Effect of solution ions on the charge and performance of nanofiltration membranes

The study investigates how solution ions influence the effective charge and separation performance of polyamide nanofiltration (NF) membranes, focusing on Donnan exclusion. While NF membranes separate solutes via size exclusion and molecular-level effects, the contribution of membrane charge—arising from carboxyl and amine functional groups—to selectivity remains complicated by ion-specific interactions. Prior observations have shown inconsistencies in interpreting zeta potential (ZP) and rejection vs pH behavior, indicating that membrane charge is not intrinsic but depends on solution composition. The purpose here is to systematically evaluate ZP across a wide range of pH, ionic strength, and ionic composition and to correlate these measurements with ion permeability, elucidating ion-specific adsorption ("stickiness") effects—particularly of chloride and monovalent cations—on membrane effective charge and Donnan exclusion. This work aims to clarify discrepancies in the literature and guide the design of NF membranes with improved ion-specific selectivity.

Zeta potential measured by streaming potential is commonly used to infer membrane surface charge. However, literature reports show discrepancies: the minimum in salt rejection vs pH curves is often equated to the isoelectric point, yet ZP is frequently negative at that pH for commercial NF membranes; and the effect of ionic strength on ZP is reported inconsistently, with some studies observing increased negative charge at higher salinity and others the opposite. These discrepancies suggest that membrane charge is influenced by solution ions beyond fixed-group density and pKa, including ion-specific adsorption related to hydration and polarizability ("stickiness"). Prior studies also highlight differences in transport between similar ions (e.g., Li+ vs Cs+) that cannot be explained by size/valence alone, implicating molecular-level interactions and interfacial phenomena. The paper builds on this literature to resolve conflicting interpretations by explicitly testing ion identity, ionic strength, and pH and by correlating ZP with transport.

Materials: Thin-film composite polyamide NF270 (Dow Filmtec) membranes. Single-salt solutions of LiCl, NaCl, KCl, CsCl were prepared in ultrapure/DI water. pH was adjusted with 0.05 M HCl/NaOH. Membrane coupons were pretreated in 25% v/v isopropyl alcohol (30 min) and rinsed with DI (30 min, repeated three times). Each coupon was used for a single ZP or filtration experiment, except for the temperature-cycling test. Zeta potential measurements: Streaming potential was measured using a SurPASS3 analyzer. ZP (ζ) was calculated via the Helmholtz–Smoluchowski equation: ζ = εrε0R/(4πηAL) · dUstr/dp. Four chloride salts (LiCl, NaCl, KCl, CsCl) at 0.1–75 mM were tested. For each solution, ZP was stabilized at initial pH (~6), then pH was raised to ~9 with NaOH and decreased stepwise to ~2.5, measuring ZP at ~0.5 pH intervals. Conditions: 23 ± 3 °C; continuous CO2 degassing by N2 bubbling; channel gap 100 ± 10 μm. Quality criteria: streaming potential vs pressure linearity correlation >0.989, |Asymmetry| <10 mV, static resistance <10 kΩ. X-ray photoelectron spectroscopy (XPS): Dow NF270 samples were soaked 24 h in DI water or NaCl solutions (1 mM and 10 mM) at 40 °C, then rinsed and soaked in DI at room temperature for ≥6 h; dried overnight. Spectra were acquired on a Kratos Supra (Al Kα, 1486.69 eV, 15 mA). Survey spectra (0–1200 eV, resolution 160) were used for surface atomic percent; high-resolution scans at resolution 20. Permeability and rejection tests: Crossflow setup with a 10 L feed tank, high-pressure pump, and three stainless steel cells (15.3 cm² each). Reject and permeate were recirculated to the feed tank. Operating parameters (pressure, crossflow velocity, pH, temperature) were maintained constant; stabilization 30 min after changes. Membranes were compacted overnight in DI at ≥2 bar above operating pressure. Typical conditions for permeability comparison: ΔP = 5 bar, 24.8 ± 0.1 °C, crossflow velocity 2.13 m s⁻¹. A separate temperature-cycling test at 1 mM NaCl, 18 bar, 2.13 m s⁻¹, temperature ramped 20 → 40 °C and repeated for three cycles to probe irreversible chloride adsorption. Analytical and calculations: Single-salt feeds; feed and permeate concentrations from conductivity; permeate mass used to compute water flux Jw; salt flux Js = Jw·Cp. Concentration polarization accounted for using film theory: Cm = (Cf − Cp)exp(Jw/k)+Cp, with k from Sherwood correlations for turbulent rectangular channels. Water and salt permeabilities were computed using the solution–diffusion model: Jw = Pw(ΔP − Δπ) and Js = Ps(Cm − Cp). Ion properties (hydrated radii, hydration enthalpy, polarizability) used for interpretation. ZP–permeability correlations were examined at selected pH and ionic strengths for LiCl and CsCl.

- Fixed-group pKa on NF270: Carboxylic groups pKa ≈ 4.5; amine groups pKa estimated 8.5–9.0. ZP vs pH exhibits a steep increase in negative charge around pH 4.5 with plateau above ~6.5, consistent with carboxyl deprotonation; slight decrease in ζ (more negative) above pH ~8.5 suggests amine deprotonation at lower density than carboxyls. - Rejection vs pH does not directly mark the isoelectric point: Minimum NaCl rejection occurs near the carboxyl pKa (~4.5) rather than at the pH of minimal total charged groups (<3). At pH < pKa, increased swelling can dominate over charge effects, reducing rejection as pH rises from 3.0 to 4.5. - Negative ζ at low pH explained by chloride adsorption: Despite expectations of near-zero/slightly positive ζ at pH ~3.5 (given low carboxyl deprotonation and low amine density), ζ remained clearly negative, attributed to adsorption of highly polarizable, weakly hydrated Cl− to neutral hydrophobic polyamide segments. - XPS evidence for ion adsorption: Chlorine surface atomic fraction increased with pre-soak salinity: 5.3% (DI), 9.3% (1 mM NaCl), 16% (10 mM NaCl), indicating ion sorption within/on the membrane. - Irreversible chloride adsorption at elevated temperature: In three sequential temperature cycles (20→40 °C) at 1 mM NaCl and 18 bar, salt permeability at a given temperature decreased in subsequent cycles, consistent with irreversible Cl− adsorption increasing negative charge and Donnan exclusion; water permeability showed no systematic change. - Ionic strength and composition effects on ζ: For I > 1 mM, increasing salt concentration reduced |ζ| due to charge screening. For I < 1 mM, increasing I led to more negative ζ, consistent with Cl− adsorption dominating at low salinity. At pH ~2.6, ζ decreased monotonically with I, as both screening and Cl− adsorption act in the same direction. - Ion-specific cation effects (stickiness): At moderate concentrations (~0.5–1 mM) and pH > ~6, ζ was less negative with CsCl than LiCl, attributed to Cs+ having lower hydration enthalpy (softer hydration, higher stickiness) and greater adsorption/neutralization of negative sites; at high concentration (50 mM), screening dominates and ζ converges for LiCl and CsCl. - ZP correlates inversely with salt permeability: Conditions with larger |ζ| exhibited lower salt permeability, supporting ZP as an indicator of Donnan exclusion. Quantitatively, LiCl/CsCl ratios showed close correspondence between ζ and permeability ratios: e.g., at pH 4.0 and 1 mM, rP ≈ 1.08 with rζ ≈ 0.99; at pH 7.5 and 50 mM, rP ≈ 0.94 with rζ ≈ 1.00; at pH 7.5 and 1 mM, rP ≈ 0.74 with rζ ≈ 1.17, reflecting stronger Donnan effects and ion-specific interactions at low salinity and higher pH.

The findings demonstrate that membrane effective charge—and thus Donnan exclusion—is governed not only by the density and ionization (pKa) of fixed carboxyl and amine groups but also by ion-specific adsorption from solution. Chloride’s high polarizability enables adsorption to neutral hydrophobic polyamide segments, accounting for negative ζ even at low pH and for increased |ζ| with increasing dilute chloride concentration until adsorption site saturation. Monovalent cations differ in their ability to neutralize negative charge according to their hydration and stickiness, with Cs+ more prone than Li+ to adsorb/screen membrane charges and reduce |ζ| at moderate salinity. These ion-specific interactions reconcile conflicting prior reports regarding the impact of ionic strength on ZP and the interpretation of rejection vs pH curves. The strong inverse correlation between |ζ| and salt permeability across conditions supports ZP as a practical predictor of Donnan exclusion and transport behavior, especially at pH ≤ 7.5 where carboxyl deprotonation controls fixed charge. Overall, incorporating solution-ion adsorption into models of membrane charge clarifies observed performance trends and aids the rational tuning of NF separations.

This work resolves key inconsistencies in interpreting zeta potential and rejection behavior of polyamide NF membranes by explicitly accounting for solution-ion adsorption. It identifies the effective pKa of fixed groups on NF270 (carboxyl ~4.5; amine ~8.5–9.0), shows that chloride adsorption increases negative effective charge (particularly at low salinity and elevated temperature), elucidates how ionic strength transitions from adsorption-dominated to screening-dominated regimes, and demonstrates subtle but meaningful cation-specific effects (Li+ vs Cs+) tied to ion stickiness. A clear inverse relationship between |ζ| and salt permeability establishes ZP as a robust indicator of Donnan exclusion. These insights improve understanding of ion-specific interactions at membrane interfaces and can guide the design and operation of NF systems for enhanced ion selectivity. Potential future directions include extending analyses to other anions and multivalent ions, mixed-salt systems, and additional membrane chemistries to generalize the ion-specific adsorption framework and further link effective charge to transport selectivity.

- The exact pKa of amine groups could not be precisely determined due to their low surface density and technical challenges above pH ~10; estimated between 8.5 and 9.0. - Correlation between ZP and permeability was evaluated primarily for pH ≤ 7.5 (where carboxyl deprotonation dominates) and for chloride-based single-salt solutions; generalization to other ions and mixed electrolytes requires further study. - Most experiments focused on a single membrane type (NF270), which may limit generalizability to other polyamide chemistries and morphologies. - ZP reflects effective charge (including contributions from adsorbed solution ions) rather than fixed charge alone; interpretation depends on solution composition and ionic strength. - Irreversible chloride adsorption at elevated temperature alters membrane charge-state history, which may complicate comparisons across thermal exposure histories. - Although carbonate/water ions were considered and found to have minor effects, measurements were performed with CO2-degassed solutions; different gas equilibria could slightly alter local pH near the membrane.