Reading Nanofiltration Data Before You Size the Skid: Rejection, Pressure, Charge

A nanofiltration membrane removes material at the nanometre scale — the conventional shorthand is 0.001 micron — and its working range sits between ultrafiltration and reverse osmosis. That position in the middle is exactly why the datasheet is easy to misread. Rejection of dissolved salts runs somewhere between 20% and 98% depending on the salt, and removal of soluble monovalent ions is lower than removal of multivalent ions. A single rejection figure quoted without the test ion tells you almost nothing.

What the membrane is actually asked to do

The everyday duties are narrower than the marketing suggests. Nanofiltration removes organics and colour from surface water, hardness and radium from groundwater, and partially removes dissolved salts in the process. It also serves extraction and concentration of valuable substances in food and pharmaceutical production, where the absence of a phase change is the attraction. Operating pressure generally falls in the 3.5-30 bar band, which is low enough to keep energy cost modest and high enough that pretreatment cannot be skipped.

Membrane materials have to satisfy a long list at once: good film-forming behaviour, thermal and chemical stability, high mechanical strength, resistance to acid, alkali and microbial attack, resistance to chlorine and other oxidants, high water flux, high salt rejection, resistance to colloidal and suspended-solid fouling — and a low price. Most membranes in service are aromatic or polyamide composites. A composite is an asymmetric structure with two jobs split between two layers: a porous support that works by sieving and a thinner dense layer that performs the actual separation, describable through solution-diffusion theory. The advantage of the arrangement is that skin and support can each be optimised in material and structure independently.

Module choice is a maintenance decision

Four module types are on the market: hollow fibre, spiral wound, plate-and-frame and tubular. Hollow fibre and spiral wound give high packing density, low cost and good hydrodynamics inside the element, but manufacturing is technically demanding, sealing is difficult, fouling resistance in service is poor and feed pretreatment requirements are high. Plate-and-frame and tubular modules are easy to clean and tolerant of fouling, but packing density is low and cost is high. Most nanofiltration systems therefore settle on hollow fibre or spiral wound, and the pretreatment specification becomes the de facto warranty on membrane life.

Development history explains the supply pattern. Theoretical work in China started relatively early, but membrane development remains at a preliminary stage. In the United States and Japan the technology reached commercial maturity sooner, and the widely used commercial composite membranes include Filmtec's NF series in the US, Nitto Denko's NTR-7400 series and Toray's UTC series in Japan.

Why salt rejection is an incomplete indicator

For an ordinary reverse osmosis membrane, salt rejection is the key performance indicator. For nanofiltration it is not sufficient, and in some cases it is actively misleading: a nanofiltration membrane can retain a larger-molecular-weight substance at a lower rate than a smaller one. The filtration mechanism is complex, and because nanofiltration is a young branch of reverse osmosis, most published explanations borrow reverse osmosis transport models — the solution-diffusion theory of Lonsdale and Merten, the hydrogen-bond theory of Reid and Breton, Sherwood's diffusion-pore-flow theory, the preferential adsorption-pore-flow theory of Loeb and Sourirajan, and Glueckauf's pore theory.

Two further variables push actual performance away from the datasheet. Membrane charge and manufacturing process give different products different selectivity: ordinary nanofiltration membranes retain divalent ions at a higher rate than monovalent ions, and in multi-component mixtures the rejection of monovalent ions may even fall. Operating conditions then move the result again — permeate flux rises with pressure, while rejection falls as solution concentration rises. A skid sized on datasheet rejection at a dilute bench concentration will underperform on a concentrated process stream.

The concentration duty, stated plainly

Where nanofiltration is used for concentration and purification, the case rests on process economics rather than on rejection percentage:

• Concentration and purification run at room temperature with no phase change and no chemical reaction, so no impurities are introduced and the product does not decompose or denature — which is what makes the process usable for heat-sensitive substances.
• Salts are removed from the product, ash content falls and purity rises; against solvent desalting the quality is better and yield can improve.
• Recovery is high and losses are low.
• Acids, alkalis, alcohols and other valuable substances in solution can be recovered for reuse.
• Equipment is simple and compact, with a small footprint and low energy consumption.
• Operation is automatic, stable and easy to maintain.

Drinking water: which numbers to trust

In drinking water treatment, nanofiltration beyond softening is used mainly for decolourisation and for removing natural and synthetic organics such as pesticides, trihalomethane precursors, disinfection by-products and their precursors, and volatile organics, with the aim of biological stability in the distribution network. The mutagenicity evidence is the most concrete part of the claim: studies show removal of most toxic and harmful organics and Ames mutagens, with the mutagenicity ratio (MR) of TA98 and TA100 strains below 2 at all tested doses and Ames test results negative.

On disinfection by-products — trihalomethanes (THMs), haloacetic acids (HAAs) and possible trichloroacetaldehyde hydrate (CH) — the average rejection rates reported for their precursors are 97%, 94% and 86% respectively. Those three figures are the ones worth carrying into a design basis, because they are precursor rejections rather than finished by-product rejections.

There is a secondary benefit that rarely appears in tenders but matters to utilities: nanofiltration effluent is low in corrosivity, which affects the service life of pipe networks and the dissolution of metal ions from pipe surfaces. Tests indicate that a nanofiltration system with appropriate post-treatment can cut lead dissolution in distribution networks by 50%, with other dissolved metal concentrations meeting drinking water standards.

Softening, and the limits of the technology

Membrane softening exploits selective permeability to ions of different valence. While removing hardness it also removes turbidity, colour and organics, so effluent quality is clearly better than other softening routes; it needs no regeneration, generates no regeneration brine, is simple to operate and saves space. In the US it is already common — over the past decade or so, new softening plants in Florida have chosen membrane softening over conventional lime softening and ion exchange. As performance improved and module prices fell, membrane softening became superior to, or comparable with, conventional methods on investment and on operation and maintenance.

The boundaries are worth stating because they are frequently crossed in proposals. Nanofiltration retains divalent and higher ions and other particles while passing water molecules and some monovalent ions such as sodium, potassium and chloride; that is why it can produce direct drinking water that still retains useful ions at lower cost than reverse osmosis. It is not a desalination technology, and it is not a disinfection barrier. In rural China and small-town plants, where management is often lax enough that bacteria can appear in the effluent, advanced treatment and disinfection are still required after the membrane.