Enhanced Machine Learning Molecular Simulations for optimization of flotation selectivity: A perspective paper

Abstract

The recovery of valuable minerals in froth flotation industry relies on finding inexpensive and environmentally friendly reagents that selectively adsorb upon surfaces and interfaces. Computer simulations, especially when extended, provide access to detailed mechanistic information on solvent configurations and may ascertain crucial dynamical events over the adsorption process. Further, in silico throughput screening can prevent both the high cost of experiments and the related risks to the environment. Yet, a better compromise between accuracy and computational cost must be met. Machine learning (ML) simulations may ease the latter and suggest solidophilic reagents able to improve the flotation efficiency, shedding new light on discerning descriptors able to accurately capture the nature of the molecule-surface interaction. In this work, our recent advancements in modeling of new accurate mineral-water interfaces based on active learning of ab initio molecular dynamics trajectories have been introduced. The case of some habitual oxides and minerals liberated in mining industry will be taken as examples.