Machine Learning-Based Modeling and Multi-Objective Optimization of Magnetron-Sputtered Platinum Coatings

Platinum coatings produced by magnetron sputtering are highly valued due to their exceptional properties, including excellent electrical conductivity, high catalytic activity, and superior corrosion resistance. The quality of these coatings, however, is strongly dependent on the sputtering parameters. This study performs optimization of platinum thin film deposition on stainless steel substrates by systematically varying magnetron sputtering parameters. Experimental data were obtained under different conditions of discharge current, pressure, and deposition time. The results were analyzed using both classical regression techniques and advanced machine learning approaches to assess the influence of process parameters on deposition rate and coating thickness. Among the tested models, Gaussian Process Regression (GPR) demonstrated the highest accuracy and stability. The findings indicate that deposition time is the dominant factor influencing coating thickness, while discharge current primarily governs the deposition rate. Furthermore, multi-objective optimization and active learning approaches highlighted the potential of combining artificial intelligence methods with experimental design to reduce the number of required trials and improve process efficiency.