(364q) Revealing Molecular Mechanisms in Polymeric Membranes through Molecular Simulation and Modeling

I am looking to direct materials discovery efforts through molecular dynamics simulations and theoretical and machine learning models. My research interests are focused on understanding how nanoscale structure and dynamics influence material properties and performance. My goals include incorporating molecular insights into engineering problems, leveraging predictive and interpretable machine learning models for the molecular sciences, and implementing accessible best practices for scientific software development.

My expertise lies in developing atomistic and coarse-grained models of soft materials and employing these models to explain the complex transport mechanisms in water purification membranes. I have applied enhanced sampling techniques to study governing mechanisms in membrane transport. In addition to these molecular simulations, I have derived a theoretical model that expands transition-state theory to heterogeneous membrane transport. These analytical expressions connect molecular jumps to observable membrane performance, and I have written software to numerically test this framework in detail. I have experience designing, applying, and teaching machine learning models for particle trajectories, images, and large tabular datasets.

Abstract

Major efforts in recent years have been directed towards understanding the molecular determinants of transport in reverse osmosis (RO) and nanofiltration (NF) membranes in order to design highly selective membranes for specific applications, including desalination, wastewater treatment, and critical resource recovery. In this presentation, I discuss our work elucidating some of the molecular mechanisms involved in transport in RO and NF membranes.

We expand the transition-state theory approach to model transmembrane permeation to better represent the chemical and physical heterogeneity within realistic membranes. This theoretical framework relates molecular jumps to free energy barriers to membrane transport. However, most investigations treat all these free energy barriers in the membrane as equal, which does not isolate the associated molecular mechanisms. We develop a novel expression relating individual barriers to experimentally observable free energy barriers. We show that experimentally observed energy barriers must be interpreted in terms of the underlying molecular barriers, and naive interpretations of these effective barriers can lead to incorrect assumptions about the transport mechanisms within polymeric membranes.

We provide an in-depth exploration of one such molecular mechanism - ion dehydration. Specifically, we use atomistic molecular simulations and enhanced sampling techniques to characterize the ion dehydration mechanism at the high salinity and high pressure required for membrane operating conditions. We quantify the energy barriers to remove one or more waters from the hydration shell for many relevant salts and operating conditions. Our comprehensive search of the design space will provide a valuable reference for identifying the primary mechanisms for ion transport in industry-standard membrane materials.

We build atomistic models of a polyamide RO membrane and a lyotropic liquid crystal NF membrane and observe solute and water transport within these membranes. We apply a nonparametric Bayesian classification model to automatically detect and fit parameters for hidden states in the solute and water trajectories. We then determine the molecular mechanisms associated with these hidden states, which would dictate the overall membrane performance. Once trained, our machine learning model can project the nanoscale dynamics to experimentally relevant time scales.