High permeability sub-nanometre sieve composite MoS₂ membranes

Global demand for potable water necessitates materials and processes that can efficiently remove salts and small neutral organics, which are particularly challenging due to their size similarity to water molecules. Two-dimensional membranes based on graphene oxide (GO) have been widely explored due to ease of fabrication and tunable chemistry but suffer from critical drawbacks: swelling in water that increases interlayer spacing and reduces ion selectivity, low water transport from surface friction with oxygenated groups, and short operational lifetimes before mechanical or performance failure. MoS2 has emerged as a promising alternative owing to zero swelling in water, favorable balance of van der Waals and hydration forces, and simulations predicting robust structure with ultrafast water permeation and high ion selectivity. Intrinsically charged nanopores in MoS2 can enhance ion selectivity via electrostatic effects, and prior studies show superior water transport kinetics and osmotic power generation with MoS2 nanopores. However, experimental demonstrations of nanoporous MoS2 membranes with high performance and durability have been limited, with earlier approaches requiring lengthy processing and offering poor control over interlayer spacing and short operational lifetimes. This study addresses these gaps by developing composite laminate multilayer MoS2 membranes using porous nanosheets and nanodisks with tunable pore size, surface charge, and interlayer spacing to achieve high ion rejection, high water permeance, and long-term stability.

Graphene and GO membranes offer mechanical and chemical stability and controllable interlayer spacing, but key limitations include water-induced swelling that reduces selectivity, friction-limited water transport due to functional groups covering a large fraction of the surface, and limited operational lifetimes. Interlayer spacing control in GO has been attempted via crosslinking, epoxy casting, electrical control, and prolonged salt immersion, with mixed success. Simulations predict that MoS2 membranes, especially with nanopores, can provide higher water flux than graphene while maintaining selectivity; nanopores carry intrinsic charge that can enhance ion rejection. Prior experimental MoS2 studies showed stability in water and potential for sieving and osmotic power generation, but practical desalination performance was limited by fabrication complexity (e.g., month-long dye functionalization) and poor control over interlayer configuration. Other reports indicate that even larger nanopores in carbon membranes can achieve high salt rejection via surface-charge-mediated effects, and that angstrom-scale slits and ion-sieving behaviors are strongly influenced by electrostatic and steric interactions. Commercial thin-film composite polyamide (TFC-PA) membranes set benchmarks but suffer from chlorine susceptibility and fouling issues.

Materials: MoS2 powder (<2 μm, >98% purity) and NMP solvent were used as received. Two octapeptides were designed for MoS2 binding and charge modulation: KFKFKFKF (cationic) and EFEFEFEF (anionic), >95% purity. Commercial TFC-PA RO membranes (SW30 HR) and NaOCl were used for benchmarking and chlorine exposure tests. Synthesis of porous MoS2 nanosheets and nanodisks (NSNDs): A two-step sonication method was employed. Step 1: bath sonication (Branson 2510) of 1 g MoS2 in 100 mL NMP for 4 h (delivered energy ~2880 J mL−1) to exfoliate and mill bulk MoS2 into thin nanosheets (NSs). Step 2: probe sonication (Hielscher UIP 500 H, 500 W, amplitude 100%, continuous operation) for 2 h (delivered energy ~45,000 J mL−1) to induce cavitation, creating nanoholes in NSs and producing nanodisks (NDs) that match pore sizes. Pore diameter was tuned by varying bath/probe sonication durations and pulse settings. Peptide functionalization: Peptides self-assemble on MoS2 surfaces into β-tape monolayers, conferring positive (KFKFKFKF) or negative (EFEFEFEF) surface charge and stabilizing aqueous dispersions for months. For oppositely charged composites, cationic and anionic peptide-coated MoS2 suspensions were mixed and allowed to react for ~20 min prior to membrane formation. Membrane fabrication: Laminate membranes (LMs) were prepared by vacuum filtration of aqueous suspensions (typically 3 mL at 6.2 ± 0.5 mg mL−1) through Anodisc alumina supports (0.02 μm pores; 25 or 47 mm diameter), yielding ~1 μm thick laminates confirmed by cross-sectional SEM. After baking at 50 °C for 12 h, an additional 1 mL suspension was filtered to restore surface uniformity. Membranes included: non-porous NS LMs; NS LMs with added NDs; porous NSND LMs with different pore size distributions (<10, <25, <60 nm); and peptide-decorated variants [pep(+), pep(−), and mixed pep(+)/pep(−)]. Characterization: AFM imaged NS thickness (1–2 layers: ~0.7 and 1.4 nm). XRD assessed interlayer spacing and stacking order; HAADF-STEM measured interlayer spacings and visualized stacking faults, pores, and voids; SEM imaged cross-sections. Zeta potential measured surface charge. Contact angle assessed wettability. Fouling tests used BSA (0.5 g L−1) in loop filtration cycles. Performance testing: Forward osmosis (FO) tests used 0.5 M NaCl feed and 2 M sucrose draw; ion concentrations were monitored over time to calculate rejection. Long-term FO stability was evaluated over 5–30 days, including multiple salts relevant to seawater (NaCl, KCl, MgCl2, Na2SO4, etc.). Reverse osmosis (RO) tests (dead-end configuration) evaluated NaCl (0.5 M) rejection and derived intrinsic water permeability (Lp) and salt permeability (B). Organic pollutant rejection in RO mode was tested at 1 bar for dyes with known hydrated radii: methyl red (neutral, Rh = 4.87 Å), methyl orange (negative, Rh = 4.97 Å), methylene blue (positive, Rh = 5.04 Å), and rhodamine B (neutral, Rh = 6.15 Å). Chlorine susceptibility was evaluated by exposing membranes to 10,000 ppm active chlorine (NaOCl) and tracking rejection, water permeability, and mass loss pre/post 1 h exposure.

• Structural control and assembly: The two-step cavitation process yields 1–2 layer porous NSs (thickness peaks at ~0.7 and 1.4 nm) and NDs; pore and disk diameters are tunable via sonication conditions. HAADF-STEM shows interlayer spacing ~6.2 Å in laminates without peptide; peptide-modified thin laminates show average interlayer spacing 7.8 ± 1.6 Å and undulating gaps indicative of peptide intercalation. XRD indicates peptide decoration narrows and sharpens the (002) peak and slightly increases interlayer spacing, consistent with more ordered stacking; re-wetted peptide-decorated membranes show ~0.17 nm larger interlayer spacing than porous MoS2 without peptide.
• Salt rejection in FO: Non-porous MoS2 NS laminates and bare Anodisc supports showed low NaCl (0.5 M) rejection (<18%). Introducing porosity markedly improved rejection: porous MoS2 <60 nm pores >57% NaCl rejection; porous MoS2 <10 nm pores ~80%. Incorporating oppositely charged peptides [pep(+), pep(−)] into porous NSND-LMs increased NaCl rejection to >99% at 0.5 M. Across several salts at 0.5 M, rejection remained >98% after 5 days, following steric trends with hydrated radius (K+ < SO4^2− < Cl− < Na+ < Mg^2+). Below 0.5 M NaCl, rejection was >99.99% after 7 days.
• Water transport: Adding NDs to non-porous NS laminates increased water permeance ~4×, indicating ND-induced nanochannels enhance transport. In porous NSND-LMs, larger pore diameters increased water permeance (e.g., from 432 ± 25 to 603 ± 38 LMH/bar when increasing characteristic pore size from <10 to <60 nm), while mixed-peptide functionalization preserved high rejection (>99%) with some reduction in flux due to tighter assembly and increased interfacial charge.
• Long-term stability: Continuous FO with 0.5 M NaCl for >30 days showed stable performance with >95% rejection and steady water permeance, indicating minimal clogging and robust laminate assembly.
• Benchmarking: FO water permeance of the peptide-modified porous NSND-LM was ~5 L m−2 h−1 (1 μm thickness), exceeding commercial TFC-PA SW30 tested in the same setup (~0.9 L m−2 h−1) by ~6×, and surpassing functionalized GO (~0.3 L m−2 h−1, 0.5 μm) and epoxy-encapsulated GO (~0.5 L m−2 h−1, 5 μm) by ~10× and ~17×, respectively.
• RO mode performance: In dead-end RO with 0.5 M NaCl, porous MoS2 membranes showed 63 ± 12% rejection, lower than FO (>99%), consistent with known pressure-induced selectivity reductions in laminar 2D membranes. Despite this, the B–Lp tradeoff compared favorably with literature membranes and a commercial SW30HR tested under similar conditions.
• Organic contaminant removal (RO, 1 bar): Nearly 100% rejection per pass for methyl red (Rh = 4.87 Å), methyl orange (Rh = 4.97 Å), methylene blue (Rh = 5.04 Å), and rhodamine B (Rh = 6.15 Å).
• Fouling resistance: With BSA (0.5 g L−1), average water flux decreased from 232 ± 8.92 to 198 ± 14.8 LMH/bar; flux recovery after rinsing was 96 ± 2%, attributed to charged, smooth, and hydrophilic surfaces (surface roughness ~1.88 nm; low contact angles).
• Chlorine exposure: At 10,000 ppm active chlorine for 1 h, commercial SW30HR lost 5.8% mass and degraded rapidly in performance; porous MoS2 lost 3.4% mass with slower performance degradation, indicating better chlorine tolerance though some oxidation is expected.

The composite laminate design combines through-pores within nanosheets, finite sheet lengths, and interspersed nanodisks acting as spacers to create a highly interconnected network of sub-nanometer pathways. This multimodal porous architecture shortens and multiplies water transport routes while enhancing ion and small-molecule exclusion via steric hindrance and electrostatic interactions at charged pore edges and intersheet interfaces. Peptide self-assembly tunes surface charge and promotes tighter, more ordered stacking, increasing interlayer charge density and reducing defects that would otherwise permit ion leakage, leading to FO NaCl rejection above 99% even at high salinity with sustained performance over a month. The observed decrease in RO rejection relative to FO is consistent with pressure-driven effects that weaken water–ion interactions and potentially compact nanochannels; nevertheless, the intrinsic permeability–selectivity balance compares favorably with state-of-the-art membranes. The membranes effectively reject both monovalent and divalent salts and small organic pollutants, and exhibit good resistance to fouling and improved chlorine tolerance relative to commercial TFC-PA. Collectively, the structural engineering of pore size, interlayer spacing, and surface charge addresses the central challenge of achieving both high water throughput and high selectivity in stable, scalable 2D laminar membranes.

This work introduces a straightforward, scalable cavitation-based route to produce porous MoS2 nanosheets and nanodisks, and assembles them into peptide-tuned laminate membranes with controllable pore size, surface charge, and interlayer spacing. The resulting NSND laminates achieve ultrahigh water permeance with very high ion and small-molecule rejection, maintain >95% NaCl rejection over month-long FO operation, effectively reject common seawater ions and organic dyes, and demonstrate favorable permeability–selectivity tradeoffs in RO mode with improved chlorine tolerance and antifouling behavior. The approach integrates benefits of ultrathin porous single-layer and stacked laminar membranes while addressing stability and scalability, suggesting strong potential for commercial water purification. Future research should optimize RO performance under pressure, further enhance chlorine resistance and oxidative stability, explore long-term fouling and cleaning protocols, and generalize the NSND-peptide strategy to other transition metal dichalcogenides and mixed-matrix architectures.

RO rejection in dead-end mode (63 ± 12% for 0.5 M NaCl) is lower than FO performance (>99%), likely due to pressure-induced effects such as reduced water–ion selectivity, concentration polarization, nanochannel collapse, and membrane compaction. While chlorine tolerance is better than commercial TFC-PA, MoS2 still shows some mass loss (3.4% at 10,000 ppm, 1 h) due to oxidation, indicating the need for further stabilization for harsh disinfection environments. Reported water permeance values depend on membrane thickness (~1 μm) and peptide-driven assembly; variability in stacking defects and interlayer alignment may influence batch-to-batch performance. Long-term stability beyond one month, performance under complex real-water matrices, and scale-up to large-area modules require additional validation.