npj Computational Materials

Element selection for functional materials discovery by integrated machine learning of elemental contributions to properties

Andrij Vasylenko, Dmytro Antypov, Vladimir V. Gusev, Michael W. Gaultois, Matthew S. Dyer & Matthew J. Rosseinsky

npj Computational Materials 9: 164 (2023)

doi.org/10.1038/s41524-023-01072-x

Published online: 04 September 2023
Abstract| Full Text | PDF OPEN

Abstract: The unique nature of constituent chemical elements gives rise to fundamental differences in materials. Assessing materials based on their phase fields, defined as sets of constituent elements, before specific differences emerge due to composition and structure can reduce combinatorial complexity and accelerate screening, exploiting the distinction from composition-level approaches. Discrimination and evaluation of novelty of materials classes align with the experimental challenge of identifying new areas of chemistry. To address this, we present PhaseSelect, an end-to-end machine learning model that combines representation, classification, regression and novelty ranking of phase fields. PhaseSelect leverages elemental characteristics derived from computational and experimental materials data and employs attention mechanisms to reflect the individual element contributions when evaluating functional performance of phase fields. We demonstrate this approach for high-temperature superconductivity, high-temperature magnetism, and targeted bandgap energy applications, showcasing its versatility and potential for accelerating materials exploration.

摘要: 组成化学元素的独特性质导致了材料的根本差异。由于组分和结构而出现特定差异之前，基于它们的相场（定义为组成元素的集合）评估材料，能够降低组合复杂性并加速筛选，利用与组成水平方法的区别。对材料类别新颖性的区分和评估与识别化学新领域的实验挑战相一致。为了解决这个问题，我们提出了PhaseSelect，一个端到端机器学习模型，它结合了相场的表示、分类、回归和新颖性排序。PhaseSelect利用从计算和实验材料数据中获得的元素特征，并采用注意机制来反映在评估相场的功能性能时各个元素的贡献。我们展示了这种方法在高温超导、高温磁性和目标带隙能量方面的应用，展示了其多功能性和加速材料探索的潜力。

Editorial Summary

From Elements to Materials: Machine Learning Unveils New Magic

The properties of materials are dependent on their composition and structure, making the selection of appropriate elemental combinations crucial for designing and synthesizing new materials. However, due to the diversity of elements and potential combinations, this selection process becomes complex and time-consuming. Research on how to efficiently evaluate the performance of materials at the elemental level based on phase fields is of significant importance, aiming to reduce combinatorial complexity and accelerate the development of functional materials. A team led by Prof. Matthew J. Rosseinsky from the Department of Chemistry. University of Liverpool, United Kingdom, proposed the PhaseSelect method through combining the concepts of elements and phase fields for optimizing elemental selection in the early stages of materials discovery. This method represents an end-to-end machine learning model that integrates the representation, classification, regression, and novelty ranking of phase fields. PhaseSelect utilizes elemental features obtained from computational and experimental materials data and employs an attention mechanism to reflect the contributions of individual elements when evaluating the functional performance of phase fields. The study validates the versatility and potential for accelerating materials exploration by applying PhaseSelect to three regression tasks involving superconducting transition temperature, Curie temperature, and bandgap energy, as well as three classification tasks related to material performance based on specific property thresholds. By combining binary classification with regression, PhaseSelect initially assign phase fields to low- and high-performing material classes and subsequently predicts the maximum expected values of the targeted properties. This innovative approach provides reliable quantitative metrics for the rapid discrimination and screening of materials phase fields at scale. In summary, this study introduces a novel method for optimizing elemental selection in the early stages of materials discovery and would accelerate the discovery of functional materials.

材料背后的元素选择之道：机器学习探索奇妙世界

材料的特性取决于其组成和结构，选择合适的元素组合对于设计和合成新材料至关重要。然而，考虑到元素的多样性以及可能的组合方式，这种选择过程变得复杂且耗时。研究如何基于相场在元素层面高效评估材料的性能，以降低组合复杂性从而加速功能材料的研发具有重要意义。该研究提出了一种集成机器学习新方法，用于在材料发现的早期阶段进行优化元素选择。来自英国利物浦大学化学系的Matthew J. Rosseinsky研究团队，将元素和相场的概念结合，提出了PhaseSelect方法。该方法是一个端到端机器学习模型，集成了相场的表示、分类、回归和新颖性排序。PhaseSelect利用从计算和实验材料数据中获得的元素特征，并采用注意机制来反映在评估相场的功能性能时各个元素的贡献。他们在超导转变温度、居里温度和带隙能量的三个回归任务以及根据性能阈值进行材料分类的三个分类任务中，验证了PhaseSelect的多功能性和加速材料探索的潜力。该方法通过将二元分类与回归相结合，先将相场分配到材料的低性能类和高性能类，然后预测感兴趣的性能的最大期望值，为大规模快速区分和筛选材料相场提供可靠的定量指标。该研究提出了一种在材料发现早期阶段进行优化元素选择的新方法，将加速功能材料的研发。