arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2203.16203cond-mat.mtrl-sciphysics.comp-phphysics.data-an
收藏

预测用于量子技术的固态材料平台

Predicting Solid State Material Platforms for Quantum Technologies

Oliver Lerstøl Hebnes, Marianne Etzelmüller Bathen, Øyvind Sigmundson Schøyen, Sebastian G. Winther Larsen, Lasse Vines, Morten Hjorth-Jensen

AI总结:

本文提出结合材料信息学与机器学习的自动发现框架,基于逾25,000种材料和近5000个特征比较三种标注方法,识别出47种量子技术候选半导体材料。

摘要:

半导体材料为量子技术(QT)提供了极具吸引力的平台,而且大量材料的性质可以在包含实验与理论探索信息的数据库中找到。然而,搜索这些数据库以寻找适用于量子技术应用的有前景候选材料是一项重大挑战。因此,我们利用材料信息学和机器学习方法,开发了一个用于自动发现QT半导体宿主平台的框架,由此形成了一个包含超过25,000种材料和近5000个物理启发特征的数据集。研究设计了三种方法,分别命名为Ferrenti、扩展Ferrenti和经验方法,用于为监督机器学习(ML)方法(逻辑回归、决策树、随机森林和梯度提升)标注数据。我们发现,在这三种方法中,完全依赖文献发现的经验方法所预测的候选材料数量明显少于另外两种方法;在比较协方差矩阵中的两个最大特征值时,该方法在合适与不合适候选材料之间表现出清晰区分。与文献预期以及聚焦带隙和离子特性的Ferrenti和扩展Ferrenti方法所得结果相反,来自经验方法的ML方法强调,与对称性和晶体结构相关的特征(包括键长、取向和径向分布)在预测一种材料是否适合QT时具有重要影响。所有三种方法和全部四种ML方法都共同认可一个包含47种合格候选材料的子集,其中包括8种元素化合物、29种二元化合物和10种三元化合物,这为进一步面向量子技术的材料探索提供了基础。

英文摘要:

Semiconductor materials provide a compelling platform for quantum technologies (QT), and the properties of a vast amount of materials can be found in databases containing information from both experimental and theoretical explorations. However, searching these databases to find promising candidate materials for quantum technology applications is a major challenge. Therefore, we have developed a framework for the automated discovery of semiconductor host platforms for QT using material informatics and machine learning methods, resulting in a dataset consisting of over $25.000$ materials and nearly $5000$ physics-informed features. Three approaches were devised, named the Ferrenti, extended Ferrenti and the empirical approach, to label data for the supervised machine learning (ML) methods logistic regression, decision trees, random forests and gradient boosting. We find that of the three, the empirical approach relying exclusively on findings from the literature predicted substantially fewer candidates than the other two approaches with a clear distinction between suitable and unsuitable candidates when comparing the two largest eigenvalues in the covariance matrix. In contrast to expectations from the literature and that found for the Ferrenti and extended Ferrenti approaches focusing on band gap and ionic character, the ML methods from the empirical approach highlighted features related to symmetry and crystal structure, including bond length, orientation and radial distribution, as influential when predicting a material as suitable for QT. All three approaches and all four ML methods agreed on a subset of $47$ eligible candidates %(to a probability of $>50 \ \%$) of $8$ elemental, $29$ binary, and $10$ tertiary compounds, and provide a basis for further material explorations towards quantum technology.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑