Inverse Design of Pore Wall Chemistry To Control Solute Transport and Selectivity

Abstract

Next-generation membranes for purification and reuse of highly contaminated water require materials with precisely tuned functionality to address key challenges, including the removal of small, charge-neutral solutes. Bioinspired multifunctional membrane surfaces enhance transport properties, but the combinatorically large chemical space is difficult to navigate through trial and error. Here, we demonstrate a computational inverse design approach to efficiently identify promising materials and elucidate design rules. We develop a combined evolutionary optimization, machine learning, and molecular simulation workflow to spatially design chemical functional group patterning in a model nanopore that enhances transport of water relative to solutes. The genetic optimization discovers nonintuitive functionalization strategies that hinder the transport of solutes through the pore, simply by patterning hydrophobic methyl and hydrophilic hydroxyl functional groups. Examining these patterns, we demonstrate that they exploit an unexpected diffusive solute hopping mechanism. This inverse design procedure and the identification of novel molecular mechanisms for pore chemical heterogeneity to impact solute selectivity demonstrate new routes to the design of membrane materials with novel functionalities. More broadly, this work illustrates how chemical design is a powerful strategy to modulate water-mediated surface-solute interactions in complex, soft material systems that are relevant to diverse technologies.

Figure 1

(a) Model nanopore system allows for precise patterning of nonpolar methyl groups and polar hydroxyl groups along the CNT backbone. For ease of visualization, we present the flattened pore patterns with blue triangles representing methyl groups and red circles representing hydroxyl groups. (b) We first simulate the finite pore to compute the density of water, which we then set in the infinite pore to compute water and boric acid transport properties. (c) A genetic algorithm incorporating MD simulations and a learned surrogate function allows for automated identification of pore wall chemical patterns optimizing pore transport properties.

Figure 2

Genetic algorithm trajectories maximizing the difference between water flux and boric acid flux for (a) fixed hydroxyl group fraction and (b) varying hydroxyl group fraction show optimization to surfaces with rows of hydroxyl groups. Each marker corresponds to a pattern discovered by the genetic algorithm. Simulated generations are shown as green stars and surrogate model-driven generations as black circles. Inset patterns show the optimal patterns found after (1) the first set of simulated generations, (2) the first set of model-driven generations, (3) the second set of simulated generations, and (4) the second set of model-driven generations. The red line shows the maximum value of the objective function at each point in the optimization.

Figure 3

Pattern of methyl and hydroxyl groups has a significant effect on the flux of (a) water and (b) boric acid in nonequilibrium simulations and on (c) the diffusivity of boric acid in equilibrium simulations. For each panel, orange triangles represent patterns with a “single patch” of hydroxyl groups, spanning the pore axis, while blue circles represent “fully dispersed” patterns (examples of both are in the inset of a). Green diamonds represent other patterns where the hydroxyl groups form smaller patches (“patchy”). Two examples of such patterns are shown in the inset of b. These patterns also show lower boric acid flux compared to the fully dispersed patterns. The inset also includes a schematic of the periodically varying free energy landscape used in the analytical model to predict the reduction in boric acid diffusivity in the ringed pores. In c, green diamonds are only shown for patterns with rows of hydroxyl groups. The dashed line shows the analytical prediction described in the text based on a model of the periodic free energy landscape depicted in the inset of b.

Figure 4

Other solutes exhibit similar trends as boric acid in a selection of rationally patterned pores. Benzene, phenol, and isopropanol show maximal water flux relative to boric acid flux in pores with rows of hydroxyl groups, while for ammonia and arsenous acid, the optimal pattern is the fully methylated one. For the simulations with solutes besides boric acid, we use the same workflow as described for boric acid except that the production simulations in the infinite pore are run for 1 μs. The reported uncertainties are the standard error of the mean from three independent simulations.