Pillared Laminar Vermiculite Membranes with Tunable Monovalent and Multivalent Ion Selectivity

Abstract

Effective membrane separation of Li⁺ from Na⁺ and Mg²⁺ is crucial for lithium extraction from water yet challenging for conventional polymeric membranes. Two dimensional (2D) membranes with ordered laminar structures and tunable physicochemical properties offer distinctive ion‐sieving capabilities promising for lithium extraction. Recently, phyllosilicates are introduced as abundant and cost‐effective source materials for such membranes. However, their water instability and low inherent ion transport selectivity hinder practical applications. Herein, a new class of laminar membranes with excellent stability and tunable ion sieving is reported by incorporating inorganic alumina pillars into vermiculite interlayers. Crosslinking vermiculite flakes with alumina pillars significantly strengthens interlamellar interactions, resulting in robust water stability. Doping of Na⁺ before the pillaring process reverses the membrane's surface charge, substantially boosting Li⁺ separation from multivalent cations via electrostatic interactions. Lithium extraction is often complicated by the presence of co‐existing monovalent cations (e.g., Na⁺) at higher concentrations. Here, by introducing excess Na⁺ into the membrane after the pillaring process, the separation of Li⁺ from monovalent cations is enhanced through steric effects. This work realizes both monovalent/multivalent and monovalent/monovalent selective ion sieving with the same membrane platform. A separation mechanism is proposed based on Donnan exclusion and size exclusion, providing new insights for membrane design for resource recovery applications.

Alumina pillared vermiculite membrane is fabricated with a laminar structure. Enhanced water stability is observed due to the strengthened interlamellar binding. Through cation doping, the surface charge and pore size of the pillared laminar membrane are tuned, demonstrating highly selective and tunable ion sieving performance promising for water treatment and resource recovery applications.

1. Introduction

The recovery of lithium from water, which contains over 70% of the total recoverable lithium on earth,^{[ 1} , ^{2 ]} is becoming increasingly critical in addressing the surging demand from electric vehicles and other technologies. Traditional lithium extraction technologies from water include solar evaporation with precipitation and solvent extraction. The solar evaporation process involves multiple precipitation and crystallization steps to produce lithium carbonate (Li₂CO₃), but it is highly time‐consuming and generates significant waste. Solvent extraction, on the other hand, utilizes specific chelating agents in an organic phase to selectively bind lithium, offering high selectivity but requiring large‐scale usage of organic solvents, which poses substantial environmental risks. In contrast, membrane technologies such as nanofiltration and reverse osmosis have been extensively studied and applied for lithium recovery from water due to their high separation efficiency, low energy consumption, and ability to operate continuously.^{[ 3} , ^{4 ]} However, traditional polymeric membranes suffer from fouling, limited long‐term stability, and trade‐off between permeability and selectivity.^{[ 5} , ^{6 ]} 2D materials such as graphene oxide (GO),^{[ 7} , ⁸ , ^{9 ]} transitional metal dichalcogenides (TMDs),^{[ 10} , ¹¹ , ^{12 ]} and MXenes^{[ 13} , ¹⁴ , ^{15 ]} have been employed as building blocks for membranes to potentially address these issues. 2D membranes are formed by restacking individual 2D flakes, with laminar interlayer channels serving as selective transport paths. The structural and physicochemical properties of the interlayer channels can be precisely tuned for fast and selective transport of ions and water molecules.^{[ 16} , ^{17 ]} However, the high material and processing costs and water instability have limited the applications of 2D membranes to date.^{[ 18} , ^{19 ]}

Phyllosilicates comprise a diverse family of naturally existing 2D materials that are abundant and inexpensive. Vermiculite is one of the most abundant phyllosilicates, the annual global production of which is estimated to be 100 000 tons with an average price of $350 per ton.^{[ 20 ]} An individual vermiculite flake is comprised of one alumina octahedra sheet sandwiched between two silica tetrahedra sheets. The isomorphic substitution of Si⁴⁺ by Al³⁺ in the tetrahedra sheets leads to a negative surface charge of vermiculite, which is balanced by interlayer cations that can be further exchanged to exfoliate the vermiculite flakes. Vermiculite membranes (VMs) with laminar interlayer galleries can be fabricated by restacking the exfoliated 2D flakes.^{[ 21 ]} The facile fabrication process and low material cost of VMs provide a potential path to large scale application for lithium extraction. Compared to traditional polymeric membrane materials, vermiculite exhibits high thermal and chemical resistance and is less prone to fouling due to its fully inorganic structure. Additionally, compared to many other 2D materials, vermiculite exhibits highly tunable surface properties, while also offering a path to significantly lower material and processing costs. The ideal membrane for lithium extraction from the water will have a uniform and stable transport channels to maintain structural integrity, and more importantly, highly selective ion sieving performance to separate Li⁺ from the co‐existing ions including Na⁺ and Mg²⁺. However, VMs readily swell in aqueous environments due to their weak interlamellar interactions and hydrophilic surfaces.^{[ 22} , ^{23 ]} Moreover, the structural and physicochemical properties of VMs have not yet been systematically investigated nor optimized for selective ion sieving. To date, researchers have reported some initial promising separation studies using VMs or membranes based on other phyllosilicates focusing on improving stability,^{[ 24} , ²⁵ , ^{26 ]} but highly selective ion sieving performance remains a challenge.

Selective ion sieving for membrane extraction of lithium can be realized based on Donnan exclusion and size exclusion mechanisms.^{[ 27} , ^{28 ]} Positively charged membranes are favored for Li⁺ separation from multivalent cation separation due to the higher electrostatic repulsion exerted on multivalent cations.^{[ 29} , ³⁰ , ^{31 ]} For example, laminar GO and phyllosilicate membranes exhibited increased Li⁺/Mg²⁺ selectivity after intercalating positively charged polymers.^{[ 32} , ^{33 ]} However, the intercalation of large polymeric species also limits control over interlayer spacing and reduces water permeance.^{[ 34 ]} Separation of Li⁺ from monovalent cations (e.g., Na⁺ and K⁺) is even more difficult due to their similar size and charge.^{[ 35 ]} Ameliorated monovalent cation separation has been demonstrated through size exclusion by creating subnanometer transport pathways in solid electrolyte membranes, but the reduced channel size inevitably results in low permeability and issues of stability and reusability.^{[ 36} , ^{37 ]} In addition, achieving tunable and highly selective monovalent/multivalent and monovalent/monovalent ion sieving using the same membrane remains an unmet goal.

Herein, we introduce pillared laminar membranes with enhanced water stability and tunable selective ion sieving for both monovalent/multivalent and monovalent/monovalent cation separations. Pillared clays (PILCs) are engineered composites comprising phyllosilicate flakes crosslinked with metal oxide pillars, which impart swelling resistance and controllable pore size. Applications of PILCs have been in bulk form, as hydrocarbon reaction catalysts or sorbents for organic pollutants.^{[ 38} , ³⁹ , ^{40 ]} Their applications as membranes with laminar structures, however, remain unexplored. In this work, alumina‐pillared laminar VMs are fabricated for the first time, demonstrating superior water stability. Cation doping has been proven to affect the pore structure and surface chemistry of bulk PILCs.^{[ 41} , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ^{46 ]} Here, by doping Na⁺ before the pillaring process, the membrane surface charge reverses from negative to positive, significantly boosting the monovalent/multivalent (e.g., Li⁺/Mg²⁺) ion transport selectivity. Lithium extraction from seawater or salt‐lake brines is often complicated by competing Na⁺ of higher concentration.^{[ 47 ]} Examination of the membrane ion sieving behavior under concentrated Na⁺ can provide insights for real‐world applications. When excess Na⁺ is introduced after the pillaring process, pre‐doped pillared VM exhibits improved Li⁺ separation from monovalent cations (e.g., Na⁺ and K⁺). Selective ion sieving for both monovalent/multivalent and monovalent/monovalent cations can be achieved with the same pillared membrane by introducing Na⁺ before and after the pillaring process. A separation mechanism based on the combination of Donnan exclusion and size exclusion is proposed, which can be applied to guide membrane design for the sieving of targeted ions for resource recovery.

2. Results and Discussion

2.1. Fabrication and Characterization of 2D Al‐Pillared Vermiculite Membranes

Alumina pillared VMs were fabricated through the process illustrated in Figure 1 . Raw vermiculite material was first exfoliated through a two‐step ion exchange process.^{[ 22 ]} A uniform dispersion of 2D vermiculite flakes was obtained with an average lateral flake size of ≈1 µm and a monolayer thickness of 1.23 nm, which were characterized by atomic force microscopy (AFM) (Figure 2a). The Al₁₃ Keggin polycation (AlO₄Al₁₂(OH)₂₄(H₂O)₁₂ ⁷⁺) is one of the Al³⁺ hydrolysis products that has been applied as a pillaring precursor to prepare bulk alumina PILCs.^{[ 48 ]} Al₁₃ was synthesized through a base hydrolysis method.^{[ 14} , ^{49 ]} The presence of Al₁₃ in the synthesized product was confirmed by ²⁷Al nuclear magnetic resonance spectroscopy (NMR), showing the signature tetrahedral AlO₄ peak located at 63.7 ppm (Figure 2b) in accordance with previous literature reports.^{[ 50} , ^{51 ]} The as‐prepared negatively charged vermiculite flakes and positively charged Al₁₃ were then mixed and crosslinked through strong electrostatic interactions.^{[ 14 ]} A laminar Al₁₃ intercalated VM (Al₁₃‐VM) was obtained after vacuum filtration of the mixture. After subsequent calcination of the Al₁₃‐VM at an elevated temperature, a pillared VM was formed by transforming the Al₁₃ precursor into alumina pillars while maintaining the membrane's laminar structure. Free‐standing membranes with laminar structure were obtained with a thickness of 20 µm (Figure 2c; Figure S1, Supporting Information). As shown in Figure S2 (Supporting Information), the pillared VM remained intact after soaking in 0.1 m Nacl solution for 30 days, demonstrating enhanced water stability compared to the pristine VM, which dissolved after only 30 min of soaking. Furthermore, the pillared laminar membranes maintained their structural integrity and morphology after three weeks of immersion in 1 and 5 m NaCl solutions and exposure to pH conditions ranging from 3 to 12 (Figure S3, Supporting Information), demonstrating exceptional stability under harsh aqueous environments.

Figure 1.

Schematic illustration of the fabrication process of pillared laminar VM and doped pillared VM.

Figure 2.

Characterizations of pillared laminar membranes. a) AFM image and height profile of exfoliated vermiculite flakes. b) ²⁷Al NMR spectra of synthesized Al₁₃ Keggin polycation. c) Cross‐section SEM image of the pillared laminar membrane. d) XRD patterns of pristine VM, pillared VM, and doped pillared VM. e) Al2p XPS spectra of pristine VM and pillared VM. f) Na1s XPS spectra of pillared VM and doped pillared VM.

Doping of cations such as Na⁺, K⁺, Ca²⁺, and transitional metal cations into PILCs, either before or after the formation of pillars, has been proven to significantly alter the pore structure and surface properties.^{[ 42} , ⁴³ , ⁴⁴ , ^{45 ]} Sodium ions were selected as dopants due to their ubiquitous presence in water sources for various membrane application scenarios. The doped pillared VM was fabricated by adding Na⁺ into the mixture of Al₁₃ precursor crosslinked vermiculite before the calcination process. From the X‐ray diffraction (XRD) patterns in Figure 2d, compared with the pristine VM, the d‐spacing of both the pillared VM and the doped pillared VM increased from 1.23 to 1.73 nm, indicating the formation of pillars and a channel height of 0.75 nm (subtracting the vermiculite monolayer thickness of 0.98 nm).^{[ 52 ]} The small subpeak at 2θ = 6.54° was indexed to unconsumed monomers,^{[ 53 ]} which acted as building blocks for Al₁₃ via self‐assembly.^{[ 48 ]} Due to the slow kinetics of this process, some monomers remained unreacted even after prolonged aging; additional aging and washing steps could further reduce their presence if desired. The newly emerged peaks at 2θ = 9.2° with a d‐spacing of 0.96 nm for pillared VM and doped pillared VM, corresponding to the monolayer thickness of vermiculite, can be attributed to collapsed vermiculite layers resulting from the loss of surface charge and dehydroxylation during calcination in the pillaring process.^{[ 54} , ^{55 ]} The broadening of the (100) peaks for pillared VM and doped pillared VM was consistent with previous observations in pillared clays.^{[ 56 ]} Variations in the orientation and distribution of the pillaring precursor,^{[ 57 ]} local disorder caused by collapsed layers,^{[ 58 ]} and effects of washing and drying processes^{[ 59 ]} have been shown to contribute to this decrease of structural uniformity. Suppressed swelling was also evidenced by XRD, as the d‐spacing of pillared VM stayed unchanged at both dry and wet states (Figure S4, Supporting Information). X‐ray photoelectron spectroscopy (XPS) analysis was performed to further confirm the formation of pillars and cation doping. The Al2p peaks after the pillaring process shifted from 73.1 to 75.8 eV, corresponding to the higher binding energy region of alumina (Figure 2e).^{[ 60 ]} Further XPS depth profiling of the pillared VM and the doped pillared VM showed higher Al/Si ratios than the pristine VM, demonstrating the formation of alumina pillars inside the membrane (Figure S5a, Supporting Information). Doping with Na⁺ was confirmed by the newly observed Na1s peak at 1071.5 eV for the doped pillared VM (Figure 2f). The atomic concentration of the sodium dopant was measured to be 0.3% across the membrane from XPS (Figure S5b, Supporting Information), which is at the same level as was observed in previous reports of bulk materials.^{[ 42} , ^{44 ]} The hydrophilicity of clays originates from surface hydroxyl groups and exchangeable cations, which act as water sorption sites.^{[ 61 ]} During the pillaring process, the original exchangeable cations were replaced by Al₁₃ pillaring precursors, which exhibit higher valency and stronger electrostatic interactions with vermiculite. Subsequent calcination transformed the precursor into less hydrophilic alumina pillars through dehydration and dehydroxylation. As a result, water contact angle analysis (Figure S6, Supporting Information) revealed that after the pillaring process, the surface of the pillared VM switched from hydrophilic to hydrophobic due to the reduction of the surface hydroxyl group and exchangeable cation sites.^{[ 62} , ⁶³ , ^{64 ]} However, after doping, the surface hydrophilicity was restored as the dopant cations replenished water sorption sites, which highlights the significant impact of sodium doping on the surface properties of pillared VM even at low concentrations.^{[ 42} , ^{44 ]}

2.2. Tunable Monovalent/Multivalent Ion Sieving

Effective Li⁺/Mg²⁺ separation is critical for lithium extraction from water, as co‐existing magnesium is typically present in higher concentrations and can precipitate during the lithium recovery process.^{[ 31} , ^{65 ]} The ion sieving performance of the pillared VM and the doped pillared VM was investigated by H‐cell diffusion dialysis (Figure S7, Supporting Information). Monovalent/multivalent cation separation experiments were conducted in a binary Li⁺/Mg²⁺ solution with the same concentrations to evaluate the permselectivity. The experimental results on the permeabilities for Li⁺ and Mg²⁺ and corresponding Li⁺/Mg²⁺ permselectivities are shown in Figure 3a. Although the pillared VM exhibited significantly enhanced water stability, its Li⁺/Mg²⁺ permselectivity was limited to 2.1 due to the absence of specific channel properties that facilitate selective separation between monovalent and multivalent cations. However, in the case of the sodium‐doped pillared VM, while the permeabilities of Li⁺ remained at the same level, the permeability of Mg²⁺ decreased by 34 times, resulting in a boosted Li⁺/Mg²⁺ permselectivity of 60.3. Interestingly, for the undoped pillared VM, the permeability of Mg²⁺ also decreased when Na⁺ was introduced into the feed solution, resulting in an improved Li⁺/Mg²⁺ permselectivity of 12.9. Despite the increased hydrophilicity due to the restored water sorption sites from dopants, doped pillared VM exhibited similar Li⁺ permeabilities and significantly reduced Mg²⁺ permeabilities compared to undoped pillared VM, suggesting that ion transport is unlikely to be primarily governed by surface hydrophilicity. To further isolate the influence of monomeric species on ion transport, monomer‐intercalated VM was fabricated by adjusting the precursor reaction pH to 3 and eliminating the aging process.^{[ 66 ]} As shown in Figure S8a (Supporting Information), XRD analysis revealed a single peak at 2θ = 6.50°, confirming the exclusive presence of monomeric species, consistent with the subpeak observed in the pillared laminar membrane (Figure 2d). Ion transport characterization in Li⁺/Mg²⁺ and Li⁺/Na⁺ binary systems showed low ion transport selectivity for monomer‐intercalated VM (Figure S8b, Supporting Information). Therefore, the enhanced separation performance observed in Na‐doped pillared VM is unrelated to the presence of residual monomeric species.

Figure 3.

Tunable monovalent/multivalent ion sieving. a) Li⁺ and Mg²⁺ permeabilities and corresponding Li⁺/Mg²⁺ ion transport selectivity from diffusion dialysis test with an equal amount of LiCl (0.1 m) and MgCl₂ (0.1 m) in the feed side, for pillared VM, pillared VM with additional NaCl (0.1 m) presented in feed side and doped pillared VM. b) Zeta potential of pristine VM, pillared VM, pillared VM after diffusion dialysis with NaCl (0.1 m) presented in feed side and doped pillared VM.

Zeta potential analysis was performed to elucidate the separation mechanism for monovalent and multivalent cations (Figure 3b). The zeta potential of the pristine VM was −28.1 mV, indicating a negative surface charge originating from the isomorphic substitution. During the pillaring process, protons were released from the Al₁₃ precursor to the outer tetrahedral layers of vermiculite, neutralizing the original negative charge and reducing the CEC. As a result, the zeta potential of the pillared VM decreased to nearly zero, creating a charge‐neutral surface. However, for highly selective monovalent/multivalent cation separation, a positively charged surface is preferred to repel multivalent cations through electrostatic interactions. Therefore, the Li⁺/Mg²⁺ permselectivity for the pillared VM with neutral surface charge was still limited. By introducing Na⁺ into the pillaring precursor, sodium was successfully doped into the membrane during calcination, reversing the surface charge from negative to positive. As a result, Mg²⁺ permeabilities were significantly diminished, boosting the Li⁺/Mg²⁺ permselectivity of the doped pillared VM with a zeta potential of + 40.8 mV. Cation doping of PILCs can also be achieved after pillar formation by simply soaking the material in a concentrated salt solution.^{[ 43} , ^{45 ]} When Na⁺ was introduced into the feed solution during diffusion dialysis, the pillared VM was also doped, resulting in a positive zeta potential of + 18.9 mV and improved Li⁺/Mg²⁺ permselectivity. However, the CEC of the pillared VM was reduced after the pillaring process, resulting in a loss of cation adsorption. Consequently, the surface charge modification and corresponding permselectivity enhancement were less pronounced for the pillared VM doped after the pillaring process compared to the doped pillared VM where the dopant cation was added to the precursor before pillaring.^{[ 44 ]}

Beyond structural stability in aqueous environments, maintaining consistent ion sieving performance over time is also crucial for practical membrane applications. The long‐term stability of the doped pillared VM was evaluated through ion sieving performance using binary Li⁺/Mg²⁺ diffusion dialysis. Ion transport permeabilities, dopant cation concentrations in the membrane, and zeta potentials were monitored throughout the diffusion process. As shown in Figure S9a (Supporting Information), despite a slight increase during the early stages of diffusion, the Li⁺ and Mg²⁺ permeabilities for the doped pillared VM remained stable throughout the test. Highly selective ion sieving was sustained over 80 h of diffusion, demonstrating consistency and stability in membrane performance. XPS elemental analysis revealed that the sodium concentration within the doped pillared VM remained stable (Figure S9b, Supporting Information). Additionally, no detectable changes in sodium concentrations in the permeate solution of the diffusion dialysis were observed from the inductively coupled plasma optical emission spectrometry (ICP‐OES) measurements. The consistent presence of sodium within the membrane and in the permeate indicates that the sodium dopant remained stable in the doped pillared VM without leaching or being exchanged by other cations during diffusion. As a result, the surface charge of the doped pillared VM remained positive for selective ion sieving between monovalent and multivalent cations, evidenced by the zeta potential analysis before and after the diffusion test (Figure S9c, Supporting Information). In addition, the doped pillared VM exhibited stable ion sieving performance even after prolonged water immersion. As shown in Figure S10 (Supporting Information), after soaking in water for five months, the membrane maintained a high Li⁺/Mg²⁺ selectivity of 48.8, with only a slight increase in Mg²⁺ permeability compared to the membranes presoaked for 2 h. The doped pillared VM exhibited excellent long‐term stability in both structural integrity and ion sieving consistency, making it promising for critical resource recovery applications. The interaction between various dopant cations and bulk PILCs has been investigated previously, revealing that the dopant cations were located on the pillars.^{[ 67} , ⁶⁸ , ^{69 ]} However, due to the limitations in direct characterization and lack of precise structural modeling (Supporting Information S1), the interactions of Na⁺ with the pillared VM were investigated indirectly by examining their impact on membrane properties including hydrophilicity, surface charge, ion sieving, and water transport.

The enhanced Li⁺/Mg²⁺ separation performance of pillared VM after cation doping was further confirmed through pressure‐driven crossflow filtration experiments. As shown in Figure S11 (Supporting Information), undoped pillared VM exhibited low salt rejection rates for both LiCl (16.2%) and MgCl₂ (68.3%). In contrast, sodium‐doped pillared VM showed increased rejection rates for LiCl (25.1%) and MgCl₂ (93.3%). The more pronounced increase in MgCl₂ rejection was attributed to the induced positive surface charge following cation doping, leading to an enhancement in Li⁺/Mg²⁺ selectivity from 2.76 to 11.6. The enhancement in separation performance is consistent with the transport phenomena observed in diffusion dialysis. The somewhat lower selectivity obtained from crossflow filtration was likely due to the fact that the membrane's morphology and structure were not optimized for crossflow filtration. Future studies will focus on reoptimizing both the crossflow filtration setup and membrane structure to further improve performance under pressure‐driven conditions.

2.3. Tunable Monovalent Ion Sieving

Sodium ions are the most dominant cation species in many bodies of water including seawater and salt‐lake brines.^{[ 70 ]} The extraction of lithium from water will be inevitably influenced by the presence of sodium at higher concentrations.^{[ 71 ]} Therefore, investigating how ion sieving can be tuned by the presence of excess sodium ions is essential for gaining insights into membrane performance for practical applications. For example, seawater typically contains Na⁺ concentration at ≈10 000 ppm,^{[ 72} , ^{73 ]} whereas salt‐lake brines exhibit higher Na⁺ concentrations ranging from 50 000 to 10 000 ppm.^{[ 74} , ^{75 ]} The Li⁺/Mg²⁺ ion separation performance was further investigated for the doped pillared VM under high concentrations of Na⁺ in the feed solution of 0.1 and 1 m, representing the approximate magnitudes of Na⁺ concentration in seawater and salt‐lake brines, respectively. As shown in Figure 4a, highly selective Li⁺/Mg²⁺ ion sieving was observed in the absence of Na⁺. However, as the Na⁺ concentration in the feed solution increased, ion transport was hindered as the permeabilities of all ions decreased. The reduction in Li⁺ permeability was particularly pronounced. Li⁺ permeability decreased by 2.3 times with the addition of 0.1 m Na⁺ and by 39.4 times when 1 m Na⁺ was present. The ion sieving between monovalent Na⁺ and Li⁺ was boosted, achieving a high Na⁺/Li⁺ permselectivity of 23.5 at 1m  Na⁺ concentration. Therefore, highly selective ion sieving can be tuned and realized for both monovalent/multivalent and monovalent/monovalent cation separation through controlled doping of sodium either before or after the pillaring process.

Figure 4.

Tunable ion sieving in doped pillared VM by introducing Na⁺. a) Li⁺ and Mg²⁺ permeabilities with equal amounts of LiCl (0.1 m) and MgCl₂ (0.1 m) and additional NaCl with varying concentrations in the feed side, and corresponding Na⁺ permeabilities and Na⁺/Li⁺ ion transport selectivity. b) Li⁺ and Na⁺ permeabilities and Na⁺/Li⁺ ion transport selectivity under different molar ratios with the same total concentration (0.2 m). c) Ion permeabilities of mixed salt solution for doped pillared VM. d) Water permeances for pillared VM, doped pillared VM, and doped pillared VM after Na⁺ diffusion dialysis.

The tunable monovalent ion sieving is also evidenced by the binary Na⁺/Li⁺ diffusion dialysis in the doped pillared VM (Figure 4b). With increasing Na⁺/Li⁺ molar ratio from 1 to 10, the Na⁺/Li⁺ permselectivity increased from 8.6 to 21. To examine the trend for ion transport of cations with different sizes and valencies, diffusion dialysis tests were performed for the doped pillared VM in a mixed salt solution containing equal amounts of K⁺, Na⁺, Li⁺, Ca²⁺, Mg²⁺, and Al³⁺. The permeabilities for K⁺, Na⁺, Li⁺, Ca²⁺, Mg²⁺, and Al³⁺ were 5.25, 4.92, 0.14, 0.11, 0.08, and 0.01 ×10⁻⁹ cm² s⁻¹ (Figure 4c). K⁺ and Na⁺, having the smallest hydrated ion sizes and the lowest valences, exhibited the highest permeabilities. In contrast, Al³⁺ showed the lowest permeability due to it having the largest size and the highest valence. Intermediate permeabilities were observed for Li⁺, Ca^2+, and Mg²⁺, with hydrated ion sizes larger than K⁺ and Na⁺. Effective separation of Li⁺ from monovalent cations was realized, achieving high Na⁺/Li⁺ permselectivity of 35.1 and K⁺/Li⁺ permselectivity of 37.5. Thus, the doped pillared VM showed superior monovalent/multivalent and monovalent/monovalent ion sieving performance simultaneously, demonstrating efficient separation of cations with varying sizes and valencies.

Cation doping into PILCs exerts a significant impact on their structural and transport properties. For example, sodium doping into alumina‐pillared bulk montmorillonite has been shown to reduce the average pore volume, leading to decreased permeability for gases and small molecules.^{[ 42} , ^{44 ]} To investigate the ion sieving mechanism in the doped pillared VM, water permeance was analyzed as the indicator for pore size evolution (Figure 4d). The undoped pillared VM exhibited the highest water permeance of 18.9 L m⁻² h⁻¹ bar⁻¹. Upon sodium doping, water permeance decreased to 10.4 L m⁻² h⁻¹ bar⁻¹ despite the interlayer spacing remaining unchanged compared to the undoped pillared VM. The drop in water permeance combined with the increased rejection for both Li⁺ and Mg²⁺ (Figure S11, Supporting Information) suggests pore size reduction created by the pillars. Additionally, following Na⁺ diffusion dialysis, the pore sizes were further constricted in the doped pillared VM, evidenced by the lowest water permeance observed at 3.8 L m⁻² h⁻¹ bar⁻¹. The water permeance results combined with zeta potential analysis suggest that ion sieving in the pillared laminar membrane is governed by cation doping through modulation of both surface charge and pore size.

Figure 5 illustrates the proposed ion sieving mechanism, which involves a combination of Donnan exclusion, driven by electrostatic repulsion, and size exclusion, dictated by the physical dimensions of the pores. The undoped pillared VM exhibited a large interlayer spacing and a nearly neutral surface charge, resulting in non‐selective ion sieving where Li⁺, Na⁺, and Mg²⁺ were not effectively distinguished. By introducing sodium dopants intrinsically before the pillaring process, positively charged surfaces were formed, accompanied by a reduction in pore size, as evidenced by decreased water and ion transport. Despite this pore size reduction in doped pillared VM, Li⁺ permeability remained nearly unchanged (Figure 3a), whereas Mg²⁺ permeability decreased significantly, resulting in boosted Li⁺/Mg²⁺ permselectivity. Therefore, monovalent/multivalent cation separation in doped pillared VM was primarily governed by Donnan exclusion, with the influence from reduced pore size playing a less significant role. Following additional extrinsic Na⁺ doping through diffusion dialysis, the permeabilities of both Li⁺ and Mg²⁺ decreased, with a more pronounced reduction observed for Li⁺. The reduction in Li⁺ permeability became more significant as the level of Na⁺ doping increased. The concurrent decline in monovalent cation permeability and water permeance suggests that size exclusion became the dominant ion separation mechanism, enabling selective separation among monovalent cations. This transition of the dominant separation mechanism from Donnan exclusion to size exclusion was likely driven by the progressive reduction in pore size caused by the accumulation of dopant cations around the pillars. In the initial stage, when vermiculite was pillared and intrinsically doped during the pillaring process, the introduced cations altered the surface charge, reversing it to a positive state. This charge reversal enhanced Donnan exclusion, effectively repelling multivalent cations such as Mg²⁺ due to stronger electrostatic interactions, while the impact of pore size reduction remained less significant, allowing the passage of monovalent cations. However, additional extrinsic Na⁺ doping after pillaring resulted in progressive narrowing of the transport pathways. This restriction imposed steric hindrance that increasingly limited the movement of hydrated ions and water molecules, shifting the dominant separation mechanism from Donnan exclusion to size exclusion and restricting the passage of both monovalent cations and water molecules. Consequently, the permeability of Li⁺ decreased due to its larger hydrated ion size and higher hydration enthalpy, leading to enhanced ion sieving among monovalent cations. The separation performance of the doped pillared VM was compared with previously reported membranes, as shown in Figure S12 and Table S1 (Supporting Information). The results demonstrate that the doped pillared VM exhibits highly selective and tunable ion sieving while maintaining fast water transport. Its separation performance in diffusion dialysis is among the highest reported in the literature. The separation performance in crossflow filtration is also comparable to that of polymeric nanofiltration membranes and will be further optimized for pressure‐driven conditions in future work.

Figure 5.

Schematic illustration of ion sieving mechanism in the pillared laminar membrane.

3. Conclusion

In summary, we have successfully demonstrated the concept of pillared laminar membranes using the alumina pillared vermiculite system. Although bulk PILCs have been studied for decades, this work marks the first demonstration of their application as membranes in aqueous separations. The incorporation of inorganic pillars significantly strengthened the interlamellar binding, resulting in improved stability. Furthermore, by controlled cation doping, we were able to fine‐tune the surface charge and pore size of the membrane, achieving tunable and highly selective ion sieving. This approach offers excellent permselectivity for both monovalent/multivalent and monovalent/monovalent cation separations to simultaneously separate target species from diverse interferents, promising for lithium extraction and recovery of other critical materials. The proposed ion sieving mechanism involves both Donnan exclusion and size exclusion that highlights how environmental species influence membrane properties and can be harnessed to enhance ion separation performance. The versatile approach demonstrated in this work can be extended to different combinations of phyllosilicates and pillars, enabling the tailored design of membranes for precise ion sieving. Several interesting topics are worth exploring for future studies, including elucidating ion sieving mechanisms using advanced characterization techniques and uncovering the accurate structure of the pillars with microscopic analysis.

4. Experimental Section

Materials

Vermiculite was obtained from Sigma–Aldrich and used without further treatment. LiCl (≥99.0%), NaCl (≥99.0%), KCl (≥99.0%), MgCl₂ (≥99.0%), CaCl₂ (≥98%), and AlCl₃ (≥99.0%) were obtained from Sigma–Aldrich for membrane fabrication and diffusion dialysis test. NaOH (≥98.0%) was obtained from Fisher Chemical for synthesizing the pillaring precursor.

Synthesis and Characterization of Al₁₃ Precursor

The Al₁₃ Keggin polycation was synthesized by Al³⁺ base hydrolysis. In a typical process, AlCl₃ solution (35 mL, 1 m) was first heated to 60 °C with constant magnetic stirring. NaOH solution (140 mL, 0.6 m) was then added to the AlCl₃ solution dropwise at a rate of 1 mL min⁻¹, reaching a final molar ratio of [OH⁻]/[Al³⁺] = 2.4. After the addition of NaOH was completed, a translucent white mixture was obtained. The mixture was then stirred and heated at the same temperature for another 4 h until a transparent solution was obtained. The solution was then aged at room temperature for 7 days. After aging, the solution was centrifuged at 9000 rpm for 30 min. The supernatant was replaced with DI water, followed by sonication for 30 min to redisperse the Al₁₃ particles. Three total centrifugation‐sonication washing cycles were applied to remove Na⁺. The solid after washing was finally redispersed in DI water, yielding a transparent Al₁₃ solution (100 mL) ready for use and characterization. ²⁷Al NMR spectra of the Al₁₃ solution were collected using a Bruker Ultrashield 500 Plus spectrometer. Quartz NMR tubes were used to reduce the Al background from glass.

Vermiculite Exfoliation

Bulk vermiculite was first exfoliated by a two‐step liquid phase exfoliation method. In the exfoliation process, bulk vermiculite powder (1.0 g) was added to a saturated NaCl solution (500 mL). The mixture was stirred and refluxed at 100 °C for 24 h to replace the interlayer cation with Na⁺. The mixture was then filtered and washed with DI water and ethanol three times to rinse off the excess NaCl. The collected solids were then added to a LiCl solution (500 mL, 4 m) and refluxed again at 100 °C for 24 h to replace the interlayer cation with Li⁺. The mixture was then filtered and washed using the same procedure. The collected solids were redispersed in DI water (500 mL) and sonicated for 1 h to fully exfoliate into individual 2D flakes. The final product was a stable and uniform water dispersion of 2D vermiculite flakes ready for use.

Fabrication of Alumina Pillared Laminar Vermiculite Membrane

The pillared VM was fabricated by calcining the Al₁₃ precursor intercalated vermiculite membranes. In a typical process, Al₁₃ precursor solution (10 mL) was added dropwise (1 mL min⁻¹) to the exfoliated vermiculite solution (100 mL) and sonicated at room temperature for 3 h. The mixture was then vacuum‐filtered onto a polyvinylidene fluoride (PVDF) membrane filter and oven‐dried at 75 °C overnight. After drying, the membrane was peeled off from the PVDF substrate to obtain Al₁₃‐VM. The precursor‐intercalated vermiculite membrane was then calcined at 350 °C for 4 h in the air inside a tube furnace and taken out after natural cooling. The final product was a free‐standing pillared laminar membrane. The doped pillared VM was fabricated using the same process, except Na⁺ was introduced before the pillaring process by adding NaCl (0.5 g) into the mixture of Al₁₃ crosslinked vermiculite flakes. All subsequent processes and conditions were the same as in the fabrication of the undoped pillared VM. Monomer‐intercalated VM was fabricated following the same procedure, except that the pH for precursor synthesis was adjusted to 3 by HCl and immediately used without further aging.

Characterization of Membrane Physicochemical Properties

The lateral size and thickness of the exfoliated vermiculite flakes were characterized using a Cypher‐ES atomic force microscope from Oxford Instruments. AFM samples were prepared by drop‐casting a diluted exfoliated vermiculite flake solution onto a silicon wafer. AFM images were taken in tapping mode using TITAN 300 tips from the same company. The height profile was extracted using a line scan processed by Igor Pro software. The morphologies of the surfaces and cross‐sections of the pillared laminar membranes were characterized by SEM using an FEI Nova 600 NanoLab focused ion beam scanning electron microscope at an operating voltage of 5 kV. SEM samples were prepared by the same drop casting process used for AFM samples, but with gold sputtered to reduce charging effects. The membrane d‐spacings/interlayer spacings were characterized by XRD using a Rigaku benchtop X‐ray diffractometer with a HyPix‐400 MF 2D hybrid pixel array detector (HPAD) and Cu Kα X‐ray source (λ = 1.5406 Å) operating at 40 kV and 15 mA. XPS measurements were conducted by Thermo Fisher K‐Alpha+ using a microfocused monochromatic Al Kα (1487 eV) X‐ray source with a 400 µm spot size. The survey scan used 200.0 eV pass energy with a step size of 1.000 eV. High‐resolution scans used a pass energy of 50.0 eV with a step size of 0.100 eV. XPS depth profiling was conducted under Argon sputter etching, calibrated by Ta₂O₅ to 0.51 nm s⁻¹ (10s each cycle). All XPS spectra were referenced to the C1s peak at 283.8 eV and processed by Thermo Fisher Avantage software (v. 5.977, Build 06436), with deconvolution performed by the Powell peak fitting algorithm with mixed Gaussian−Lorentzian line shapes and a Smart background. Water contact angle measurements were performed on a drop shape analyzer (DSA 25E from KRÜSS) using sessile drop mode. The angles were taken 3 s after the liquid was dropped onto the surfaces. For each membrane, at least 3 measurements were taken at different locations of the membrane to confirm uniformity. The Zeta potential values of the membranes were measured by a SurPASS 3 zeta potential analyzer from Anton Paar using the stream potential in a KCl electrolyte (0.01 m). At least 3 different samples were measured and averaged with a PVDF membrane as the reference.

Characterization of Ion Sieving Performance

The ion sieving performance of the pillared laminar membrane was characterized by diffusion dialysis in an H‐cell with the concentration gradient acting as the driving force. For Li⁺/Mg²⁺ binary diffusion dialysis tests, a salt solution (90 mL) containing the same concentration of LiCl (0.1 m) and MgCl₂ (0.1 m) was added to the feed side of the H‐cell. For Li⁺/Mg²⁺ binary diffusion dialysis tests modulated by Na⁺, different concentrations of NaCl (0.1 and 1 m) were additionally added to the feed side. For Na⁺/Li⁺ and Li⁺/Mg²⁺ binary diffusion dialysis tests with varying ratios, the total concentrations were kept at 0.2 m. For tests using mixed salt solutions, KCl, NaCl, LiCl, CaCl₂, MgCl₂, and AlCl₃ of equal concentrations were added to the feed side, with the total concentration also kept at 0.2 m. The permeate side was filled with DI water of the same volume (90 mL). Both compartments of the H‐cell were constantly stirred at 300 rpm to minimize concentration polarization. The membranes tested were soaked in DI water for 2 h prior to testing to remove any excess ions and impurities. The membranes were mounted to an O‐ring in between two half cells and secured with a chain lock. The permeate side was sampled regularly for concentration measurements by ICP‐OES using an iCAP PRO Duo from Thermal Scientific with equal volumes of solution removed from the feed side to avoid creating a pressure difference. The permeabilities for different ions were represented by the diffusion coefficient extracted from the flux, which follows Fick's first law for concentration‐driven diffusion process:

$$J_i=\frac{V}{A}\left(\frac{d{c}_i}{dt}\right)=D_i\frac{\mathrm{\Delta }{c}_i}{l}$$ (1)

where V is the volume of the half cell, A is the effective membrane area (4.9 cm²), Δc_i is the initial concentration difference of the specific salt between the permeate and feed side, l is the membrane thickness, and D_i is the permeability of the specific salt to be extracted. The unit for diffusion coefficient is cm² s⁻¹, which represents the amount of ions transported across a unit area per unit time under a unit concentration gradient. The permselectivity (S) between two salts can be calculated by their permeabilities:

$$S_j^i=\frac{D_i}{D_j}=\frac{\left(\frac{d{c}_i}{dt}\right)\mathrm{\Delta }{c}_j}{\left(\frac{d{c}_j}{dt}\right)\mathrm{\Delta }{c}_i}$$ (2)

Crossflow Filtration Test

The Li⁺/Mg²⁺ separation performance of the pillared laminar membrane was also characterized by pressure‐driven crossflow filtration using a Sterlitech Sepa membrane test skid. The membrane was mounted onto a customized crossflow cell with an effective filtration area of 4.9 cm². The applied pressure was 2.0 bar, and all tests were conducted at 25 °C. The mixed salt solution containing MgCl₂ and LiCl with total concentration of 2000 ppm and Mg²⁺/ Li⁺ mass ratio of 10 was used as feed. The permeate was collected for 24 h and three tests were conducted for each type of membrane. The ion concentrations in the permeate were measured by ICP‐OES. The rejection rate can be calculated by the ion concentrations in the feed (c_f) and permeate side (c_p):

$$R=\left(1-\frac{c_p}{c_f}\right)\times 100\%$$ (3)

The selectivity obtained from crossflow filtration was calculated by:

$$S_{Mg}^{Li}=\frac{c_p^{Li}/c_f^{Li}}{c_p^{Mg}/c_f^{Mg}}$$ (4)

where c^Li and c^Mg are the concentrations of Li⁺ and Mg²⁺ in the feed (f) and permeate (p) side, respectively.

Water Permeance Test

Water permeances of the pillared laminar membranes were characterized by dead‐end filtration using a stirred cell from Amicon. Membranes were placed at the bottom holder of the cell with an area of 4.9 cm². DI water was then added to the cell and pressure was applied (2.0 bar) as the driving force for water to transport across the membrane. The water permeance (Lp) can be calculated by:

$$Lp=\frac{V}{At\mathrm{\Delta }P}$$ (5)

where V is the volume of water permeated, A is the membrane area, t is the permeation time, and ΔP is the pressure difference across the membrane.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Liu Y., Wang Y., Sengupta B., Kazi O. A., Martinson A. B. F., Elam J. W., Darling S. B., Pillared Laminar Vermiculite Membranes with Tunable Monovalent and Multivalent Ion Selectivity. Adv. Mater. 2025, 37, 2417994. 10.1002/adma.202417994

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.