High-Affinity Detection and Capture of Heavy Metal Contaminants using Block Polymer Composite Membranes.

Adsorptive membranes offer one possible solution to the challenge of removing and recovering heavy metal ion contaminants and resources from water supplies. However, current membrane-based sorbents suffer from low binding affinities, leading to issues when contaminants are present at trace concentrations or when the source waters have a high concentration of background electrolytes that compete for open binding sites. Here, these challenges are addressed in the design of a highly permeable (i.e., permeability of ∼2.8 × 104 L m-2 h-1 bar-1) sorbent platform based on polysulfone and polystyrene-b-poly(acrylic acid) composite membranes. The membranes possess a fully interconnected network of poly(acrylic acid)-lined pores, which enables the surface chemistry to be tailored through sequential attachment of polyethylenimine brushes and metal-binding terpyridine ligands. The polyethylenimine brushes increase the saturation capacity, while the addition of terpyridine enables high-affinity binding to a diversity of transition metal ions (i.e., Pd2+, Cd2+, Hg2+, Pb2+, Zn2+, Co2+, Ni2+, Fe2+, Nd3+, and Sm3+). This platform removes these metal contaminants from solution with a sorbent capacity of 1.2 mmol g-1 [based on Cu2+ uptake]. The metal capture performance of the functionalized membranes persists in spite of high concentrations of competitive ions, with >99% removal of Pb2+ and Cd2+ ions from artificial groundwater and seawater solutions. Breakthrough experiments demonstrate the efficient purification of feed solutions containing multiple heavy metal ions under dynamic flow conditions. Finally, fluorescence quenching of the terpyridine moiety upon metal ion complexation offers an in situ probe to monitor the extent of sorbent saturation with a Stern-Volmer association constant of 2.9 × 104 L mol-1. The permeability, capacity, and affinity of these membranes, with high-density display of a metal-binding ligand, offer a chemically tailored platform to address the challenges that arise in ensuring clean water.

Introduction

Accelerating urbanization, expanding mining operations, defense munition remediation, and aging infrastructure have led to growing concern regarding heavy metal contamination in fresh water supplies.[1−4] Treatment processes specializing in heavy metal remediation can be broadly categorized into chemical precipitation, biocatalytic processes, size-selective membrane separations, or adsorptive separations.[5−10] Size-selective membrane separations, such as reverse osmosis (RO) and nanofiltration (NF), remove dissolved metal ions by sterically restraining the hydrated ions, whose size is larger than that of water molecules, from passing through a thin selective barrier.[11] This approach can achieve exceptionally high purity in the permeate water. However, such purity is possible at the expense of substantial energy demands that result from the large transmembrane pressure drops needed to overcome the osmotic pressure of the feed solution.[12,13] Additionally, the management of the heavy metal enriched concentrate is a significant challenge because further evaporation and crystallization are required to minimize the volume of this waste prior to its disposal.[14,15] Even then, there are concerns that the metal ions may leach from the disposal site and return as an environmental contaminant.

Adsorption-based processes offer the ability to efficiently capture dissolved metals using ligands that sequester metal ions on the sorbent surface through a variety of chemical interactions.[16−21] Conventional adsorption operations utilize packed beds filled with microporous beads of resin that are 100–500 μm in diameter.[22] While these beads provide large saturation capacities, their relatively large diameters and the associated mass transfer resistance result in an inefficient use of the available binding sites.[23] Adsorptive membranes, based on porous templates (e.g., pore diameters, dp ≈ 10–1000 nm), direct flow through the porous matrix and offer shorter diffusion distances between the solutes and the binding moieties on the sorbent surface.[19,24−29] This configuration reduces mass transfer resistances and, consequently, affords higher throughput operation while utilizing available binding sites more efficiently. However, despite the advantageous mass transfer associated with membranes, the performance of many state-of-the-art adsorptive membranes is hindered by low binding affinities and limited saturation capacities. To overcome these challenges, the design of an ideal sorbent would have a high density of binding sites with high equilibrium binding constants for transition metal ions to enable nonspecific heavy metal removal. Moreover, it should operate at low hydraulic pressures with high throughput and afford reuse through binding site regeneration using external stimuli.

Previous studies that have targeted membrane adsorbers for metal ion remediation have addressed the issue of low saturation capacity by attaching a high density of binding sites on the membrane surface. For example, poly(glycidyl methacrylate)-lined cellulose nanofiber membranes were further functionalized with poly(acrylic acid) (PAA) brushes via an epoxide ring-opening reaction. The resulting films achieved a high cadmium (Cd2+) saturation capacity of 1.4 mmol g–1 and a hydraulic permeability value of ∼1000 L m–2 h–1 bar–1.[26,27] In another approach, the pore walls of a mesoporous membrane fabricated from a self-assembled polyisoprene-b-polystyrene-b-poly(N,N-dimethylacrylamide) block polymer were converted to PAA moieties, providing the membrane with a high copper (Cu2+) saturation capacity of 4.1 mmol g–1, an equilibrium binding constant 200 M–1, and a Cu2+ to nickel (Ni2+) selectivity of 10.[29] Though membranes in these prior studies demonstrated large saturation capacities for individual heavy metals under ideal conditions, these membranes had limited binding affinity due to the use of ligands with only a modest ability to coordinate metal ions. Furthermore, heavy metal contamination is frequently present as a component in complex aqueous solutions,[30] and adsorptive membranes typically display a lower binding capacity in the presence of competitive background electrolytes due to precluded access to available binding sites. Therefore, it is crucial to develop a membrane adsorber with high-affinity ligands that efficiently capture many heavy metal contaminants in the presence of competing ions.

In addition to tailoring binding to improve sorbent performance, adsorption processes would benefit from enhanced in situ monitoring and process control.[31−34] For example, in reverse osmosis processes, it has been demonstrated that the intensity of fluorescent tracers such as rhodamine-WT or uranine could be utilized as a reliable means to monitor the integrity of the membranes while in operation.[35,36] In adsorptive processes, data on the saturation of the sorbent would inform decisions regarding process regeneration or replacement. While prior studies have remarked on the tendency of transition metals to quench the fluorescence of some binding ligands upon complexation,[37−40] little work has been devoted to the development of such a capability in membrane adsorbers due to limited chemistries available for the attachment of an appropriate fluorescent ligand. Thus, through molecular design of the sorbent and ligand chemistries, it could be possible to leverage the quenched fluorescence of the metal–ligand complex to provide real-time detection of the saturation of the membrane adsorber as well as an accurate representation of the local concentration of the heavy metal. Meeting the design challenge of such an adsorptive membrane requires deliberate selection of membrane templates for advantageous transport performance, as well as binding ligand chemistry with observable characteristics that change as a function of metal binding.

In this study, we address these challenges in the development of membrane adsorbers through a bottom-up design approach to fabricate a polysulfone (Psf) block polymer composite membrane template that can be further tailored to enhance binding affinity and expand sorbent functionality. A PAA-lined membrane with a fully interconnected bicontinuous network consisting of pores ∼1 μm in diameter was prepared through the use of surface-segregation and vapor-induced phase separation (SVIPS) methodology. Then, through facile coupling reactions, the pore wall was chemically tailored to attach heavy metal binding groups, which increased binding capacity and enhanced the sorbent affinity for a diversity of transition metal ions. Specifically, terpyridine (TerP) moieties with innate fluorescence were introduced through a polyethylenimine (PEI) intermediate. This route additionally turned the membrane into a sensitive heavy metal probe, and demonstrates that the appropriate choice of pore wall chemistry can target multiple design criteria simultaneously. The Psf block polymer composite serves as the foundation for a variety of affinity or adsorption-based separations with potential in situ saturation monitoring capability. Furthermore, the surface-segregated pore wall functional group can be tailored according to the specific demands of targeted applications, such as biomolecular recognition,[41] or trace metal mining from seawater.[42,43]

Results and Discussion

Fabrication of Highly Porous Composite Membranes with Surface Segregated PAA Brushes

The processing of functional polymers into thin films with bicontinuous porous structures is critical to generating adsorptive membranes with high permeability values and large surface area-to-volume ratios. Here, the SVIPS methodology, which is based on the controlled intrusion of humid air into a polymer solution (Figure ) was used to generate the desired microstructure. A casting solution of 8% (by weight) Psf and 2% (by weight) PS–PAA copolymer dissolved in 2-pyrrolidinone (2P) was used to fabricate composite Psf/PS–PAA membranes. This solution was cast into a 305 μm-thick film on a glass substrate using a doctor’s blade and exposed to a humid environment at ∼95% R.H. and 25 °C for a predetermined period of time. As the water vapor from the environment dissolves into the thin film, it initiates phase separation (i.e., liquid–liquid demixing) and the formation of an interconnected spongy structure.[44,45] Because the nascent pore connectivity of the spongy structure is critical to producing a membrane with high hydraulic permeability, the coalescence of the transient morphology formed by the phase separation was hindered through solvent selection.[46] Namely, 2P was chosen as the solvent because it forms hydrogen bonds with both the Psf matrix and the incoming water vapor. As a result, the viscosity of the polymer-rich phase is increased,[47] which sufficiently elongates the characteristic coalescence time such that the bicontinuous morphology persists over the course of the vapor intrusion process. At the end of the vapor intrusion step, the film is precipitated in a DI water bath to fix the microstructure of the thin film in place. Using this approach, the morphology of the membrane was regulated through tuning the humid air exposure time.

Figure 1

Schematic of the surface-segregation and vapor-induced phase separation (SVIPS) membrane fabrication process. (a) The polymer solution was prepared by dissolving the polysulfone (PSf) and PS-b-PAA in 2-pyrrolidone. (b) The polymer solution was drawn into a uniform thin film on a glass substrate. (c) The casting solution thin film was exposed in a humid environment (with a relative humidity ∼95%) for a predetermined amount of time. The intrusion of water vapor from the humid air into the casting solution contributes to the formation of a uniform cross-sectional architecture comprised of spongy cells. (d) The film was subsequently plunged into a nonsolvent water bath that caused the hydrophobic polymers to precipitate and vitrify the membrane nanostructure. Simultaneously, due to their hydrophilic nature, the PAA brushes preferentially segregate toward the surface of the pore wall. (e) The composite membrane was annealed in a bath of DI water at 80 °C to allow the PAA brushes to extend toward the center of the pore.

The formation of a spongy, bicontinuous network of pores is associated with phase separation induced by the penetration of water vapor into the film. This phase separation occurs slowly as the dissolved water vapor traverses the film thickness as a diffusion front. In cases where the film was plunged into the nonsolvent bath before the diffusion front of dissolved water was able to reach the lower portion of the thin film near the glass substrate, the polymers precipitated following a nonsolvent induced phase separation (NIPS) process. In this case, due to the rapid intrusion of nonsolvent, the NIPS process results in asymmetric membranes with a finger-like macroporous substructure.[44,48] The contrast in these microstructures can be observed in Figure . For films not exposed to humid air (i.e., a vapor exposure time of 0 s), a microporous finger-like structure persists over the whole membrane cross-section (Figure a), while for a 45 s vapor exposure time the spongy bicontinuous structure spans part of the membrane thickness before the finger-like structures develop (Figure b). The dissolved water penetrates through the entire film thickness with an exposure of 90 s and thus a spongy membrane is seen throughout the thickness with cross-sectional pore sizes that slowly taper from ∼500 nm surface features to ∼1 μm at the bottom. The progression of the spongy microstructure through the cross-section of the films is accompanied by evolving surface structures. The NIPS process generates membranes with surface pore features ∼100 nm in diameter. Vapor intrusion initially results in high porosity which decreases with higher exposure times due to the coalescence of the polymer-rich phase near the surface. The coalescence also results in a broader range of pore sizes on the membrane surface, with features ranging from 200 to 2000 nm in diameter.

Figure 2

SEM micrographs show the structure of the composite membranes as a function of the time the cast films are exposed to humid air. At the end of the vapor exposure period, the membranes were plunged into a nonsolvent bath to fix the membrane nanostructure in place. (a–c) The cross-sectional architecture of the membranes. (a) Membranes with no exposure to humid vapor precipitated in a manner consistent with the nonsolvent induced phase separation (NIPS) process with delayed demixing and slow precipitation. In this case, the cross-section possesses an asymmetric morphology with an active layer that tapers into a finger-like macroporous substructure. (b) A short vapor exposure time generates asymmetric membranes that are composed of spongy cells supported by finger-like pores underneath. The transition from a spongy microstructure to a finger-like microstructure occurs in the vicinity of the penetration depth of the water vapor. (c) Membranes with a sponge-like cross-section and graded cellular substructure were obtained with prolonged exposure time to the humid environment. (d–f) The surface morphology of the membranes undergoes a transition from (d) a mesoporous structure that is induced by the NIPS process to (e) a macroporous structure with a higher density of pores produced by short vapor exposure period and (f) macroporous feature with a lower density of pores generated by a longer exposure time.

In addition to generating the desired morphology, the SVIPS approach allows for the introduction of targeted chemical functionality along the pore walls of the membrane. The components of the PS–PAA block copolymer were specifically selected to achieve this aim. The aromaticity of the polystyrene repeat units caused the PS block to preferentially mix with the polysulfone matrix during phase separation.[49−51] Meanwhile, the hydrophilic PAA block preferentially segregated toward the surface of the pore walls as it prefers contact with the nonsolvent. Thermal annealing of the films in a water bath at 80 °C enabled the PAA blocks to further segregate into the surrounding water. These surface-segregated PAA brushes were homogeneously distributed over the pore surfaces, as demonstrated by the high magnification micrograph in Figure S1, thereby allowing full access to the PAA repeat units as functional sites for further coupling reactions. Thus, a bicontinuous, porous membrane template was established which possessed carboxylate groups that could be easily modified to impart functionality directed toward target applications. Here, such functionality is aimed at heavy metal adsorption studies.

pH-Responsive Hydraulic Permeability of PAA-Lined Template

The pH-responsive hydraulic permeability of the membranes confirmed that the pore walls were PAA-lined and demonstrated the stable attachment of the PAA brushes. At pH 5.5, the composite membrane fabricated with 45 s vapor exposure exhibited a permeability of 2.5 × 104 L m–2 h–1 bar–1 ,while the membrane fabricated with 90 s exposure had a permeability value of 2.2 × 104 L m–2 h–1 bar–1 (Figure S2). The similarly high permeability values suggest the membranes prepared with a symmetric morphology (90 s) and those fabricated with an asymmetric morphology (45 s) both possess a fully interconnected pore structure. Subsequently, measuring the hydraulic permeability of the membranes while subjecting them to cyclic changes in solution pH (Figure S3) demonstrated that at pH 13 the hydraulic permeability was ∼1.8 × 104 L m–2 h–1 bar–1, while at pH 1 the hydraulic permeability was ∼3.2 × 104 L m–2 h–1 bar–1. These reversible changes over the course of four cycles were consistent with the scaling analysis suggested in the limit of a low Reynolds number flow. Specifically, in this limit, the permeability depends on the effective pore diameter to the fourth power, dp4, where dp can be affected by the conformation of the polymer brushes that line the pore walls. On the basis of the molecular weight of the PAA, the observed changes in permeability were consistent with the extension and contraction of PAA brushes that results, respectively, from the deprotonation and protonation of the carboxylic acid repeat units (see Supporting Information for details).[49,52]

Chemically Tailored Surface Chemistries for Targeted Ion Capture

The reaction scheme in Figure a outlines the coupling reactions utilized to tailor the chemistry of the PAA-lined polysulfone template for application as a heavy metal adsorber. A detailed reaction scheme is listed in Figure S4. The PAA brushes of the parent membrane exhibit modest metal binding capacities as a result of ion-exchange mechanisms. However, the pKa of the carboxylic acid (∼4.3, based on the value for the PAA monomer) may compromise cation removal performance at moderate pH (i.e., pH 4 to pH 7) due to the reduced number of charged repeat units available to interact with cations.[53] Establishing a covalent linkage between the template wall and a strong transition metal complexing group that can be regenerated is critical to the design of an efficient heavy metal purification device. To achieve this end, the pore wall was further functionalized through two simple chemical reactions. The first reaction attached branched poly(ethylenimine) (PEI), a chemical linker that expands the number of sites available for further modification and heavy metal capture, to the pore wall. PEI was covalently linked to the PAA block through a carbodiimide coupling reaction that was executed by immersing the membrane in an aqueous solution containing branched PEI (Mn ≈ 60 kg mol–1), 1-ethyl-3-(3-(dimethylamino)propyl) carbodiimide hydrochloride (EDC·HCl), and hydroxybenzotriazole (HOBt) at room temperature for 4 days. This chemical conversion was monitored using Fourier transform infrared (FT-IR) spectroscopy, and the resulting spectra are shown in Figure S5. The appearance of a broad peak at ∼1630 cm–1 is associated with the existence of primary amines and the formation of amide carbonyl groups.[54] The PEI possesses its own innate metal-coordinating capabilities through amine electron donation.

Figure 3

(a) The coupling reaction schemes utilized to convert the pore wall chemistry from a parent PAA-lined pore to a terpyridine-lined pore through an intermediate polyethylenimine-lined (PEI-lined) pore. (b) Copper binding isotherms for the three types of membranes. The concentration of Cu2+ bound to the membrane is reported as a function of the concentration of CuCl2 in solution. The dashed lines through the data represent the corresponding Langmuir isotherm determined from linear regressions. The isotherm suggests a saturation capacity of 0.36 mmol g–1 for the Psf-PAA membrane and 0.80 mmol g–1 for the Psf-PEI membrane. (c) A magnified view of the experimental isotherm for the membranes with terpyridine-lined pore walls. The isotherm suggests a saturation capacity of 1.2 mmol g–1. Inset: photographs of the parent and copper-saturated membranes situated. (1) bare Psf-PAA, (2) Cu2+ saturated Psf-PAA, (3) Cu2+ saturated Psf-PEI, and (4) Cu2+ saturated Psf-TerP. The error bars are propagated standard deviation derived from multiple experiments.

The second reaction anchors the strong heavy metal coordinating group 6-(2,2′:6′,2′′-terpyridin-4′-yloxy) hexanoic acid (TerP) to the primary amines of the PEI following a similar carbodiimide coupling mechanism as that utilized in the first functionalization step. The synthetic procedure and molecular characterization (i.e., 1H NMR and mass spectroscopy) of the TerP ligand are provided in Figures S6 and S7. Using ethanol as a solvent, the heterogeneous membrane–reaction mixture was heated to 70 °C for 12 h because TerP only partially dissolves in polar protic solvents at room temperature. The enhanced peak intensity observed at ∼1560 cm–1 and ∼3300 cm–1 of the FT-IR spectrum demonstrates the presence of additional secondary amines adjacent to amide bonds at high density,[55] which in turn indicates the successful linkage of the TerP moieties with strong binding affinity for transition metals to the pore walls of the membrane. Furthermore, SEM micrographs in Figure S8 suggest the morphology of membranes and the pore connectivity that results from the SVIPS process were retained despite modification of the pore wall chemistry. Additionally, it should be noted the chemically modified membranes were mechanically robust and easily handled prior to and after the pore wall coupling reactions (Figure S9).

Elemental analysis by X-ray photoelectron spectroscopy (XPS) further corroborates the successful surface modification and provides valuable information to support the heavy metal binding performance of the membranes. While both grafted functional groups, PEI and TerP, contain nitrogen, survey scans of each successively modified membrane (Figure S10) suggest an increasing trend in nitrogen content as represented by the N 1S photoelectron intensity and decreasing oxygen concentration as characterized by the O 1S intensity. The ratio of nitrogen to oxygen content increased from 8.4 × 10–2 for Psf-PAA to 0.76 for Psf-PEI and 0.90 for Psf-TerP. Meanwhile, the cation adsorption performance of each ligand was screened using Cu2+ as a model solute. Prior to the XPS study, pieces of membrane saturated with Cu2+ were rinsed with excess DI water to remove unbound copper ions. The progression of the PEI-functionalization scheme corresponds with an increasing photoelectron intensity of Cu2+, suggesting the incorporation of branched PEI enhanced the number of sites available for cation binding.

To fully elucidate the Cu2+ binding performance of these membranes, the concentration of Cu2+ bound to the membrane was systematically determined through static binding experiments. Circular sections of membranes that were 2.5 cm in diameter were immersed in aqueous solutions of cupric chloride at varying concentrations and allowed to adsorb cations for 8 h. The membranes reached 90% of their saturation capacity value within 90 s of exposure to solution but were left overnight to ensure equilibrium was approached (Figure S12 and associated video). The bound copper was subsequently released, which regenerated the membranes for repeated use. While the implementation of a pH 1 hydrochloric acid solution was sufficient to protonate the membrane and release the bound Cu2+ cations from the Psf-PAA and Psf-PEI membranes, 50 mM ethylenediaminetetraacetic acid (EDTA) solution was needed to remove cations from the Psf-TerP membranes. The Cu2+ concentrations in the resulting retentate and release solutions were quantified using ultraviolet–visible (UV–vis) spectroscopy for samples treated with the Psf-PAA and Psf-PEI membranes or inductively coupled plasma optically emitting spectroscopy (ICP-OES) for solutions originating from the use of the Psf-TerP membranes.

The copper binding isotherms of the membranes determined from these measurements are reported in Figure with the equilibrium concentrations of bound Cu2+ (q) plotted as a function of the retentate concentrations (c). In this manner, the maximum binding capacity (Q) as well as the binding affinity (K) of the three membranes were quantified by fitting the isotherms to the Langmuir model.

The linearization of the Langmuir isotherm model used to fit the values of Q and K is shown in Figure S13; the resulting fit is plotted as the dashed lines in Figure . The leftward horizontal shift of the copper binding isotherm for the Psf-TerP membranes (orange circles), which reaches saturation at a retentate concentration of ∼1 mM (64 ppm of Cu2+), demonstrates the strongest copper binding affinity. The thermodynamically favored binding between copper and TerP is characterized by a high equilibrium binding constant of 6400 M–1. In contrast, the bound Cu2+ concentrations of Psf-PAA (red rhombus) and Psf-PEI (blue square) membranes increase slowly with respect to increasing retentate concentrations with saturation only being approached at higher retentate concentrations near 100 mM (6.4 × 103 ppm of Cu2+). This lower copper binding performance is directly correlated to lower K values of 110 and 82 M–1, for PAA-lined and PEI-lined membranes, respectively. These trends in metal binding affinity are consistent with the Pearson acid–base concept that classifies pyridines as a borderline Lewis base that favors binding with Cu2+, a borderline Lewis metal ion. Conversely, carboxylate and aliphatic amines are defined as hard binding sites that do not favor binding with Cu2+.[56,57]

While binding affinity directly contributes to the heavy metal capture performance at lower cation concentrations, the maximum binding capacity is another critical parameter in determining the utility of a sorbent for a given process. The Q value determined from the Langmuir isotherm provides a direct comparison between the maximum capacity of the three types of membranes studied here. The original PAA-lined membrane demonstrates a Q value of 0.36 mmol g–1, which is slightly lower than the number of acrylic acid repeat units incorporated within the membrane (0.43 mmol g–1). That is, the copper binding mechanism at saturation likely follows a one-to-one stoichiometry since ∼80% of the PAA repeat units are charged in an aqueous CuCl2 solution with a pH in the range of pH 4 to pH 5 if one-to-one binding stoichiometry is assumed.[29] The covalent attachment of branched PEI to the pore walls introduces additional amine binding sites for cation adsorption, which is consistent with the increased saturation capacity of 0.80 mmol g–1. Albeit a higher capacity is achieved for Psf-TerP relative to the Psf-PEI, it is likely that the number of active binding sites remains nearly constant during the second functionalization reaction, and the cation binding mechanism with Cu2+ transitioned from a four-to-one or six-to-one ligand-to-ion ratio for the branched PEI chemistry to a two-to-one or even one-to-one ratio for the terpyridine chemistry.[58,59] The Q value of the terpyridine-lined membrane was experimentally determined as 1.2 mmol g–1, which is comparable to the values of commercial resins in the range of 1.1–2.2 mmol g–1.[60,61] It should be noted that such capacity leaves room for further improvement; block polymer sorbents have reported capacities as high as 4.1 mmol g–1.[29] Photographs of the three membranes at their saturation capacity are displayed in Figure c. The noticeably blue and cyan colors of the PEI-functionalized and TerP-functionalized membranes are consistent with changes expected for Cu2+ binding by these different chemistries.[62,63] The higher affinity and high capacity Cu2+ binding exhibited by membranes functionalized with tailored chemical moieties suggest that the binding characteristics of these films can be systematically altered or optimized through the appropriate selection of metal binding groups. No significant change in the sorbent capacity was observed upon recycling these membranes through at least 10 adsorption and regeneration cycles (Supporting Information, Figure S14). Moreover, the value of the hydraulic permeability of the Psf-TerP membranes was similar to that of the Psf-PAA membranes (Figure S15), further suggesting the pore structure (e.g., membrane morphology) was not significantly changed during the pore wall modifications.

Transition Metal Ion Capture Performance

Engineering the pore wall chemistry manipulates the transition metal binding profile of the three membranes over a broad spectrum of cationic species, as highlighted by the results presented in Figure . The cation removal performance of membranes was assessed by immersion in single component solutions containing one of eight transition metal cations at a concentration of 18 ppm. Further details of the batch uptake experiments are provided in Figure S16. A packing ratio of 2.0 g of membrane (L of solution)−1 was used. The pH of these solutions was unadjusted except for experiments with PdCl2, which were executed at pH 1. After the solutions were left unstirred for 8 h, the membranes were removed, and the concentrations of metals in the retentate solutions were determined using ICP-OES. The extent of cation removal was determined by measuring the concentration of the targeted ion in solution before and after the adsorption experiment (e.g., as shown in Figure S17). The Psf-TerP membranes removed the eight cations indiscriminately (Figure c), as demonstrated by metal ion removals in excess of 95% for all metal species. This performance is likely attributed to the high affinity of the cations to the terpyridine rings.[59] Critically, this performance, which was achieved with highly permeable thin films that exhibit minimal mass transfer resistance, is similar to recently reported metal–organic framework based (MOF-based) ion traps.[16]

Figure 4

(a–c) A survey of metal ion removal efficiency for the Psf-PAA membranes, the Psf-PEI membranes, and the Psf-TerP membranes in deionized water. The static adsorption experiments were performed with an initial cation concentration of 18 ppm and with a packing ratio of 2.0 g of membrane per L of solution. (d–f) A comparison of Pb2+ and Cd2+ removal by the Psf-PAA membranes, the Psf-PEI membranes, and the Psf-TerP membranes. The metal ion removal experiments were executed in deionized (DI, labeled as red column), synthetic groundwater (Ground, labeled as blue column), and artificial seawater (Sea, labeled as gray column). A single metal salt at an initial cation concentration of 6 or 18 ppm was utilized in each experiment. The membranes were added to these solutions at a packing ratio of 2.0 g of membrane per L of solution. The initial and final cation concentrations were assessed using ICP-OES. Error bars are propagated standard deviation derived from multiple measurements.

The other membranes in this study, Psf-PAA and Psf-PEI, display less efficient ion removal due to lower binding affinities and saturation capacities. In the case of PAA-lined membranes, the moderate pKa of acrylic acid makes the concentration of deprotonated carboxylate groups susceptible to protonation, which reduces the number of available binding sites. In an extreme example, due to the hydrochloric acid incorporated to facilitate the dissolution of PdCl2, the binding solution was pH 1. At this low pH, most of the PAA repeat units are protonated, and thus <5% of the Pd2+ was removed from solution. Furthermore, the carboxylate groups of PAA offers hard binding sites with higher equilibrium constants when complexing with hard Lewis metal ions [such as Nd3+ or Sm3+, which were >90% removed (Figure a)]. However, the binding between soft and intermediate Lewis acids to the carboxylate groups of PAA is less favored, resulting in lower equilibrium binding constants, as shown by the Cu2+ binding, and only partial removal of many of the cations selected in the screening experiment. In contrast, the Psf-PEI material displayed more efficient cation removal as a result of its increased saturation capacity. Through the covalent attachment of densely packed amine moieties onto the parent Psf-PAA membrane, the PEI-lined membrane efficiently purified a broad range of heavy metal ions (Figure b), while partially (∼70%) removing Co2+ and Zn2+. Given its ability to surpass the removal efficiency of carboxylate and amine-based membrane adsorbers, the indiscriminate complexion of TerP lined membrane adsorbers to heavy metal ions demonstrates its potential in applications related with wastewater treatment and trace metal recovery.

The ability to reversibly adsorb heavy metal ions of concern to human health and the environment, especially with high concentrations of competing or interfering dissolved ions in the background, is critical to the development of membrane sorbents for use in saline sources of water. To this end, experiments were conducted to screen lead (Pb2+) and cadmium (Cd2+) removal by membranes immersed in solutions with cation concentrations of 6 and 18 ppm. Both of these concentrations are higher than the cation concentration typical of contaminated water, which is on the order of ∼10–100 ppb. A packing density of 2.0 g of membrane L–1 was used in these experiments. As shown in Figure e,f, both Psf-PEI and Psf-TerP remove 99+% of metal cations, even in the presence of monovalent and multivalent ions present in synthetic groundwater and seawater, which have total dissolved solids (TDS) values of 820 ppm and 36 000 ppm, respectively. The ability of Psf-PEI and Psf-TerP to remove lead and cadmium even in solutions containing high concentrations of competing ions and/or background electrolytes suggests utility in a wide variety of environments. In addition, the high permeability of the Psf-TerP membrane adsorbers (Figure S15) on the order of ∼2.8 × 104 L m–2 h–1 bar–1 could make them candidates for multistage operations, such as in conjunction with RO/NF separations where adsorbers could be used to remove metal ions from the concentrated brine prior to its disposal.

Because deprotonated PAA operates by a different binding mechanism that more closely resembles ion-exchange,[64] the performance of the Psf-PAA membrane in the presence of background electrolytes is different, and in this case, worse than that of PSf-PEI or PSf-TerP. As a control, Psf-PAA shows a lower removal performance when there is no background electrolyte and can only partially remove heavy metal ions contained in synthetic seawater. At high ionic strength (e.g., seawater), the number of available charges may be screened or neutralized by Manning condensation initiated by counterions from the background solution.[65] In contrast, at low ionic strength (i.e., DI water), the charge density of deprotonated PAA may be affected by charge regulation that decreases the number of available binding sites.[66,67]

Dynamic Metal Ion Capture Performance

The rapid uptake kinetics demonstrated in batch adsorption experiments are consistent with the idea that diffusion from the membrane surface to unoccupied binding sites is the dominant resistance to ion capture (Figure S12). Therefore, as demonstrated by the breakthrough curves in Figure , the functionalized Psf membranes are well-suited to capture metal ions from contaminated feed solutions under dynamic flow conditions. Breakthrough curves were obtained by stacking three membranes in a stirred cell and passing heavy metal contaminated feed solutions through the stack at a volumetric flux (i.e., superficial velocity) of ∼200 L m–2 h–1. Under these conditions, the average residence time of solution within the stack was ∼5 s. Still, due to the rapid uptake kinetics, this was a sufficient period of time to reduce the concentration of metal ions from 1 ppm in the contaminated feed solutions to the single digit ppb level in the permeate solutions.

Figure 5

Dynamic breakthrough experiments demonstrate metal ion removal in pressure-mediated flow. The concentrations of heavy metal ions in the aliquots of solution that permeated through a test bed membrane stack are plotted versus the cumulative volume of contaminated feed solution treated. The test bed consisted of one Psf-PEI membrane atop of two Psf-TerP membranes. In many instances, the concentration of heavy metal contaminants in the efflux from the test bed were below the LOQ of ICP-OES. In such cases, the ion concentration in the permeate solution were plotted at the LOQ for the various metal ions, and this region is also shaded gray. Cd2+ and Pb2+ readings below their LOQ values were adjusted up to 10 ppb, while the Cu2+ and Hg2+ readings were adjusted up to 5 and 20 ppb, respectively. The concentration measured by the ICP-OES is shown in Figure S18. No dramatic breakthrough was observed for the ternary feed solution containing a mixture of Cd2+ (orange star), Pb2+ (purple circle), and Hg2+ (red rhombus) metal ions. A binary mixture of 1 ppm of Cd2+ (green triangle) and 1 ppm of Cu2+ (blue square) in the feed solution demonstrates a Cd2+ breakthrough starting at ∼25 mL, while efficiently removing the incoming Cu2+. Photographs of the colorless feed solution from this experiment as well as the surface of the membrane taken after the breakthrough experiment demonstrate the ability of these membrane stacks to effectively capture and concentrate metal ions under dynamic flow conditions.

The breakthrough curves plot the concentration of metal ions in the permeate solution versus the volume of contaminated solution permeated. Two different feed solutions were utilized in these breakthrough experiments. One feed solution contained 1 ppm each of Pb2+, Cd2+, and Hg2+ ions. In this instance, breakthrough of Cd2+ initiated after 80 mL of solution had been treated, while the concentrations of Pb2+ and Hg2+ remained below the limit of quantification (LOQ) over the whole course of the experiment (105 mL of solution permeated). The second feed solution contained 1 ppm of Cu2+ and 1 ppm of Cd2+. Here, there is a sharp breakthrough of Cd2+ starting at 25 mL of permeated solution. However, the concentration of Cu2+ remained below the LOQ, and no breakthrough was observed over the course of the experiment. The earlier onset of Cd2+ breakthrough in this second feed solution suggests that the incoming Cu2+ displaces bound Cd2+ ions due to its higher affinity for the terpyridine ligands lining the pore wall. The high affinity of the membrane stack for Cu2+ gains support from visual comparison of the 1 ppm feed solution, which is colorless and transparent, and the clearly blue membrane surface after 100 mL of the feed solution is permeated through the membrane. This dramatic color change is consistent with the efficient heavy metal removal in the flow-through configuration that is observed in the ICP-OES analysis.

Psf-TerP as a Sensitive Fluorescent Probe to Detect Heavy Metal Ions

The composite membranes demonstrated efficient ion removal performance in multiple background electrolyte conditions through the attachment of higher binding affinity ligands. The manipulation of pore wall chemistry may facilitate function beyond enhanced metal binding affinity and capacity. For example, the ability of metal ions to rapidly quench the original fluorescence of the TerP moiety via complexion, even at low concentrations, offers an opportunity to use the Psf-TerP membrane as a sensitive fluorescent probe to provide online information regarding sorbent saturation and the local concentration of heavy metal ions.[38,39,68] This function of the Psf-TerP membrane, relying on fluorescence quenching of TerP by Cu2+ metal ions, is demonstrated in Figure . The fluorescence emission at 408 nm was significantly attenuated relative to the bare membrane fluorescence even with a trace concentration of Cu2+ of only 20 μM (1.3 ppm). This observation is supported by the fluorescent micrographs at various background Cu2+ concentrations in Figure b; the intensity of the blue fluorescence emission gradually decreased with increasing bound Cu2+ concentration.

Figure 6

Quenching of Psf-Terp membrane fluorescent intensity with increasing bound Cu2+ concentration (q). (a) Fluorescent emission spectra (λex = 358 nm) as a function of retentate Cu2+ concentrations. (b) The fluorescent intensity collected from emission wavelength 408 nm follows a linear relation with bound Cu2+ concentration, suggesting the fluorescence is sensitive to the extent of Cu2+ saturation. The micrographs inset shown were taken from fluorescent microscopy with background Cu2+ concentration (i) 0, (ii) 0.05, (iii) 0.2, and (iv) 1 mM, respectively.

The variation in fluorescent intensity (I) relative to the intensity observed in a solute-free system (Io) is a function of the concentration of the fluorescent quencher and can be quantified using the Stern–Volmer eq (eq ). In the case of the Psf-TerP membrane (Figure b), the reduction in fluorescence intensity, plotted as Io/I, exhibits a linear relationship (R2 = 0.97) with the concentration of bound Cu2+ (q), which implies that bound copper is quenching the fluorescence of the TerP moieties; KSV is the association constant based on the concentration of bound copper.

As such, fluorescence detection enables facile determination of the concentration of Cu2+ bound to the membrane. Moreover, because the binding isotherm for copper has been established above, these measurements can additionally be utilized to identify the concentration of copper in solution. The sensitivity of this detection was quantified using the quenching efficiency at low Cu2+ concentrations where the isotherm is approximately linear (Figure S19). In this regime, the association constant relating fluorescence quenching to the concentration of copper in solution (KSV′ = KSVQK) was ∼2.9 × 104 L mol–1, indicating detection in the ppm range. Detection of heavy metal ions by the Psf-TerP membrane in this concentration range is more sensitive than luminescent MOF-based sensors that utilize a Lewis base as the functional component.[69,70] Taken together, the fluorescence quenching of Cu2+ when bound to TerP-lined pore wall affords convenient and accurate heavy metal detection, and simultaneously provides a means to readily monitor the extent of sorbent saturation. The capability to execute such in situ monitoring of the sorbent saturation and metal ion concentration is an important advantage for quality control that is possible through specific design of the pore wall chemistry.

Concluding Remarks

A high flux and high capacity membrane adsorber with fully interconnected bicontinuous morphology was prepared through surface-segregation and vapor-induced phase separation methodology. The membrane fabrication was combined with straightforward coupling reactions to covalently introduce transition metal binding moieties on the pore walls for efficient and nonspecific heavy metal adsorption. While exhibiting exceptionally high hydraulic permeabilities, the functionalized membranes provide for indiscriminate metal ion removal by sequestering a broad spectrum of 10 heavy metal ions with percent removals in excess of 95%, including in the presence of background electrolytes at a high ionic strength representative of that found in seawater. The quenching of terpyridine fluorescence observed when heavy metal ions were bound to the sorbent provided an in situ monitoring strategy to quantify adsorption effectiveness and device life. Furthermore, most metal ion binding is reversible, facilitating membrane regeneration protocols. Additionally, examining the size-selective and potential fouling characteristics of the functionalized Psf membranes will be critical toward the systematic design of next-generation hybrid water treatment systems.[71] The facile modification of the pore wall chemistry using straightforward coupling reactions affords a new membrane fabrication paradigm for high performance heavy metal removal based upon the selection of ligand chemistry, with additional potential applications in biomolecular recognition, biopharmaceutical separations, munitions remediation, or supported catalytic functions.