Where Nanofiltration Sits Among Pressure-Driven Membranes
A nanofiltration (NF) system separates dissolved species using a thin-film composite membrane with pore sizes of 1–10 nm and a molecular weight cut-off of 100–2000 Da, operating at 2–30 bar. Rejection is driven by three coupled mechanisms — size sieving, Donnan (electrostatic) exclusion, and dielectric exclusion — which together retain divalent ions such as SO₄²⁻ (up to 99%) and Ca²⁺ (up to 81%) while letting monovalent salts pass. NF occupies the middle band between ultrafiltration (UF) and reverse osmosis (RO), providing partial desalination and hardness removal at a lower energy cost than RO.
The Springer 2025 review on NF in wastewater treatment defines the membrane class by its 1–10 nm pore size and 100–2000 Da MWCO (source: Springer, 2025-04). UF sits above it with pores of 2–100 nm and MWCO of 1,000–500,000 Da, removing colloids, bacteria, proteins, and oil emulsions but not ions. RO sits below with sub-nm pores and near-total salt rejection, including monovalent Na⁺ and Cl⁻. NF is the only one of the three that delivers a meaningful rejection gap between divalent and monovalent ions. Operating pressure tracks the same order: UF runs at 1–10 bar, NF at 2–30 bar, RO at 10–80 bar. Templated polyamide (PA) NF membranes have been demonstrated at only 2 bar with 90 LMH flux (source: Nature Communications, 2018-06, industrial RO system reference for the high-pressure end of the family).
| Parameter | Ultrafiltration (UF) | Nanofiltration (NF) | Reverse Osmosis (RO) |
|---|---|---|---|
| Pore size | 2–100 nm | 1–10 nm | < 1 nm (dense) |
| MWCO | 1,000–500,000 Da | 100–2000 Da | < 200 Da |
| Operating pressure | 1–10 bar | 2–30 bar | 10–80 bar |
| Rejects divalent ions (SO₄²⁻, Ca²⁺) | No | Yes (60–99%) | Yes (≥ 99%) |
| Rejects monovalent ions (Na⁺, Cl⁻) | No | Partial (10–50%) | Yes (≥ 99%) |
| Removes organics / dyes / pharmaceuticals | Partial (high MW only) | Yes (above MWCO) | Yes |
| Typical placement in a reuse train | After biological / clarifier | After UF, before RO or reuse | Polishing or ZLD pre-concentration |
Engineers select NF to remove hardness, sulfates, dyes, pesticides, and most dissolved organics while retaining monovalent salts in the permeate. If the discharge or reuse limit is total dissolved solids (TDS), RO is required. In a typical industrial reuse train — biological → UF/MBR → NF → RO polish — the NF step drops scaling potential and extends RO membrane life by 2–4×.
The Three Separation Mechanisms Inside an NF Membrane
NF rejects solutes through three coupled mechanisms — size sieving, Donnan (electrostatic) exclusion, and dielectric exclusion — with the dominant mechanism shifting based on solute type, charge, and feed concentration. The accepted engineering framework that combines all three is the DSPM-DE model (Donnan-steric-pore model with dielectric exclusion), which uses the extended Nernst-Planck equation to predict ion flux as the sum of diffusion, convection, and electromigration through the membrane pore (source: Springer, 2025-04).
Size sieving (steric exclusion) blocks solutes whose hydrated diameter exceeds the membrane's effective pore diameter. The 100–2000 Da MWCO threshold (source: Springer, 2025-04) defines this limit, where solutes above the cut-off are rejected regardless of charge. This mechanism dominates for uncharged organics, large dye molecules, humic substances, pesticides, and pharmaceuticals with molecular weights above ~300 Da. Even at the lower end of the NF MWCO range, a membrane with a 200 Da cut-off rejects most textile dyes, which typically range from 600–1500 Da.
Donnan exclusion occurs because most commercial NF membranes carry a net negative surface charge in aqueous solution. Co-ions — anions such as Cl⁻ and SO₄²⁻ in a NaCl/Na₂SO₄ feed — are electrostatically repelled by the membrane; counter-ions like Na⁺ are partially rejected to maintain electroneutrality. Divalent co-ions are repelled more strongly than monovalent co-ions, allowing NF to reject SO₄²⁻ more effectively than Cl⁻. The Xinjiang No. 1 Oil Production Plant study measured maximum rejections of 99% for SO₄²⁻, 81% for Ca²⁺, and 94% for Mg²⁺ on synthetic divalent feed (source: Membranes/PMC, 2025). Rejection falls as feed ion concentration rises because elevated ionic strength screens the membrane's fixed surface charge, reducing the Donnan potential.
Dielectric exclusion is a quantitatively important mechanism where water's dielectric constant drops inside the nanoconfined pore, raising the Born solvation energy barrier for an ion crossing the membrane. Ions with higher charge density — Al³⁺, Mg²⁺, SO₄²⁻ — face a stronger penalty than Na⁺ or Cl⁻, so dielectric exclusion amplifies the divalent-over-monovalent selectivity generated by Donnan effects (source: Springer, 2025-04). In the DSPM-DE model, this is captured through the ratio of pore-water dielectric constant (εₚ) to bulk dielectric constant (ε_b) as an additional partitioning barrier at each membrane-solution interface.
Rejection rises with applied pressure as flux and driving force increase, but falls as feed concentration rises due to charge screening. The Xinjiang results confirm this: when feed divalent-ion concentration increased, Ca²⁺ and Mg²⁺ rejection dropped while SO₄²⁺ rejection rose slightly, reflecting the interplay between concentration-dependent screening and electrostatic repulsion.
A Worked Example: Softening Oilfield Produced Water with NF
The Xinjiang No. 1 Oil Production Plant case serves as a benchmark for NF performance, pairing a real industrial stream with a novel membrane to report rejection numbers. The feed is produced water rich in Ca²⁺, Mg²⁺, and SO₄²⁻, with a scaling tendency that fails reinjection standards. Researchers fabricated a negatively charged NF membrane via interfacial polymerization using 2-carboxypiperazine and trimesoyl chloride, characterizing it by SEM, XPS, and FTIR (source: Membranes/PMC, 2025).
On synthetic single-salt feed, maximum rejections reached 99% for SO₄²⁻, 81% for Ca²⁺, and 94% for Mg²⁺. On real produced water, the membrane delivered 100% SO₄²⁻ removal, 91% Ca²⁺ removal, and 95% Mg²⁺ removal. The scaling tendency of the treated effluent was eliminated, rendering the water directly reusable for reinjection (source: Membranes/PMC, 2025). These results demonstrate the synergy of Donnan exclusion, surface adsorption, and mass transfer resistance under applied pressure.
High-hardness, high-sulfate industrial wastewater streams are ideal candidates for NF application. Mining leachate, coal-to-chemical brine requiring gypsum-scaling control, RO concentrate requiring hardness removal, and textile effluent with reactive dyes all map onto this mechanism profile. If water quality issues focus on hardness and sulfate rather than bulk TDS, NF provides a more efficient economic solution than RO.
How a Complete NF System Is Built Around the Membrane
An NF skid consists of a high-pressure feed pump (centrifugal for flows above 20 m³/h; multistage centrifugal or positive-displacement for lower flows), cartridge filters or strainers rated at 5 µm, FRP or stainless-steel pressure vessels holding 2–6 spiral-wound 4040 or 8040 elements in series, pressure and flow transmitters, a concentrate control valve, and a clean-in-place (CIP) skid (source: Springer, 2025-04).
Pretreatment is essential to maintain system integrity. Suspended solids above the silt density index (SDI) limit of ~3 blind the feed spacer, oil and grease above ~0.1 mg/L foul the polyamide layer, and free chlorine above 0.1 mg/L oxidizes the selective layer. A typical pretreatment chain is a multi-media pretreatment filter followed by a 5 µm cartridge guard, sodium bisulfite dosing for chlorine reduction, and antiscalant injection. For streams with biological fouling potential, an upstream UF or MBR stage is standard — for example, an MBR bioreactor producing SDI < 2 permeate feeds directly into NF.
The selective layer is thin-film composite polyamide, typically made by interfacial polymerization of piperazine (PIP) or 2-carboxypiperazine with trimesoyl chloride (TMC) on a polysulfone (PSf) or polyethersulfone (PES) support. Vessels are FRP for brackish duty or 2205/2507 duplex stainless for higher-pressure or chloride-rich service. Emerging materials, such as PA with graphene oxide (GO), TiO₂ nanofillers, MoS₂ interlayers, cellulose nanocrystals, and g-C₃N₄-regulated IP layers, aim to improve flux and antifouling properties (source: Springer, 2025-04).
Energy, Operating Cost, and Where NF Beats RO
NF operates at 2–30 bar versus 10–80 bar for RO, with specific energy demand typically 30–60% lower than RO, depending on feed salinity and recovery. Templated PA NF has been demonstrated at 90 LMH flux with only 2 bar applied pressure — an order of magnitude lower than a typical brackish RO system running at 15 bar (source: Nature Communications, 2018-06). NF is the primary choice for water softening and partial desalination where full RO is unnecessary.
Because NF does not reject monovalent salts (NaCl rejection is 10–50% versus > 99% for RO), a stream requiring TDS reduction still needs RO. Placing NF as a softener and partial desal stage after UF/MBR and before RO extends RO membrane life by 2–4× and reduces cleaning frequency. For COD-heavy streams, NF pairs with biological treatment and upstream separation; see the broader picture in our COD removal technologies guide.
Frequently Asked Questions
What is the difference between NF pore size and MWCO?
Pore size (1–10 nm) describes the physical opening in the polyamide selective layer, while molecular weight cut-off (100–2000 Da) is the operational definition—the molecular weight of a reference solute at which the membrane achieves 90% rejection. Pore size indicates geometry; MWCO indicates the practical separation boundary (source: Springer, 2025-04).
Why does NF reject divalent ions much better than monovalent ions?
Rejection is achieved through Donnan exclusion (negatively charged membranes repel divalent co-ions like SO₄²⁻ more strongly than Cl⁻) and dielectric exclusion (higher charge density ions face a larger Born solvation energy penalty). Together, these mechanisms push SO₄²⁻ rejection above 99% while NaCl rejection remains at 10–50% (source: Springer, 2025-04).
How much less energy does NF use compared to RO?
NF operates at 2–30 bar versus 10–80 bar for RO, resulting in 30–60% lower specific energy demand per m³ of permeate. Advanced templated PA NF has been demonstrated at 90 LMH flux with only 2
