Guided experiment on design of Si solar cells assisted by ZnO nanostructures using machine learning technique

Abstract This study presents an investigation on efficiency enhancement of Si solar cells by adding ZnO nanostructures using Machine Learning. It employs characteristics of obtained x-ray diffraction (XRD) patterns and photoluminescence (PL) emission spectra as predictive variables for a regression model, with the efficiency enhancement of the solar cell as the target. The model is trained using three different sets of features: XRD parameters, PL characteristics, and a combination of both (plus ZnO droplets named Drop feature herein). The Gradient Boosting Regressor (GBR) model demonstrated strong predictive capabilities, achieving a high coefficient of determination (R 2 ) of around 0.90 and a low Root Mean Square Error (RMSE) of approximately 0.03 on unseen data.Our analysis, using SHAP values, clearly identified the quantity of ZnO nanostructures applied (referred to as ‘Drop’) as the most significant factor, demonstrating the largest importance. Furthermore, specific structural features from XRD, such as the properties of certain peaks (e.g., Peak 1 and Peak 3), and optical properties from PL, like the Peak Position Peak 1 PL and Band gap (eV), were also found to play important roles in the relative efficiency of Si solar cells.This methodology not only shed light on the experimental procedure and the most contributing factor but also paves the way for upcoming advancements in the field of data-driven solar cell design and fabrication.