Using machine learning to guide the synthesis of supported palladium catalysts with desired palladium dispersion

Abstract

Supported palladium catalysts are indispensable in a wide range of industries, including petrochemicals, pharmaceuticals, and the automotive sector. The dispersion of palladium within these catalysts, primarily determined by the average nanoparticle size, significantly influences both the catalytic properties and the utilization efficiency of palladium. This study explores the relationships between various catalyst synthesis parameters and the resulting Pd nanoparticle size/dispersion. We developed a machine learning (ML) model to guide future synthesis efforts aimed at achieving specific palladium dispersion levels. Data were collected from previous studies on supported Pd catalysts published between 2000 and 2023, encompassing 1543 distinct catalysts. Of these, 1295 data points were used to construct the ML model. Key synthesis parameters-such as synthesis method, metal loading, support type, support surface area, metal precursor, solvent, solvent pH, support's point of zero charge, and calcination/reduction conditions-were identified as independent variables, while dispersion and average Pd nanoparticle size served as dependent variables. A random forest (RF) regression model was employed to predict dispersion (in %), validated through 5-fold cross-validation. The model achieved root mean squared errors (RMSE) of 9.5 (training) and 14.9 (testing) in Pd dispersion (in %) prediction. Experimental synthesis of new supported palladium catalysts using different synthesis parameters confirmed the model's predictions, yielding an RMSE of 5.4. Additionally, data from the literature published in 2024 were also used to validate the model, the comparison resulted in an RMSE of 5.9. This ML approach offers significant potential for precisely controlling palladium dispersion during catalyst synthesis, moving beyond traditional trial-and-error methods. It holds a broad potential to significantly improve palladium utilization across a variety of industrial applications.