【作者】 梁敏；

【导师】 谢文辉；

【作者基本信息】 华东师范大学 ， 凝聚态物理， 2024， 硕士

【摘要】 在自旋电子学领域,寻找具有远高于室温的居里温度并且能够实现完全自旋极化的材料一直是重要研究内容。而惠斯勒合金由于其种类中,性质丰富,在自旋电子学材料研究中一直受到关注。本文选择双半惠斯勒合金作为研究对象,运用机器学习方法辅助的高通量计算对其进行计算研究。本文工作首先从OQMD收集到203种材料用于机器学习模型训练。研究中采用极端决策树、梯度决策树等五个机器学习,利用回归模型的评价指标对模型性能进行了评估。结果显示在预测稳定性方面,极端决策树模型的表现最佳;而在预测形成能方面,梯度决策树模型的表现最好。基于上述机器学习模型,本文对包含3d过渡金属组成的5040种双半惠斯勒合金化合物进行分析,预测并筛选出3681种热力学稳定的材料。然后再采用基于密度泛函理论的第一性原理计算,对上述材料进行了高通量计算研究,通过研究其电子结构后找到了98种半金属材料。通过应用平均场理论对居里温度进行了估算,并发现了69种具有接近或高于室温的磁性材料。在对这些材料的动力学稳定性进行研究后,最终筛选出20种具有高居里温度的磁性半金属材料。上述研究表明,在现有数据库基础上,通过机器学习模型辅助的高通量第一原理计算方法,可以在双半惠斯勒合金中找到具有热力学和动力学稳定性、高自旋极化以及高居里温度的一些材料,这为进一步的研究提供了帮助。更多还原

【Abstract】 In the field of spintronics,the search for materials with Curie temperatures significantly higher than room temperature and capable of achieving complete spin polarization has been a focal point of research.Heusler alloys,due to their diverse properties among various compositions,have consistently garnered attention in spintronics material research.This study focuses on double half-Heusler alloys as the research object,utilizing machine learning-assisted high-throughput calculations for computational research.The work begins with the collection of 203 materials from OQMD for training machine learning models.Five machine learning algorithms,including extreme gradient boosting and gradient boosting decision tree,were employed,and the performance of the models was evaluated using regression metrics.The extreme gradient boosting model exhibited the best performance in predicting stability,while the gradient boosting decision tree model also showed excellent performance in predicting formation energies.Based on these machine learning models,an analysis was conducted on 5040 double Half-Heusler alloy compounds composed of 3d transition metals,predicting and screening out 3681 thermodynamically stable materials.Subsequently,high-throughput computational research was carried out using first-principles calculation software based on density functional theory.By studying their electronic structures,98 half-metallic materials were identified.Applying meanfield theory to estimate the Curie temperatures,69 magnetic materials with properties close to or above room temperature were discovered.After investigating the dynamic stability of these materials,20 magnetic half-metallic materials with high Curie temperatures and stability were ultimately selected.The research demonstrates that with the help of machine learning models and highthroughput first-principles calculations based on existing databases,materials with thermodynamic and dynamic stability,high spin polarization,and high Curie temperatures can be identified in double Half-Heusler alloys,providing valuable insights for further research.更多还原