(557f) Design and Synthesis of Fused Particle and Minimal Surface Microstructure Polymer Membranes with Machine Learning and Experiments

To optimize the filtration performance of synthetic polymer membranes, their internal microstructure needs to be designed based on theoretical principles and not empirically, as occurs at present, using phase inversion and interfacial polymerization. Here we present a rational approach to design membrane microstructure using a 3D in-silico tool. We develop and use a 3D in-silico tool to simulate separation of particles in (a) commercial microporous polymeric membranes, and (b) membranes with novel computer-generated microstructures. To validate these predictions, experiments are underway. The in-silico tool comprises computational fluid and particle drag mechanics combined with particle and membrane force measurements in aqueous solutions, modeled by the extended DLVO (xDLVO) theory¹, to study particle intrusion and capture in microporous polymer membranes. We have previously demonstrated² that (a) the 2D in-silico tool predictions for a commercial membrane show substantial flow channeling and qualitatively agree with experimental filtration measurements using scanning electron microscopy with particle tracking via energy dispersive X-ray spectroscopy, (b) tear-drop microstructures of different orientation weakly separate particles that differ in size by a factor of 2, and (c) interactions between proteins (streptavidin) and polymeric membranes can be described by the xDLVO theory³.

In this work, the 3D microstructure of commercial polymeric membranes (Express and Durapore, MilliporeSigma, Bedford, MA) was digitally reconstructed through focused ion beam – scanning electron microscopy (FIB-SEM). This 3D membrane microstructure along with intermolecular force parameters which are obtained through AFM force measurements and a computational fluid mechanics program (Filtration module, GeoDict, Math2Market GmbH, Kaiserslautern, Germany) constitute the 3D in-silico tool. Applying this tool to study the performance of commercial polymeric microporous membranes showed that channeling could be a ubiquitous feature that severely limits selectivity. The tool was then used to obtain predictions of filtration performance for separating suspended colloids that differ in size by a factor of 10 (i.e., the holy grail of selectivity for membrane filtration) using membranes with (a) fused particle microstructures composed of same or different sizes of spherical particles and (b) minimal surface packing microstructures. Microstructures composed of spherical particles can be synthesized by controlled deposition and crosslinking of spherical colloidal particles. 3D printed minimal packing microstructures have high tortuosity and have been extensively used in column packings to enhance heat and mass transfer. These novel microstructures allow adjustment of the tortuosity of the flow path to maximize performance. The geometrical design parameters used to generate these microstructures and the corresponding predicted in-silico selectivities and permeabilities can be used to train a machine learning model to predict an optimal microstructure of a target separation.

In-silico selectivities for separating particles of 0.1 and 1 mm diameter, for commercial PES microporous membrane and membranes with fused particle microstructure were ~ 4 and 13, respectively, illustrating that removing channeling will increase separation of solutes. To our knowledge, this is the first report combining particle drag mechanics with intermolecular force measurements in 3D to design and synthesize microfiltration membranes based on theoretical principles. The in-silico tool can be used to characterize membranes for separation performance, and it can be combined with machine learning to guide improved microstructure design, synthesis, and testing of new microporous membranes.

References

1. van Oss, C., Long-range and short-range mechanisms of hydrophobic attraction and hydrophilic repulsion in specific and aspecific interactions. Journal of Molecular Recognition 2003, 16, 177-190.
2. Sorci, M.; Wookcock, C. C.; Andersen, D. J.; Behzad, A. R.; Nunes, S.; Plawsky, J.; Belfort, G., Linking microstructure of membranes and performance. J. Membr. Sci. 2020, 594, 1-9.
3. Karla, S.; Sorci, M; Tice, L; Hersey, J; Giglia, S; Petersen, P; Moussa, B; Plawsky, J; Belfort, G., Short-range hydration forces are critical in describing molecular interactions between streptavidin and polymeric membranes. Oral presentation, Session: Surface forces between biological and non biological interfaces, ACS Spring 2023, Indianapolis.