Chemical modification of zeolites for the recovery of rare earth elements evaluated by machine learning algorithms

Abstract

Rare earth elements (REE) are a significant group of valuable elements used in diverse and relevant applications in our daily lives. The mining and processing of the original ores, as well as the final wastes disposal, produce wastewater with variable concentrations of REE to be recovered. Sorbent materials like zeolites have been employed in adsorption processes to capture diverse pollutants from wastewater. The objective of this study is to identify the most effective modified zeolite for the adsorption and desorption of REE from aqueous solutions. To achieve this, the processes evaluation by machine learning (ML) algorithms was explored through both supervised and unsupervised analyses. The purpose of the usage of such tools was to assist in the selection of the optimal zeolite for REE recovery and to assess the predictive capabilities of the models. Modified zeolites were obtained by acid and alkali treatments in order to increase their sorption capacity compared to the controls and they were characterized by SEM/EDS, FTIR and pH zero point charge. Kinetic modelling and desorption assays were also performed, these last ones to evaluate the REE leaching from the sorbent for the best suitable modified zeolites. An overall removal of 80% for adsorption and over 90% recovery for desorption were achieved for the best modified zeolites. ML algorithms helped to classify the adsorption results and allowed the selection of the best suitable modified zeolites. It is concluded that alkali modification of the zeolites surfaces increases their natural adsorption capacity for recovering REE.