(622a) Inverse Design of Pore Wall Chemistry to Control Solute Transport through Molecular Simulation Coupled Optimization

Next-generation membranes for water treatment that move beyond simple desalination require materials with precisely tuned functionality. In particular, the purification and reuse of highly contaminated waters, such as oilfield-produced water, face key challenges in removing a wide array of solutes, including small neutral solutes that are difficult to separate. Multifunctional membrane surfaces potentially provide a vast, underexplored design space to improve membrane transport properties, but are difficult to design through trial-and-error. Here, we demonstrate an inverse design computational approach to efficiently identifying promising materials. We develop an optimization workflow coupled to molecular simulations to engineer the transport of water relative to that of boric acid in a model nanopore by spatially patterning the pore wall with nonpolar methyl and polar hydroxyl groups. The genetic algorithm optimization identifies non-intuitive functionalization strategies that hinder the transport of boric acid through the pore, simply by altering the functional group patterning. We show that a machine-learned surrogate function can predict the performance of novel patterns based only on features of the patterns and significantly accelerates the optimization. Clustering of the patterns reveals design rules which we use to create rationally designed patterns revealing the distinct mechanisms by which water and boric acid transport through the pore. Finally, we demonstrate that these inverse design techniques can also be applied to multi-objective optimization and show that the optimal patterns may help break the pernicious permeability-selectivity tradeoff in membrane design. This inverse design procedure and the accompanying molecular-level understanding of the effect of chemical heterogeneity on solute selectivity demonstrate new routes to the design of membrane materials with novel functionalities, and more broadly, show that surface chemical patterning can modulate water-mediated surface-solute interactions in systems relevant to many other technologies.