Enhanced prediction and optimization of thin metal film optical properties using optimized ensemble learning models

Area of Science:

• Materials Science
• Optical Engineering
• Artificial Intelligence

Background:

• Thin metal films are critical for advanced technologies like sensors, optoelectronics, and photovoltaics.
• Optimizing optical properties (transmittance, reflectance, absorptance) of these films is challenging due to complex relationships between composition, thickness, and fabrication.
• Accurate prediction of optical characteristics is essential for efficient material design and device performance.

Purpose of the Study:

• To develop and validate an AI-driven framework for the simultaneous prediction and optimization of thin metal film optical properties.
• To explore the effectiveness of various machine learning models, including ensemble methods and multitask learning, for this task.
• To provide a user-friendly, deployable tool for real-time optical property prediction and material characterization.

Main Methods:

• A dataset of 1320 experimentally measured samples of gold, aluminum, nickel, tin, copper, and molybdenum films (200-2000 nm wavelength) was utilized.
• Ensemble models (Random Forest, Gradient Boosting, XGBoost, Extra Trees) and a multitask learning model were trained and optimized using GridSearchCV with stratified K-fold cross-validation.
• Feature importance analysis was conducted, and the best-performing model was deployed via a web-based GUI.

Main Results:

• The CatBoost Regressor model exhibited superior performance, achieving high accuracy with R² = 0.99928, MAE = 0.21924, and MSE = 0.28203 on average across all optical properties.
• Feature importance analysis indicated that Material Type had a slightly greater predictive influence than Wavelength.
• Material-specific error analysis identified challenging prediction zones at spectral extremes, guiding further refinement.

Conclusions:

• The AI-driven framework offers a scalable, interpretable, and deployable solution for AI-assisted material design and optical characterization of thin metal films.
• This data-driven approach significantly accelerates the optimization process by enabling rapid material property identification and enhancement.
• The developed tool facilitates efficient exploration of the design space for thin metal films, benefiting sensor, optoelectronic, and photovoltaic applications.