Machine Learning Prediction of Defect Structures in Amorphous Silicon Dioxide

Abstract

Defects in insulators can have a highly detrimental impact on the performance of semiconductor devices. The study of defect formation in these amorphous insulating materials is a computationally challenging task, due to the relatively large model sizes required and their stochastic nature. Here, we propose a novel machine learning framework to predict the formation and structure of defects in amorphous materials. Our approach aims at significantly reducing the computational costs, while maintaining a high level of accuracy. We present the results of applying our workflow to the formation of hydroxyl <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$E^{\prime}$</tex> center defects in amorphous silicon dioxide, which have recently been suggested to be responsible for random telegraph and 1/f noise, as well as the bias temperature instability. The process of predicting a particular defect structure is studied in full-detail and statistical results are presented for a testing data-set.