Electronic Bandgap Prediction of Functionalized MX2‐based Materials Using Machine Learning‐aided and First‐principles Studies

Abstract Owing to the increasing interest and inherent properties, two‐dimensional (2D) transition metal dichalcogenides (TMDs) emerged as suitable candidates for diverse material applications. The precise characterization of a material toward a desired property is often governed by its bandgap. The accurate prediction of bandgap remains critical due to the limitation of conventional approaches, where experiment demands a lot of resources and the density functional theory faces constraints due to the in‐built estimation errors. To address this, the present study leverages on developing a robust machine learning model, harnessing both structural and compositional features to predict the accurate bandgap of 2D TMDs. The advancement in the accuracy of TMDs bandgap is reinforced by the selection of a suitable database screened out via hierarchical down selection methodology, incorporating bandgaps from both experimental and hybrid functional, which significantly improves the prediction capabilities. This is evidenced by the exceptional statistical regression metrics observed for the model across the training and test datasets. Moreover, to validate the predictive capability and versatility of the developed model, we predicted the bandgap of six materials that are unknown to the learning algorithm. Notably, the predicted bandgaps are found to be very close to the previously reported experimental as well as hybrid functional values. These findings underscore the adeptness of the model to accurately predict the bandgap, making it applicable for the fine‐tuning of a broad spectrum of 2D TMDs toward their desired applications.