论文标题
通过基于物理的机器学习
First-Principles Prediction of Electronic Transport in Experimental Semiconductor Heterostructures via Physics-Based Machine Learning
论文作者
论文摘要
近年来，电子运输属性预测的第一原理技术已经取得了迅速的进步。但是，建模由制造过程引起的可变性的异质结构模型仍然是一个挑战。基于机器学习（ML）的材料信息学方法（MI）越来越多地用于加速具有目标特性的新材料的设计和发现，并将第一原理技术的适用性扩展到较大的系统。但是，很少有研究利用MI学习电子结构特性，并使用知识来预测各自的运输系数。在这项工作中，我们提出了一个电子传输信息学（ETI）框架，该框架在小型系统的初始模型上训练，并预测硅/葡萄片异质结构的热电器，而不是长度尺度，可与第一原则技术访问，并匹配衡量数据。我们证明了MI的应用来提取重要的物理学，这些物理学决定了在半导体异质结构中电子传输，从而从特别针对热电材料的组合策略中脱颖而出。我们预计ETI将对各种材料类别具有广泛的适用性。
First-principles techniques for electronic transport property prediction have seen rapid progress in recent years. However, it remains a challenge to model heterostructures incorporating variability due to fabrication processes. Machine-learning (ML)-based materials informatics approaches (MI) are increasingly used to accelerate design and discovery of new materials with targeted properties, and extend the applicability of first-principles techniques to larger systems. However, few studies exploited MI to learn electronic structure properties and use the knowledge to predict the respective transport coefficients. In this work, we propose an electronic-transport-informatics (ETI) framework that trains on ab initio models of small systems and predicts thermopower of silicon/germanium heterostructures beyond the length-scale accessible with first-principles techniques, matching measured data. We demonstrate application of MI to extract important physics that determines electronic transport in semiconductor heterostructures, breaking from combinatorial strategies pursued especially for thermoelectric materials. We anticipate that ETI would have broad applicability to diverse materials classes.
