Enhancing Photovoltaic Parameters through Anti-Reflective Coatings: A DFT and Machine Learning-Based Study

Abhinav Thakur, Manasvi Raj, Shubham Kumar, Miran Mehta, Megha Singh, Adarsh Jain, Neeraj Goel,
Enhancing Photovoltaic Parameters through Anti-Reflective Coatings: A DFT and Machine Learning-Based Study,
Journal of Physics and Chemistry of Solids,
2025,
113320,
ISSN 0022-3697,
https://doi.org/10.1016/j.jpcs.2025.113320.
(https://www.sciencedirect.com/science/article/pii/S0022369725007735)
Abstract: ABSTRACT
Improving light absorption in solar cells remains a critical challenge in advancing photovoltaic (PV) efficiency. This study explores the role of anti-reflective coatings (ARCs) in enhancing solar cell performance through a dual approach combining first-principles simulations and machine learning (ML) analytics. Using Density Functional Theory (DFT), we investigated the optical, electronic, and vibrational properties of five ARC materials—SiO2, Al2O3, HfO2, ZnS, and MgF2. Our simulations reveal that SiO2 exhibits the most favorable characteristics, including low reflectivity, stable refractive index, a wide electronic bandgap, and dynamic vibrational stability. To assess real-world performance, an ML model was trained on experimental and simulated solar cell data of 4500 data point to predict the impact of ARC application on key PV parameters. The results showed significant improvements in short-circuit current (Jsc), open-circuit voltage (Voc), fill factor (FF), and power conversion efficiency (PCE) with 19.6%, 30.4%, and 14.9% increase in PCE for lead-based, cesium-based perovskite and non-perovskite solar cell respectively. ML model achieves a strong predictive correlation (R2 = 0.85) with Root Mean Square Error (RMSE) of 0.021 and Mean Absolute Error (MAE) of 0.017. This integrated computational-simulated framework highlights SiO2 as a robust, high-performance ARC candidate. Results shows that SiO2 has ultra-low reflectivity (<0.04) and a nearly constant real refractive index (∼1.45) which makes it as a promising candidate for ARC. The novelty of this work lies in the dual methodology, where DFT insights are paired with an ML model trained on a combination of simulated and real-world data to predict improvements in solar cell parameters.
Keywords: ARCs; Anti-Reflective Coatings; DFT; Density Functional Theory; SiO2; Silicon Dioxide; ML; Machine Learning