Machine learning assisted layer-controlled synthesis of MoS2

Abstract

Two-dimensional (2D) transition metal dichalcogenides (TMDs) have attracted significant interest due to their intriguing physical, chemical, electronic and optical properties. However, the practical applications of TMDs are limited by challenges related to controlling the thickness of atomic layers. Machine learning (ML), a data-driven approach characterized by extensive search capabilities and accurate classification, offers a promising approach to address this limitation. In this study, a prediction model was constructed using four machine learning algorithms, namely XGBoost, Support Vector Machine (SVM), Naïve Bayes (NB), and Multilayer Perceptron (MLP), to explore the growth mechanism of MoS₂ material layers prepared through chemical vapor deposition (CVD). Furthermore, the models were evaluated using performance assessment metrics such as recall, specificity, accuracy, and ROC curve. The results showed that the MLP model had the highest prediction accuracy, up to 75%, and an AUC of 0.8. The XGBoost model was used to extract the feature importance of growth parameters, revealing that the temperature of the precursor molybdenum source (Mo_T), reaction temperature (T), and reaction time (t) were the main factors affecting the growth of MoS₂ layers. Finally, we use virtual data to predict the results and delineate the range of each growth condition, with 50% of predicted results as the dividing line. The optimization of growth conditions through machine learning algorithms holds promise for enhancing control over the preparation of MoS₂ layers, thereby facilitating the development of electronic and optoelectronic devices.