[1]杨小渝,郝德博,舒城,等.MatCloud-QE:基于云原生理念的高通量第一性原理计算程序包[J].中国材料进展,2024,43(11):1007-1015.[doi:10.7502/jissn.1674-3962.202202023]
 YANG Xiaoyu,HAO Debo,SHU Cheng,et al.MatCloud-QE: A Cloud-Native High-ThroughputFirst-Principles Simulation Software Package[J].MATERIALS CHINA,2024,43(11):1007-1015.[doi:10.7502/jissn.1674-3962.202202023]
点击复制

MatCloud-QE:基于云原生理念的高通量第一性原理计算程序包()
分享到:

中国材料进展[ISSN:1674-3962/CN:61-1473/TG]

卷:
43
期数:
2024年第11期
页码:
1007-1015
栏目:
出版日期:
2024-11-29

文章信息/Info

Title:
MatCloud-QE: A Cloud-Native High-ThroughputFirst-Principles Simulation Software Package
文章编号:
1674-3962(2024)11-1007-09
作者:
杨小渝郝德博舒城许立芳梁亚男马新杰
1. 中国科学院计算机网络信息中心,北京 100190 2. 中国科学院大学,北京 100049 3. 北京迈高材云科技有限公司,北京 100190
Author(s):
YANG XiaoyuHAO DeboSHU ChengXU LifangLIANG YananMA Xinjie
1. Computer Network Information Centre,Chinese Academy of Sciences,Beijing 100190, China 2. University of Chinese Academy of Sciences, Beijing 100049, China 3. Beijing Maigao MatCloud Co., Ltd., Beijing 100190, China
关键词:
MatCloud-QE第一性原理计算程序包MatCloud+材料云第一性原理计算高通量材料计算云原生
Keywords:
MatCloud-QE first-principles simulation software package MatCloud+ cloud platform first-principles simulation high-throughput materials simulationCloud native
分类号:
TB30;TP319
DOI:
10.7502/jissn.1674-3962.202202023
文献标志码:
A
摘要:
国内开展第一性原理计算总是局限于某一两个商业软件,不仅需要软件使用许可,而且还需要下载、安装到本地,再通过人工部署到高性能计算集群才能开展使用,极大地限制了科研人员开展第一性原理计算。为了解决这个问题,作者团队基于Quantum ESPRESSO第一性原理计算内核,进一步开发了高通量第一性原理计算加速算法、参数智能推荐算法、输入结构统一算法、自动后处理引擎、组件化引擎、图形化引擎、更多物性计算功能以及MatCloud+接口等算法程序,形成了具有云原生特点的第一性原理计算程序包:MatCloud-QE;进一步通过接入MatCloud+材料云,使用户仅通过浏览器就可在线开展第一性原理计算和数据自动化采集和管理,极大地方便了用户开展第一性原理计算以及高通量计算筛选,且无需软件使用许可。以面向太阳能电池应用的无机钙钛矿Rb1-xCsxPbI3的材料特性的第一性原理计算为例,说明通过MatCloud+材料云平台使用MatCloudQE第一性原理计算程序包可进行能隙、态密度和光学性质的计算,计算结果不仅和文献报道结果相符,且整个计算过程减少人工干预次数可达90%以上,极大地提高了效率。目前,MatCloud-QE第一性原理计算程序包深度集成在MatCloud+材料云平台,用户通过云端才能使用,该模式为用户提供了丰富的计算资源和一站式第一性原理计算服务。
Abstract:
In China, carrying out firstprinciples simulation is always limited to 1~2 commercial softwares, these softwares not only require software licenses, but also need to be downloaded and manually deployed to high-performance computing clusters, which greatly restricts the application of first-principles simulations. Based on fundamental Quantum ESPRESSO first-principles simulation package, we developed AI-driven high-throughput first-principles simulation acceleration algorithm, parameter intelligent recommendation algorithm, input structure unifying algorithm, automatic post-processing engine, component engine, graphical engine, more simulation functions, and application programming interface(API) to MatCloud+ cloud platform, forming a Cloud-native first-principles simulation software package, namely MatCloudQE. Through the MatCloud+ cloud platform, users can carry out QE firstprinciples simulation and data management on-line only through the Web browser without downloading or installing any third-party softwares. This paper takes the first-principles simulation of material properties of inorganic perovskite Rb1-xCsxPbI3 as an example, showing that the MatCloud-QE firstprinciples simulation package can be used through the MatCloud+ cloud platform to calculate energy gap, state density and optical properties. The calculation results are not only consistent with the results from literatures, but also reduce the number of manual interventions by more than 90% during the entire computing process, which greatly improves efficiency.

备注/Memo

备注/Memo:
收稿日期:2022-03-04修回日期:2022-08-24 基金项目:云南省材料基因工程专项资助项目(2019CLJY06,202002 AB080001-6);国家自然科学基金资助项目(52142306);信息功能材料国家重点实验室开放课题项目(SKL2021) 第一作者:杨小渝,男,1968年生,研究员,博士生导师, Email:kxy@cnic.cn
更新日期/Last Update: 2024-12-09