Nanofiltration: How It Works, Membrane Materials, and Water Treatment Uses

In a nanofiltration (NF) system, a feed stream is pushed against a thin polymer or composite membrane at pressures typically between 5 and 25 bar. Water and small, mostly monovalent ions permeate through sub-nanometre channels, while multivalent ions, dissolved organic matter with molecular weights above a few hundred grams per mole, dyes, sugars, and many pharmaceutical and pesticide residues are retained. The combination of nanopores sized close to a hydrated ion and a charged membrane surface gives NF its characteristic ability to discriminate between solutes that differ only in valence or in molecular shape.

The term nanofiltration was coined in 1984 at the FilmTec Corporation to describe a class of "loose reverse osmosis" membranes that selectively passed monovalent salts, and was formalized in the membrane literature in 1988 by Cadotte, Eriksson and others. Modern NF membranes are almost always composite structures: a roughly 20 to 200 nanometre thin film of crosslinked polyamide is formed on a porous polysulfone support, which in turn rests on a polyester nonwoven backing. Because the active layer is so thin, careful engineering of its chemistry, roughness, and charge density at the nanoscale determines almost everything about how the membrane performs.

Three transport mechanisms operate in parallel inside an NF membrane and together determine which solutes pass and which are held back. Size, or steric, exclusion blocks any species whose effective hydrated radius is larger than the channels in the active layer; this is why NF membranes are usually rated by a molecular weight cut-off in the 100 to 1000 g/mol range. Charge, or Donnan, exclusion arises because the polyamide surface and the inside of its channels carry fixed charges – usually negative at neutral pH due to deprotonated carboxyl groups – so co-ions of the same sign are repelled, dragging counter-ions with them to maintain electroneutrality. Dielectric exclusion becomes important inside very narrow channels, where the lower local permittivity of confined water raises the energy cost of bringing in a charged species.

Across a typical aromatic polyamide NF membrane, divalent salts such as Na2SO4 or MgSO4 are rejected at 95 to 99 percent, while NaCl rejection commonly falls between 20 and 70 percent depending on the membrane chemistry and the salt concentration. Neutral organics are screened mainly by size: rejection rises steeply once molecular weight exceeds the cut-off, and pharmaceutical molecules of 200 to 400 g/mol are typically rejected at 70 to 95 percent. Water permeance, the volumetric flux per unit pressure, is set by the thickness, porosity, and hydrophilicity of the active layer; commercial membranes generally deliver 5 to 15 L m−2 h−1 bar−1, while the best laboratory designs reach 50 to 100 L m−2 h−1 bar−1 with comparable salt rejection.

A central challenge in NF design is the inverse relationship between water permeance and solute selectivity. Loosening the active layer to admit more water also widens its channels, which lets unwanted solutes through. The "upper bound" describing this trade-off has been pushed back repeatedly over the last decade through routes such as crumpling the polyamide film to increase its effective area, depositing it on a nanostructured interlayer, and reducing its thickness to a few nanometres. Two-dimensional materials and porous frameworks offer further routes around the bound by replacing tortuous polymer paths with engineered nanochannels of well-defined size.

NF is best understood by comparing it with its neighbours in the pressure-driven membrane family. Microfiltration (MF) and ultrafiltration (UF) act mainly as size sieves with much larger pores; reverse osmosis (RO) is denser still than NF and rejects almost all dissolved species. NF is the only one of these four that combines sub-nanometre size sieving with pronounced charge-based selectivity.

In practice, NF and RO often share infrastructure: spiral-wound modules, polyamide thin-film composite chemistry, and similar fouling-control strategies. The advantage of NF over RO when full desalination is not required is energy: lower operating pressures translate directly into lower specific energy consumption, which can drop NF energy use to less than half that of seawater RO for comparable feedwaters.

No single material dominates research the way titanium dioxide dominates photocatalysis, but commercial NF is overwhelmingly built on aromatic and semi-aromatic polyamide chemistries. Newer architectures based on graphene oxide, MXenes, metal-organic frameworks, and covalent organic frameworks are advancing rapidly in the laboratory and beginning to enter the market for niche applications.

The dominant commercial architecture is the polyamide thin-film composite (TFC) membrane introduced by John Cadotte's team at FilmTec in the 1970s and 1980s. A porous polysulfone or polyethersulfone support is wetted with an aqueous diamine such as piperazine or m-phenylenediamine, then briefly contacted with a non-aqueous solution of trimesoyl chloride. Polymerization happens almost instantaneously at the liquid–liquid interface, depositing a 20 to 200 nm crosslinked polyamide film with a characteristic ridge-and-valley topography. Tuning the monomer pair, additives, reaction time, and post-treatment yields membranes spanning the full NF and RO range from the same basic platform.

Stacking graphene oxide nanosheets into a laminate creates interlayer galleries that act as two-dimensional channels for water and ions. Pioneering work in 2014 demonstrated that micrometre-thick graphene oxide laminates could sieve hydrated ions and small molecules with sharp molecular-weight cut-offs, while remaining vacuum-tight when dry. Closely related laminates of MXene flakes, transition-metal dichalcogenides, and boron nitride nanosheets follow the same principle. Controlling the gallery spacing under wet conditions and preventing swelling are the central engineering challenges.

Metal-organic frameworks and covalent organic frameworks are crystalline porous materials with monodisperse pores in the sub-nanometre to a few-nanometre range. Polycrystalline films of MOFs and COFs, or composite membranes with these materials embedded in a polymer matrix, can in principle deliver sharper size selectivity than statistically distributed polymer free volume. Practical NF MOF and COF membranes still face challenges around defect-free synthesis at large area, mechanical durability under cross-flow, and chemical stability in aqueous service, but performance has improved markedly since 2020.

Ceramic NF membranes built on titania, zirconia, alumina, or silica supports are robust to high temperature, extreme pH, and aggressive solvents, which makes them attractive for industrial process streams and for organic solvent nanofiltration. Their permeance is generally lower than that of state-of-the-art polyamide TFC membranes, and their capital cost is higher, but their long service life under conditions that would destroy a polyamide membrane is often decisive. Carbon nanotube and silica-pillared nanoporous membranes are research extensions of this family.

Interfacial polymerization remains the workhorse of NF fabrication because it self-limits the active-layer thickness, scales easily to roll-to-roll production, and tolerates a wide range of monomer chemistries. Recent variants include adding nanofillers such as nanofibers, MOFs, or carbon nanostructures to the casting solution to form thin-film nanocomposite (TFN) membranes; using sacrificial nanoparticle templates to crumple the polyamide film and increase its effective area; and inserting a porous nanostructured interlayer between the support and the polyamide to decouple film formation from substrate roughness.

Beyond interfacial polymerization, NF membranes are also produced by phase inversion of polyelectrolytes, layer-by-layer deposition of charged polymers, vacuum filtration of 2D-material dispersions, electrospinning of nanofibrous mats followed by densification, and atomic layer deposition of inorganic skins on porous supports. Each route trades off ease of manufacture, defect tolerance, and the kind of selective layer it can produce. Surface treatments after fabrication – surface functionalization with hydrophilic polymers, zwitterionic brushes, or biocides – tune charge, fouling resistance, and antibacterial activity without changing the underlying separation chemistry.

The original commercial use of NF was the removal of calcium and magnesium hardness from drinking water, replacing or supplementing ion-exchange softeners. Because Ca2+, Mg2+, and SO42− are rejected at 90 percent or higher while most NaCl passes through, the product water is soft without becoming over-mineralized. NF is also widely used to remove dissolved natural organic matter that causes colour, taste, and disinfection-by-product precursors in surface waters, reducing downstream chlorine demand.

Pharmaceuticals, personal-care products, pesticides, and per- and polyfluoroalkyl substances (PFAS) appear in surface and reclaimed water at trace concentrations and resist conventional treatment. NF rejects neutral and ionic micropollutants above roughly 200 to 300 g/mol with efficiencies of 70 to 99 percent, depending on charge, hydrophobicity, and membrane chemistry. Combined with biological treatment and oxidation steps, NF is increasingly central to indirect potable reuse schemes and to industrial wastewater polishing.

Placing NF upstream of RO removes hardness ions, sulfate, silica precursors, and dissolved organic foulants before they reach the tighter membrane. This reduces scaling and biofouling on the RO stage, allows higher recovery, and lowers cleaning frequency. NF–RO trains are now standard in seawater desalination, brackish-water reuse, and zero-liquid-discharge industrial water systems.

A growing application of NF is solute–solute separation rather than water–solute separation. The contrast between the high rejection of divalent Mg2+ and the low rejection of monovalent Li+ on positively charged NF membranes is the basis of magnesium–lithium separation in salt-lake brines, an attractive route to lithium for battery manufacturing. Similar logic applies to recovering nutrients from agricultural runoff, separating mono- and divalent cations in mining liquors, and concentrating valuable acids or bases in chemical processing. Performance metrics designed for water production are not always appropriate for these tasks, and the field is actively developing new evaluation frameworks.

When the solvent is not water but ethanol, hexane, dimethylformamide, or another organic liquid, the process is called organic solvent nanofiltration (OSN). Solvent-stable polyimide, polybenzimidazole, and crosslinked PEEK membranes, along with ceramic and graphene-oxide variants, allow molecular-weight-based separation in pharmaceutical purification, edible oil refining, lubricant dewaxing, and homogeneous catalyst recovery. OSN can replace energy-intensive distillation steps and is one of the fastest-growing branches of nanofiltration outside of water treatment.

Fouling is the dominant practical limitation of NF. Particulate fouling deposits suspended solids on the membrane surface; organic fouling adsorbs natural organic matter and proteins; biofouling forms biofilms that thicken over months; scaling precipitates sparingly soluble salts as the rejected stream is concentrated. Each lowers permeance, raises operating pressure, and shortens membrane life. Mitigation combines feed pretreatment, cross-flow operation to limit cake build-up, periodic cleaning with acid and base solutions, antiscalant dosing, and the design of more hydrophilic, charge-tuned membrane surfaces. The zeta potential of the membrane surface is a useful predictor of fouling tendency for charged foulants.

Beyond fouling, NF membranes have a finite tolerance to oxidants such as free chlorine, which slowly degrades aromatic polyamide; to extreme pH, which limits cleaning options; and to high temperature, which softens the polymer support. The unavoidable tension between permeance and selectivity is sharpened by these chemical constraints, since the most attractive routes to higher permeance often involve thinner, less crosslinked active layers that are also more vulnerable to attack. Brine disposal from NF concentrate streams is a further environmental issue shared with all desalination technologies and is the focus of growing regulatory attention.

The next generation of NF membranes is being shaped by three converging trends. Ultrathin polyamide films of just a few nanometres, often built on engineered nanofabricated supports, push water permeance well above the historical upper bound while maintaining commercial salt rejection. Two-dimensional and porous-framework membranes promise sharp size cut-offs and tunable surface chemistry, with the selectivity needed to separate very similar ions such as lithium from magnesium or to recover specific micropollutants for destruction or reuse. Hybrid systems combine NF with electrochemical or photocatalytic stages to break down rejected pollutants rather than only concentrate them, addressing the brine-disposal problem at its root. As membrane informatics, in situ characterization, and atomistic simulation mature together, NF design is shifting from incremental empirical optimization toward targeted engineering of pore size, surface charge, and transport pathways as a coherent whole.

Reverse osmosis membranes are denser, operate at higher pressures of roughly 10 to 80 bar, and reject nearly all dissolved salts, including monovalent ions such as sodium chloride. Nanofiltration membranes have slightly looser structures, operate at lower pressures of about 5 to 25 bar, and selectively reject multivalent ions and small organic molecules while letting most monovalent ions through. This makes nanofiltration cheaper to run when full desalination is not needed, for example in water softening or selective ion separation.

Ultrafiltration uses membranes with much larger pores, typically 2 to 100 nanometres, that remove suspended solids, bacteria, viruses, and large macromolecules but pass essentially all dissolved salts and small organics. Nanofiltration goes further: its sub-nanometre channels can sieve dissolved organics down to a few hundred grams per mole and reject ions on the basis of charge as well as size. Ultrafiltration is often used as a pretreatment step before nanofiltration to protect the tighter membrane from particulate fouling.

Nanofiltration removes hardness ions such as calcium and magnesium, sulfate and other multivalent anions, dissolved natural organic matter responsible for colour and taste, many pesticides and pharmaceutical residues, dyes, and per- and polyfluoroalkyl substances above a few hundred grams per mole. Rejection of monovalent salts such as sodium chloride is incomplete, typically between 20 and 70 percent depending on the membrane and conditions, which is why nanofiltration is not used as the final desalination step for seawater.

Most commercial nanofiltration membranes are thin-film composite structures with a porous polysulfone or polyethersulfone support, often on a polyester backing, topped by an ultrathin polyamide selective layer formed by interfacial polymerization. Research-grade membranes also use ceramic oxides, polyelectrolyte multilayers, graphene oxide laminates, MXene films, metal-organic and covalent organic frameworks, and biological channels embedded in synthetic supports.

By the strict definition of nanotechnology as the engineered use of structures below 100 nanometres, nanofiltration qualifies: the polyamide selective layer is roughly 20 to 200 nanometres thick, the effective transport channels are well under one nanometre across, and the rejection of solutes depends on chemistry and confinement at that scale. The term itself was coined in the mid-1980s, before nanotechnology became a household word, and was originally chosen to describe the size of solutes being separated rather than to invoke nanotechnology.

Desalination, Nanofiltration membranes broaden the use of membrane separation technology

Science, Precise and ultrafast molecular sieving through graphene oxide membranes

Chemical Society Reviews, Tailored design of nanofiltration membranes for water treatment based on synthesis-property-performance relationships

Nature Water, Performance metrics for nanofiltration-based selective separation for resource extraction and recovery

Chemical Society Reviews, Empowering ultrathin polyamide membranes at the water-energy nexus: strategies, limitations, and future perspectives

Nature Reviews Methods Primers, Pressure-driven membrane desalination