Two-dimensional MXene membranes with biomimetic sub-nanochannels for enhanced cation sieving

Biological ion channels regulate ion transport across cell membranes with exceptional selectivity (e.g., K+ channels exhibiting K+/Na+ selectivity >10,000). Their ultimate selectivity arises from synergistic effects of sub-nanoscale pore size, specific binding sites, and appropriate charge density. Mimicking these features in artificial membranes can enable advances in water treatment, resource extraction, energy conversion, and biosensing. Prior approaches replicated individual features—creating uniform sub-nanochannels (e.g., graphene oxide, MOFs, perforated PET), grafting functional groups to impart chemical affinity, or modulating charge density—but reproducing all three aspects precisely remains challenging. Moreover, how local charge density in sub-nanochannels governs ion transport and separation is not fully understood. Inspired by the KcsA K+ channel (~5.6 Å filter with evenly distributed carbonyl oxygens), this work uses Ti3C2Tx MXene nanosheets and EDTA molecules to construct artificial ion channel membranes with controlled 2D channel size (~6.0 Å), KcsA-like oxygen-rich binding sites, and tunable local charge density to enhance cation sieving, particularly for mono-/divalent ion discrimination.

Earlier efforts focused on mimicking biological channel structure by assembling uniform sub-nanochannels using materials such as graphene oxide, metal-organic frameworks (MOFs), and perforated PET. Due to limited selectivity from size exclusion alone, subsequent strategies grafted functional groups (e.g., sulfonate) to introduce differential chemical affinity and tailored channel charge density to modulate ion selectivity. While these approaches capture individual features of biological channels (size, affinity sites, or charge), replicating all three simultaneously with high precision has been challenging. The field lacks a clear mechanistic understanding of how local charge density in sub-nanochannels affects ion transport and separation, motivating the present biomimetic MXene–EDTA design.

Materials and synthesis: Ti3C2Tx MXene nanosheets were synthesized via an improved MILD etching method by selectively removing Al from Ti3AlC2 using LiF (1.0 g) and HCl (20 mL, 9 M) at 37 °C for 24 h, followed by repeated washing to pH ~6, delamination by ultrasonication (110 W, 30 min, ice bath, Ar flow), and centrifugation to obtain ~2 mg mL−1 monolayer dispersions. Concentration was measured gravimetrically by mass difference after deposition on PES substrates. Membrane fabrication: EDTA-2Na solutions (0–1.5 mg mL−1) were mixed with Ti3C2Tx dispersion (1.0 mg mL−1) and stirred 6 h at room temperature, enabling EDTA grafting via covalent (Ti–COO) and hydrogen bonding. The mixtures were vacuum-filtered onto PES supports (0.22 µm, 47 mm) with staged pressure (0.3 bar then 1 bar), dried (80 °C, 6 h), washed, and soaked to remove unbound EDTA. Membranes were denoted MLM (no EDTA) and MLM-EDTA-X (X=0.25–1.5 mg mL−1). Characterization: Morphology and structure were examined by SEM (ESEM Quanta 250 FEG) with EDS, AFM (thickness, lateral size), TEM/SAED (Tecnai F20; cross-sections via resin embedding and ultramicrotomy), XRD (Bruker D8, Cu Kα), XPS (PHI 5000 VersaProbe-III), FT-IR (Nicolet iS5), zeta potential of nanosheets (Malvern ZS90) and membrane surface (SurPASS). Channel spacings and anti-swelling behavior were derived from XRD d-spacings and peak FWHM in dry and various saline conditions. Ion permeation measurements: A U-shaped diffusion cell (two 30 mL chambers) was used; the membrane (MXene side facing feed) was encapsulated between perforated aluminum tapes. Feed contained mixed salts (KCl, NaCl, LiCl, CaCl2, MgCl2), each 0.2 M; permeate contained DI water. Both chambers were magnetically stirred. Ion concentrations in permeate were quantified by ICP-OES. Permeation rates Pi (mol m−2 h−1) and selectivity S were calculated: Pi = (Ci − Co)·V/(A·t); S = (P1/C1)/(P2/C2). pH-dependent tests adjusted both chambers (0.05–0.1 M HCl/NaOH). Concentration and temperature dependences were measured with controlled sequences and water-bath thermostating. Energy barrier and charge density: Arrhenius plots of ln(Pi) vs 1/T (20–40 °C) gave energy barriers Ea via ln(Pi) = ln(α) − Ea/(RT). Membrane charge density σ was estimated using Gouy–Chapman: σ = (εκε)·sinh(Fζ/2RT)1/2 with Debye length κ−1 = [2εRT/(F2Co)]1/2, using membrane surface zeta potential; the channel charge density was assumed equal to surface charge. Computational methods: DFT (Gaussian 09, UB3LYP/DEF2SVP, water solvation) computed adsorption energies Eads of EDTA with K+, Na+, Li+, Ca2+, Mg2+ for varying EDTA deprotonation states (reflecting pH ~8 and beyond). MD simulations (GROMACS 2020.6) modeled ion transport through Ti3C2(OH)2 laminar channels with and without EDTA intercalation; d-spacings of 18.4 Å (pristine) and 16.0 Å (EDTA-intercalated). Force fields: UFF (Ti3C2(OH)2), SPC/E water, OPLS-AA for ions and EDTA. Systems contained ~0.2 M ions in feed, equilibrated (NPT, 2 ns), followed by 20 ns production runs at 300 K (Nosé-Hoover). Charge densities were tuned by deprotonation counts of hydroxyl and carboxyl groups. Ion entry and passing times, hydration numbers, and RDFs were analyzed.

- Biomimetic channel and anti-swelling: EDTA crosslinking reduced swelling and stabilized sub-nanochannels. MLM d-spacing increased from 2.5 Å (dry) to up to 9.2 Å (aqueous), while MLM-EDTA-1.0 increased from 2.8 Å to 6.0 Å across DI water and 0.2 M salt solutions, yielding channel size comparable to the KcsA filter (~5.6 Å). - Ion sieving performance: MLM showed limited selectivity (K+/Mg2+ ~10; K+/Li+ ~1.4) due to larger channels (~8.4 Å). MLM-EDTA membranes selectively impeded divalent and smaller monovalent ions while maintaining K+ transport, achieving K+/Li+ up to 5.4 and K+/Mg2+ up to 121.2 in mixed 0.2 M salt feeds, surpassing previously reported membranes under similar conditions. - Transport order and dehydration: Due to partial dehydration required to enter ~6.0 Å channels and corresponding hydration energies (K+ > Na+ > Li+ >> Ca2+ > Mg2+), observed permeation rates followed K+ > Na+ > Li+ >> Ca2+ > Mg2+. - EDTA ion recognition (DFT): Calculated adsorption energies (more negative indicates stronger binding) showed preferential binding of EDTA to divalent ions: Mg2+ −3.50 eV, Ca2+ −2.42 eV vs Li+ −1.49 eV, Na+ −1.20 eV, K+ −0.98 eV. This elevates divalent ion transport barriers through EDTA-decorated channels. - Charge density effect: Increasing membrane charge density (by pH tuning) from −7.0 to −22.0 mC m−2 for MLM-EDTA-1.5 raised K+/Mg2+ selectivity from 53.8 to 112.5, with linear correlations between charge density and K+/Mn+ selectivity (M = Na+, Li+, Ca2+, Mg2+). Energy barrier differences between K+ and Mg2+ increased from 3.09 to 5.78 kcal mol−1 over the same charge-density range. The contribution of non-charge effects (size, hydrogen bonding) to Mg2+ transport resistance remained ≤8.4%. - MD insights: At low channel charge density (−0.78 e nm−2), entry is rate-limiting: Mg2+ loses ~2.4 waters (14.3 → 11.9) vs K+ ~0.5 (6.1 → 5.6) upon entry into MLM-EDTA, prolonging Mg2+ entry time; K+ entry time slightly decreases due to favorable electrostatics. As charge density increases to −11.26 e nm−2, Mg2+ entry time decreases (electrostatics overcome entry dehydration), but passing time increases dramatically (+440% for Mg2+ vs +130% for K+), enhancing K+/Mg2+ separation. Elevated charge density induces additional in-channel dehydration and increased hydrogen bonding between hydrated cations (especially Mg2+) and deprotonated channel oxygens, increasing friction and promoting Mg2+–EDTA complexation. - Overall mechanism: Cation selectivity is entry-dominated at low charge density, shifting to passing-dominated at high charge density. The combined effects of size-controlled partial dehydration, EDTA-mediated ion recognition, and tunable local charge density yield ultrahigh mono-/divalent selectivity with maintained K+ permeability.

The study addresses the challenge of simultaneously emulating size, binding sites, and charge density of biological ion channels in synthetic membranes. By integrating EDTA within Ti3C2Tx MXene laminar channels, the authors construct sub-nanochannels (~6.0 Å) enriched with negatively charged oxygen-containing groups that mimic KcsA’s selectivity filter. Experiments and simulations show that selectivity arises from two synergistic mechanisms: (1) size-controlled partial dehydration at the channel entry, which disproportionately hinders ions with high hydration energies (especially Mg2+), and (2) chemical recognition by EDTA, which preferentially complexes divalent cations, further increasing their transport resistance. Tuning local charge density (via pH) enhances these effects by deprotonating EDTA carboxyls and MXene hydroxyls, strengthening electrostatic interactions and EDTA–Mg2+ complexation. This results in linear improvements in K+/Mn+ selectivity with charge density and record-high K+/Mg2+ selectivity (121.2) at practical permeation rates. MD and DFT corroborate the mechanistic picture, revealing a transition from entry-limited to passing-limited transport as charge density increases, with additional in-channel dehydration and hydrogen bonding increasing Mg2+ friction and complexation. These findings provide a mechanistic basis for designing ion-separation membranes that combine precise channel size control with tailored affinity sites and charge density.

EDTA-functionalized MXene laminar membranes (MLM-EDTA) replicate key features of biological ion channels—sub-nanochannel size, oxygen-rich binding sites, and tunable charge density—yielding compact, water-stable channels with ultrahigh mono-/divalent cation selectivity. The membranes achieve a K+/Mg2+ selectivity of up to 121.2 in mixed salt feeds while maintaining favorable K+ permeability. Experimental characterization, DFT, and MD simulations indicate that selective sieving stems from the interplay of partial dehydration at channel entry, EDTA-mediated ion recognition, and increased local charge density that strengthens divalent cation complexation and hydrogen bonding. The work offers a strategy for rationally designing affinity groups and local charge density in sub-nanochannels for applications in resource recovery, clean water production, and power generation.

- Charge density estimation assumes channel charge density equals membrane surface charge density (Gouy–Chapman model), which may deviate from the true in-channel environment. - MD simulations required exposing a row of oxygen atoms at the channel exit to increase detectable ion permeation rates, an artificial condition that may not fully represent actual channels. - Performance metrics were primarily evaluated in mixed 0.2 M salt solutions and around neutral-to-alkaline pH; behavior under different ionic strengths, compositions, or extreme pH was not comprehensively detailed in the provided text.