Water quality sensors are essential to predict and diagnose acute shocks and chronic fluctuations in various water systems. Polyvinyl chloride (PVC)-based ion-selective membrane (ISM) sensors have evolved into well-established water quality assessment tools. However, current studies about ISM fabrication rely on exhaustive experimental investigations by conventional screening procedures from an educated guess and empirical observation, which results in heavy development costs and time.

This study addresses these prominent challenges by combining machine learning (ML), molecular fingerprint (MF), Bayesian optimization technologies, and experimental results to develop transformative PVC-based ISM cation sensors with exceptional performance based on 1745 datasets collected from literature in the last 20 years. Appropriate ML models trained with literature-based datasets were developed to enable accurate prediction and a deep understanding of the relationship between ISM components and sensor performance (R² = 0.75). A reference MF based on the chosen atomic groups derived from ML model interpretation was constructed for the rapid screening of ionophores. Further, Bayesian optimization was applied to identify optimal combinations of ISM materials with the potential to deliver desirable ISM sensor performance. The Na⁺, Mg²⁺ and Al³⁺ sensors based on the Bayesian optimization results were evaluated and exhibited excellent Nernst slopes with deviations < 7% from the ideal value among all the predicted sensors, and superb detection limit at 10^-7 M level from the experimental validation results.

This study will potentially shift the paradigm of the conventional experiment-driven sensor development process to a more time-efficient, cost-effective, and rational membrane design strategy guided by ML-based techniques.

In this study, we demonstrate the successful layer-by-layer (LbL) deposition of a novel sacrificial membrane coating composed of molybdenum disulfide (MoS₂) nanosheets and polyelectrolytes. Its sacrificial nature comes from the ease of removal and subsequent regeneration of the layers on the surface of a commercial polyamide nanofiltration membrane, owing to the electrostatic interactions between the alternately charged components. Sacrificial layers have garnered research interest in recent years particularly as a method of membrane fouling control; during a cleaning stage, the sacrificial layers and attached foulants can be thoroughly removed, leaving the underlying membrane clean for subsequent redeposition of the sacrificial layers. MoS₂ has shown promise in water treatment and membrane applications due to the material’s smooth layer structure, strong negative surface charge, and antimicrobial properties, which make it well-suited for use in sacrificial LbL-assembled membranes. The LbL assembly and removal processes are modeled using quartz crystal microbalance with dissipation monitoring (QCM-D) combined with ellipsometry. The MoS₂ -polyelectrolyte multilayers are deposited on the surface of commercial polyamide membranes, which are then tested in a bench-scale reverse osmosis system with model foulants. The fouling propensity of the membrane for different types of foulants are evaluated, and layer removal strategies are compared. The successful incorporation of two-dimensional nanomaterial MoS₂ with polyelectrolytes in these sacrificial multilayers unlocks further potential beneficial uses of MoS₂ in membranes for water treatment.

The ability to selectively permeate a single ion from a mixture of ions with similar size and charge, such as selecting lithium from a mixture of alkali cations, is a critically-needed capability for enabling resource recovery applications such as battery recycling. Polymer membranes are a proven technology capable of other types of ion separations, but historical methods of characterizing membrane performance and rationalizing ion transport do not provide enough insight to explain observed ion-ion selectivity. This makes engineering membranes for ion-ion separations extremely challenging.

We devised a new framework for understanding ion transport by combining approaches drawn from fields outside the traditional bounds of environmental engineering, including battery research, computational chemistry, and chemical kinetics. Our general and comprehensive kinetic modeling framework makes it possible to quantitatively estimate the activation energy barriers of partitioning and diffusion of each individual ion in a mixture using traditional membrane characterization methods. We used this framework to study alkali cation transport in a water-stable polymer of intrinsic microsporosity (AquaPIM) membrane and a commercial cation exchange membrane, which we characterized via ionic conductivity and partitioning measurements.

Our results show that partitioning has a higher activation energy barrier than diffusion in almost all cases, in contrast with historical assumptions. Furthermore, we find evidence that activation entropy may play an important role in competitive ion transport. Informed by our results, we discuss the meaning of “rate limiting” in the context of membrane transport and present implications for more rational design of ion-selective membranes.

Antifouling membranes are key for water and resource recovery from challenging streams. Zwitterionic polymer brushes have attracted significant interest for strong and versatile antifouling surface modification, including membrane functionalization. However, scaling up conventional fabrication methods, which require deoxygenation and occur in bulk solvent, is a major challenge.

In this presentation, we report systematic development and characterization of polymer brush membranes. We recently developed brush grafting from membranes at ambient conditions for the first time, via Cu⁰-mediated surface-initiated atom transfer radical polymerization (Cu⁰-SI-ATRP). Here, we report relationships between polymerization methods, brush properties, and hydrophilic/hydrophobic foulant adhesion via force microscopy, for well-defined brush layers on Si wafers and for poly(vinylidene fluoride) (PVDF) membranes. Brush growth can be accelerated by increasing confinement, via rapid oxygen consumption and reduced activator concentrations. However, fouling resistance is significantly lower, for Si wafer and membrane substrates, associated with greater heterogeneity. We have also demonstrated independent control of brush thickness and density to investigate their effects on fouling resistance for the first time, for multiple foulant and membrane types. Finally, we report relationships between adhesion force measurements and membrane fouling in dynamic filtration tests. Cu⁰-SI-ATRP results in PSBMA brush layers that increase water flux (by up to 18%) and improve fouling resistance (reducing adhesion by up to 99% and increasing fouled water flux by up to 30%). Results also demonstrate Cu⁰-SI-ATRP scalability for membrane substrates from 2.25 to >150 cm².

This work provides new understanding of material property-performance relationships, guiding design and scalable fabrication of brush-modified membranes.

Hydrophilic modification with thermosensitive polymers is one of the promising techniques to alleviate membrane fouling and to easily regenerate a fouled surface via simple temperature-change cleaning. However, it is intrinsically unfavorable to detach foulants above the lower critical solution temperature (LCST) because of the enhanced hydrophobicity of thermosensitive polymers. Despite many attempts to utilize thermosensitive polymer for membrane modifications, the effects of increasing hydrophilicity of polymers above the LCST, such as the incorporation of hydrophilic poly(ethylene glycol)-based polymer and changing a structure of copolymer, remain unclear and controversial. Here we examine the role of polymer structure (i.e., homo, random, and block), consisting of thermosensitive polymer (poly(N-isopropylacrylamide, PNIPAM) and hydrophilic polymer (poly(2-[2-(2-methoxyethoxy)ethoxy]ethyl acrylate, PMEO₃A), on antifouling and surface cleaning performance of ultrafiltration polyethersulfone (PES) membrane. We synthesized different polymer structures with the same degree of polymerization and grafted the polymers to the polydopamine-coated PES membrane. Because of the presence of hydrophilic PMEO₃A moieties in the copolymers, copolymer-grafted membranes displayed less hydrophobicity and protein adsorption capacity above the LCST compared to PNIPAM-grafted membrane. Dynamic protein fouling experiments also showed that the copolymer-grafted membranes had a higher water flux recovery ratio (FRR) after cleaning above the LCST, and the random copolymer had the smallest irreversible fouling with the highest FRR. Our findings indicate the incorporation of hydrophilic polymer into thermosensitive polymer enhances antifouling properties and surface cleaning performance of the membrane, which emphasizes a sophisticated polymer design should be further investigated to maximize its antifouling ability.