(520e) Molecular Engineering and Machine Learning Guided Design for Ion-Ion Separation
AIChE Annual Meeting
2023
2023 AIChE Annual Meeting
Separations Division
Membranes Designed for Ion-Ion Separations I
Wednesday, November 8, 2023 - 1:54pm to 2:15pm
Ion-ion separation has profound practical interests in many different areas, such as extraction of lithium from salt lake, water demineralisation, and recovery of resources from the chlor-alkali process. Nanofiltration membranes with the capability to precisely differentiate ions are one key approach to realising the target. In this presentation, I'll introduce our recent work on the molecular engineering and machine learning studies of nano filtration membranes for ion-ion separation.
First, by spatial and temporal control of interfacial polymerization via graphitic carbon nitride (g-C3N4), we were able to construct a nanofiltration membrane with nanoscale ordered structure, low thickness and high charge density. Therefore, fast permeability, high rejection, and precise Cl-/SO42- separation were achieved.
Second, inspired by the structure and function of arginine, we fabricated positively charged NF membranes using triaminoguanidine (TG) as an aqueous monomer through interfacial polymerization. The optimal membrane exhibits the ever-reported highest Li+/Mg2+ selectivity. Separation factors (SLi,Mg) arrive at 83.0, 46.4 and 36.4 at Li+/Mg2+ mass ratios ranging from 1:20, 1:50 to 1:120 at pH=5.6. Moreover, exceptional Li+/Mg2+ selectivity were maintained at alkaline conditions, i.e., pH 7.6 and 8.6.
Third, we applied machine learning to guide the fabrication of nanofiltration (NF) membranes with high solute-solute selectivity. By gradually increasing the experimental dataset, the machine learning models kept improving the prediction accuracy for selectivity. At the dataset size of 114, the models well guided the synthesis of NF membranes with outstanding selectivity for both positively and negatively charged monovalent/divalent ion pairs.