Collaborative Research - CDMR: Informatics Guided Data Driven Computational Design of Multifunctional Materials

技术摘要本项目将建立一种新的材料模拟和建模方法，将信息学方法与密度泛函理论(DFT)计算联系起来。具有高居里转变温度的钙钛矿新化学的发现是这项研究的模板；尽管正在开发的计算方法对任何一类材料问题都是通用的。该方法利用和开发统计学习和图论方法，以发现已知材料中的行为模式，这些模式可用于生成识别新化合物中结构-性质关系的设计规则；并将用于指导和针对具有所需性质的特定化学成分进行DFT计算。具体地说，该项目正在开发一种以信息学为基础的基于规则的设计方法，通过探索巨大的化学空间来确定高温压电钙钛矿的新化学成分，这些空间如此之大，否则仅靠实验和/或计算方法就无法进行研究。这项工作的智力影响将是改变计算材料科学的传统使用，从建立大型数据库存储库，其中信息学主要用作化合物的搜索引擎，转变为将数据科学方法用作学习引擎，以发现新物理，并为对原本仍未确定的有前途的化学物质进行详细计算提出新的方向。这项工作的更广泛影响是建立一个被称为“材料数据铸造”的数据科学网络基础设施(CI)，其中将包括可通过安全的网络基础设施平台广泛访问的模型、材料元数据和模拟能力。该项目还将涉及对研究生和本科生的多机构、多学科培训，对高中生和教师进行外联，并将代表性不足的群体纳入教育和外联工作。非技术摘要原子尺度材料建模领域使用基于量子力学的复杂理论，称为密度泛函理论或DFT，这些理论在计算高效的软件中实现，尽管如此，计算密集。该项目利用了一种名为材料信息学的新兴方法，该方法能够通过探索数据的统计性质来检测和提取结构与性质的相关性，而不必利用先前存在的理论公式。在这个项目中，材料信息学被用来为高温电子器件确定具有最佳性能的新材料，而不必对每一种可能的材料结构和成分进行计算。开发的工具将通过一个网络基础设施平台广泛提供，我们称之为“材料数据铸造”。这个门户网站还将促进爱荷华州立大学(ISU)和佛罗里达大学(UF)之间的联合在线课程，学生们在课程中使用项目中开发的工具来分析真实数据和问题。在密歇根大学，外展将通过密歇根大学学生科学培训计划面向高中生，而在ISU，我们将利用爱荷华州EPSCoR计划，进入由社区和部落学院组成的全国网络。我们的计划是利用材料数据铸造厂的教育能力与教师联系，他们将在这个项目的过程中接触到数百名学生。

Technical Abstract This project will establish a new approach for materials simulation and modeling that links informatics methods to Density Functional Theory (DFT) calculations. The discovery of new chemistries of perovskites that possess high Curie transition temperatures are the template for this study; although the computational methodology that is being developed is generic to any class of materials problems. The approach exploits and develops statistical learning and graph theoretic methodologies to discover patterns of behavior among known materials that can be used to generate design rules for identifying structure - property relationships in new compounds; and will serve to guide and target DFT calculations to specific chemistries with the desired properties. Specifically, the project is developing a rule-based design approach, based on informatics, to identify new chemistries of high temperature piezoelectric perovskites by exploring vast chemical space, so large, that it would otherwise have been prohibitive to study by solely experimental and/or computational methods. The intellectual impact of this work will be to transform the traditional use of computational materials science from one of building large data repositories where informatics is primarily used as a search engine for compounds, to one where data science methods are used as a learning engine to uncover new physics and suggest new directions for conducting detailed computations on promising chemistries that would otherwise remain unidentified. The broader impacts of this work is to build a data science cyberinfrastructure (CI) that called a "Materials Data Foundry" that will include models, materials metadata, and simulation capabilities that will be broadly accessible through a secure cyber infrastructure platform. The project will also involve the multi-institutional, multi-disciplinary training of graduate and undergraduate students, outreach to high-school students and teachers, and inclusion of underrepresented groups in both education and outreach efforts. Non-Technical Abstract The field of atomistic scale materials modeling makes use of sophisticated theories based on quantum mechanics, called Density Functional Theory or DFT for short, that are implemented in computationally efficient software that are, nonetheless, computationally intensive. This project makes use of a new and emerging methodology called Materials Informatics that is able to detect and extract structure-property correlations by exploring the statistical nature of data without necessarily utilizing pre-existing theoretical formulas. In this project materials informatics is used to identify new materials with optimum properties for high-temperature electronic devices without having to carry out calculations for every possible material structure and composition. The tools developed will be made broadly available through a cyberinfrastructure platform, we term a "Materials Data Foundry". This web portal will also facilitate joint online courses between Iowa State University (ISU) and the University of Florida (UF), where the students use the tools developed in the project to analyze real data and problems. At UF, outreach will be focused on high school students through the UF Student Science Training Program while at ISU we will leverage the Iowa EPSCoR program to have access to a national network of community and Tribal Colleges. Our plan is use the educational capabilities of the Materials Data Foundry to connect with teachers who in turn will reach hundreds of students over the course of this project.