A high throughput screening model of solidophilic flotation reagents for chalcopyrite based on quantum chemistry calculations and machine learning

Flotation reagents are critical to realizing selective separation of different minerals in the flotation process. The current “trial and error” strategy for screening effective flotation reagents is time-consuming and inefficient. Herein, a combined machine learning (ML) + quantum chemistry (QC) model has been proposed to accelerate the screening of solidophilic flotation reagents. The accurate QC features of an ethyl-functional group (EFG) set involving 47 molecules have been obtained and collected as a database to describe their bonding reactions with the surface Cu(II), Fe(II), and Cu(I) ions at the B3LYP/def2-TZVP level under solvation effects. 15 QC feature descriptors and 4 ML algorithms have been adopted to establish the high throughput ML screening. QC results show the affinity of EFG molecule with Cu(II) is the strongest, followed by Fe(II), and the weakest is Cu(I). ML results show that the gradient boosting regression can successfully predict these molecules with the highest selective bonding index. The atom type, frontier molecular orbital, molecule charge, and dipole moment have significant effects on the bonding interactions. ML has shown an extremely lower time cost than the QC-based models. This work sheds new light on the development and discovery of efficient, selective, and green flotation reagents by accurate and low-cost artificial intelligence-based computational methods.