A Machine Learning-Based Study of Li+ and Na+ Metal Complexation with Phosphoryl-Containing Ligands for the Selective Extraction of Li+ from Brine

Abstract

The growth of technologies concerned with the high demand in lithium (Li) sources dictates the need for technological solutions garnering Li supplies to preserve the sustainability of the processes. The aim of this study was to use a machine learning-based search for phosphoryl-containing podandic ligands, potentially selective for lithium extraction from brine. Based on the experimental data available on the stability constant values of phosphoryl-containing organic ligands with Li+ and Na+ cations at 4:1 THF:CHCl3, candidate di-podandic ligands were proposed, for which the stability constant values (logK) with Li+ and Na+ as well as the corresponding selectivity values were evaluated using machine learning methods (ML). The modelling showed a reasonable predictive performance with the following statistical parameters: the determination coefficient R2= 0.75, 0.87 and 0.83 and root-mean-square error RMSE = 0.485, 0.449 and 0.32 were obtained for the prediction of the stability constant values with Li+ and Na+ cations and Li+/Na+ selectivity values, respectively. This ML-based analysis was complemented by the preliminary estimation of the host–guest complementarity of metal–ligand 1:1 complexes using the HostDesigner software.