The Ab-Initio Prediction of Crystal Structure: Combining Data Mining Ideas with Quantum Mechanics

晶体结构的从头算预测：数据挖掘思想与量子力学的结合

• 批准号: 0606276
• 负责人: Gerbrand Ceder
• 金额: --
• 依托单位: Massachusetts Institute of Technology
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2006
• 资助国家: 美国
• 起止时间: 2006-08-01 至 2011-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0606276&HistoricalAwards=false
• 关键词: Ab; Initio; Prediction; Crystal; Structure

TECHNICAL SUMMARY:This award supports computational and theoretical research that aims to develop a rigorous formalism to capture "knowledge" from past experimental and computed data, and use it to rapidly guide accurate quantum mechanical energy or free energy methods towards the most stable structure in binary and ternary metallic alloys and oxides. In a departure from traditional computational approaches, the PI will merge ideas from data-mining to extract knowledge from the large body of existing crystal structure information with the predictive power of quantum mechanical calculations. The PI will pursue a probabilistic approach. The probability of particular structure to appear in a new alloy is expanded in terms of correlations between structures at different compositions and between structures and elements. To capture structure correlations present in nature, the PI will data mine some of the largest databases available for metallic alloys and oxide compounds and construct a maximum entropy representation of it. This will enable predictions for many alloys for which currently little or no characterization is present. The resulting structure prediction tool and all in-house generated data will be made available to the research community as a web-based structure predictor so that these new methods can be most efficiently disseminated.The new developments gained from this research, and the ab-initio database that will be created, will be integrated with the freely available (on the web) electronic course on Computational Materials Science the PI teaches. This contributes to the cyberinfrastructure of the materials research community.NON-TECHNICAL SUMMARY:This award supports computational and theoretical research that aims to develop a rigorous formalism to capture "knowledge" from past experimental and computed data, and use it to rapidly guide accurate computations that aim to predict how atoms will organize themselves in materials. The PI will focus on classes of alloys and oxide materials. Crystal structure plays a fundamental and widely applicable role in materials science. Many relevant physical properties of inorganic materials are directly tied to, and sometimes prohibited by, the underlying symmetry of the way atoms arrange themselves in a crystal. In computational materials science where one tries to predict properties of materials before they are synthesized, the prediction of structure is a key but a missing cornerstone of materials design by computer. This work contributes to efforts to develop computational methods that can predict the way atoms will arrange themselves in a crystal.This work contributes to the cyberinfrastructure of the materials research community. It involves the novel application of data mining to materials computations and the use of the resulting integration to solve complex materials problems. The successful completion of this research project will lead to an approach that can determine the stable arrangement of atoms in a material with a high confidence level, and to the creation of a database available to the public that contains the results of computations for a large number of crystal structures and alloys that can be queried by theorists, computational materials researchers, experimentalists, students, and materials educators.

技术概述：该奖项支持计算和理论研究，旨在开发一种严格的形式主义，从过去的实验和计算数据中获取“知识”，并使用它快速指导精确的量子力学能或自由能方法，以实现二元和三元金属合金和氧化物中最稳定的结构。与传统的计算方法不同，PI将融合数据挖掘的思想，从现有的大量晶体结构信息中提取知识，并利用量子力学计算的预测能力。PI将采用概率方法。根据不同成分的结构之间以及结构与元素之间的相关性，扩大了新合金中出现特定结构的可能性。为了捕捉自然界中存在的结构相关性，PI将对一些可用于金属合金和氧化物化合物的最大数据库进行数据挖掘，并构建其最大熵表示。这将使许多合金的预测，目前很少或没有表征存在。由此产生的结构预测工具和所有内部生成的数据将作为基于网络的结构预测器提供给研究界，以便这些新方法可以最有效地传播。从这项研究中获得的新进展，以及将创建的ab-initio数据库，将与PI教授的免费（在网络上）计算材料科学电子课程相结合。这有助于材料研究界的网络基础设施。非技术总结：该奖项支持计算和理论研究，旨在开发一种严格的形式主义，从过去的实验和计算数据中获取“知识”，并使用它来快速指导精确的计算，旨在预测原子如何在材料中组织自己。PI将专注于合金和氧化物材料的分类。晶体结构在材料科学中起着基础和广泛应用的作用。无机材料的许多相关物理性质与晶体中原子排列方式的潜在对称性直接相关，有时甚至被这种对称性所禁止。在计算材料科学中，人们试图在材料合成之前预测材料的性能，结构预测是计算机材料设计的关键，但却缺少基石。这项工作有助于开发计算方法来预测原子在晶体中的排列方式。这项工作有助于材料研究界的网络基础设施。它涉及数据挖掘在材料计算中的新应用，并使用由此产生的集成来解决复杂的材料问题。该研究项目的成功完成将导致一种方法，可以确定具有高可信度的材料中原子的稳定排列，并创建一个可供公众使用的数据库，其中包含大量晶体结构和合金的计算结果，可供理论家，计算材料研究人员，实验家，学生和材料教育者查询。