Conformal active learning-aided screening of ligand-protected Cu-nanoclusters for CO2 reduction reactions

In this study, we propose a conformal active learning (CAL) method to screen ligand-protected atomically precise Cu-nanoclusters for the CO2 reduction reaction. We investigate the roles of core metals and protecting ligands in product selectivity. We demonstrate a unique machine learning model that accurately accelerates the screening of nanoclusters, calculating prediction uncertainty for unknown datasets with a rigorous coverage guarantee. This approach results in reduced error and uncertainty (mean prediction interval) in the overall dataset predictions, effectively balancing uncertainty, coverage guarantee, and prediction error for all the considered descriptors important for a catalytic reaction. Furthermore, through feature importance analysis, we determine the major influences from the considered features. By applying optimal criteria to the predicted results of each descriptor, we can identify the best selective catalysts for C1 product formation. This CAL-based method will certainly open a new direction in the field of high-throughput screening of catalysts.