计算物理 ›› 2017, Vol. 34 ›› Issue (6): 697-704.

• • 上一篇    下一篇

支撑材料基因工程的高通量材料集成计算平台

杨小渝1,2, 王娟1,2, 任杰1,2, 宋健龙1, 王宗国1, 曾雉3, 张小丽3, 黄孙超3, 张平4, 林海青5   

  1. 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. 中国科学院固体物理研究所, 合肥 230031;
    4. 北京应用物理与计算数学研究所, 北京 100088;
    5. 北京计算科学研究中心, 北京 100193
  • 收稿日期:2016-08-10 修回日期:2017-01-23 出版日期:2017-11-25 发布日期:2017-11-25
  • 作者简介:杨小渝,男,博士,研究员,中科院"百人计划",从事材料信息学,材料基因工程信息化技术研究,E-mail:kxy@cnic.cn
  • 基金资助:
    国家重点研发计划(2016YFB0700501),国家自然科学基金面上项目(61472394)及国家自然科学基金重点项目(11534012)资助

An Integrated High-throughput Computational Material Platform

YANG Xiaoyu1,2, WANG Juan1,2, REN Jie1,2, SONG Jianlong1, WANG Zongguo1, ZENG Zhi3, ZHANG Xiaoli3, HUANG Sunchao3, ZHANG Ping4, LIN Haiqing5   

  1. 1. Computer Network Information Center, Chinese Academy of Sciences, Beijing 100190, China;
    2. University of Chinese Academy of Sciences, Beijing 100049, China;
    3. Institute of Solid State Physics, Chinese Academy of Sciences, Hefei 230031, China;
    4. Institute of Applied Physics and Computational Mathematics, Beijing 100088, China;
    5. Beijing Computational Science Research Center, Beijing 100193, China
  • Received:2016-08-10 Revised:2017-01-23 Online:2017-11-25 Published:2017-11-25

摘要: 总结材料计算和数据管理面临的挑战,阐述高通量材料集成计算的内涵,介绍我们在国内率先研制出的高通量材料集成计算平台和软件框架MatCloud.平台支持与不同高性能计算集群的动态绑定,基于图形化界面的第一性原理计算流程设计、持久化、解析和执行;大批量计算作业的生成、提交、运行监控,及容错和纠错的自动流程实现;支持数据的提取、规范化加工、及自动存储等.MatCloud形成了一个高通量材料集成计算软件框架,支持个性化插件的快速研发.

关键词: 材料基因组计划, 材料信息学, 高通量材料计算, 材料基因组工程, MatCloud

Abstract: The core philosophy of Material Genome Initiative is transition of way of new material design from traditional "try-and-error" approach to in-silico material design approach where intensive computing and material informatics are employed. It aims to effectively speed up discovery, development, production and deployment of new material two times faster as it is now. It means a culture shift of new material discovery:simulation and prediction first, followed by experiment. An integrated computational material platform that can facilitate high-throughput quantum mechanical simulations and manage simulation lifecycle data is therefore vital. This paper depicts a high throughput computational material platform and software framework, namely, MatCloud, which effectively integrates individual quantum mechanical simulation tasks, data extraction and data storage into an automatic flow in an end-to-end manner without direct human control. Especially, core data curation activities are also integrated into this flow rather than happening at post-simulation stage separately. MatCloud is demonstrated in an example of disorder binary alloy design to be valid and effective.

Key words: Material Genome Initiative, material informatics, high-throughput simulation, in-Silico material design, MatCloud

中图分类号: