基于SISSO和机器学习方法的钙钛矿结构的稳定性预测:新型容许因子建立与验证

New tolerance factor based on SISSO and machine learning for predicting stability of perovskite structure

由于钙钛矿型材料具有广泛的应用前景,因此对其结构及物理、化学性质的研究一直是材料研究领域的热点之一。其中,利用容许因子(Tolerance factor)来预测钙钛矿型材料的结构稳定性可以帮助研究者发现更多的新型功能材料,而传统的基于离子半径定义的容许因子tIR存在一定的局限性。本文基于SISSO(Sure independence screening and sparsifying operator)方法和键价模型提出一种新型的容许因子τBV,其可以有效地避免由离子半径带来的局限性。本工作使用机器学习中的决策树算法建立容许因子验证模型,实验结果表明,新型容许因子τBV可以很好地预测ABO3型化合物是否具有钙钛矿结构,并大大提高了预测精度。

Due to the wide application prospects of perovskite materials,research on their structures and physical and chemical properties of perovskite materials has been one of the hot topics in the field of materials research.Among them,predicting the stability of perovskite structure with the help of tolerance factor can help researchers discover more new functional materials.The conventional tolerance factor tIR for determining the stability of the perovskite structure based on ion radius has certain shortcomings and limitation.In view of this,this work proposes a new type of tolerance factorτBV based on the bond valence model using the SISSO(sure independence screening and sparsifying operator)method which can effectively avoid the defect limitation caused by the ionic radius.This work uses the decision tree algorithm in machine learning to establish the new tolerance factor verification model and the results show that the new tolerance factorτBV can excellently predict whether the ABO3 compound is perovskite or non-perovskite,which greatly improves the prediction accuracy.