploying multiplication-less DNN on a carefully-optimized special-purpose non von Neumann (NvN) computer to mitigate the performance-limiting data shuttling (i.e., ‘memory wall bottleneck’). By testing on different molecules and bulk systems, we show that the proposed MD methodology is generally-applicable to various MD tasks. The proposed MD methodology has been deployed on an in-house computing server based on reconfigurable field programmable gate array (FPGA), which is freely available at http://nvnmd.picp.vip.

扩展阅读

Npj Comput. Mater.: 局部原子邻域描述符—非数据驱动无损压缩

Npj Comput. Mater.: 轨道耦合—调控谷能带和激子劈裂

Npj Comput. Mater.: 无损数据共享平台—助力材料科学突破

Npj Comput. Mater.: DFT中的电子密度—等变图神经网络

本文系网易新闻·网易号“各有态度”特色内容

媒体转载联系授权请看下方

微信扫码关注该文公众号作者

戳这里提交新闻线索和高质量文章给我们。

来源: qq

点击查看作者最近其他文章