Nanofiltration Membrane Applications
1. What Nanofiltration Is — and Isn’t
Nanofiltration (NF) occupies the molecular weight cutoff (MWCO) range between ultrafiltration and reverse osmosis, typically 150–1,000 Da, with a separation layer thought to carry roughly 1 nm pore structure — hence the name. In practice, a membrane is generally classified as NF rather than RO once its NaCl rejection falls to 90% or below; NF is accordingly sometimes described as “loose RO” or “low-pressure RO,” since it achieves useful separation at feed pressures as low as 0.1 MPa where a true RO membrane would pass far more salt. NF membranes reject organics with molecular weights from roughly 200–1,000 Da fairly completely, but their defining commercial value is selective ion separation — rejecting multivalent ions (Ca²⁺, Mg²⁺, SO₄²⁻) far more strongly than monovalent ions (Na⁺, Cl⁻) at a fraction of RO’s operating pressure and energy cost.
2. Separation Mechanism: Solution-Diffusion Plus the Donnan Effect
NF transport follows the same solution-diffusion framework as RO — permeate dissolves into the membrane polymer and diffuses down its own concentration gradient, so any solute crossing the membrane must overcome an effective osmotic pressure differential. What distinguishes NF is the added contribution of electrostatic (Donnan) exclusion: the membrane’s fixed surface charge interacts with the electrolyte ions in solution, and because ions of different valence carry different charge density, the membrane’s rejection of any given ion depends heavily on what else is in the feed water, not just on that ion’s own size.
This produces a counterintuitive field result worth flagging explicitly, because it runs against the common assumption that a membrane’s published rejection curve is a fixed property of the membrane alone. In single-salt bench testing, a mid-to-high-rejection NF element can reject MgSO₄ at 98% or better. Introduce a monovalent salt like NaCl into the same feed, and that same element’s magnesium rejection can drop to roughly 77% — a divalent-ion rejection loss of over 20 percentage points caused entirely by the presence of a second, unrelated ion. The mechanism: the more weakly rejected monovalent anion (Cl⁻) passes through the membrane first, and to preserve local electroneutrality on the permeate side, it effectively pulls a divalent cation (Mg²⁺) through with it. Any NF process design based on single-salt rejection data alone will therefore over-predict divalent removal in a real multi-ion feed water — mixed-ion pilot or bench data is not optional for a duty where hardness or heavy-metal removal is the design target.
3. Rejection-Class Specifications
Commercial NF elements are typically grouped into three nominal NaCl-rejection classes, independent of any particular manufacturer’s model naming:
| NF Class | Typical NaCl Rejection | Approx. Organic MWCO | Typical 8040-Format Flow (per element) |
|---|---|---|---|
| NF40-class (loose) | 20–40% | <400 Da | ~26.5 m³/d |
| NF70-class (mid) | 40–70% | <200 Da | ~26.5 m³/d |
| NF90-class (tight) | 85–95% | <100 Da | ~28.4 m³/d |
Divalent-ion (Ca²⁺, Mg²⁺, SO₄²⁻) rejection runs well ahead of monovalent rejection across all three classes — typically 90%+ for CaCl₂ and >97% for MgSO₄ even in the loosest (NF40) class, versus 20–95% for NaCl depending on class. In mixed-ion field testing against an actual multi-component feed water, a tight (NF90-class) element has been shown to reject silica, magnesium, and calcium at or near 100%, sulfate and chloride above 98%, and sodium above 96%, bringing overall TDS rejection to roughly 97%. A mid-rejection (NF70-class) element run against the same feed water rejects magnesium and calcium at 70–77%, sulfate at 95%, chloride at 53%, sodium at 35%, and overall TDS at roughly 38% — illustrating how far apart the two classes sit even though both are nominally “nanofiltration.”
4. Application Landscape
NF has moved out from under RO’s shadow into its own established niche wherever a process needs monovalent/divalent ion splitting, organic concentration, or moderate desalting at lower energy cost than full RO. Documented duties include:
Landfill leachate treatment, where NF alone commonly achieves better than 90% COD rejection; petroleum-industry phenolic wastewater, where phenol removal exceeding 95% has been demonstrated; electroplating wastewater, where NF removes cadmium, nickel, mercury, and titanium at comparatively low operating pressure; textile and dyeing wastewater, where NF decolorizes and recovers dyes and auxiliary chemicals for reuse; pulp and paper wastewater, where NF removes dark-colored lignin and chlorinated lignin compounds, including kraft (sulfate) pulping effluent treated via combined UF/NF trains; and steel-mill pickling liquor, where NF concentrates and recovers nickel, iron, copper, and zinc from spent acid while allowing the residual acid stream to meet discharge limits.
5. Case Study: Municipal Drinking Water Softening
A municipal water utility plant installed 78 tight (NF90-class) elements in an 8040 format, arranged in a 7:4:2 staged array, designed for 75 m³/h (roughly 1,800 m³/d) of finished water. The system ran stably for more than three years on the following performance:
| Parameter | Feed | Product | Result |
|---|---|---|---|
| Conductivity | 1,423 µS/cm | 122 µS/cm | — |
| System desalting rate | — | — | >85% |
| System hardness removal | — | — | >95% |
The selective rejection behavior described in Sections 2–3 is exactly what makes this duty economical: a tight NF element removes the hardness-forming divalent ions (Ca²⁺, Mg²⁺) at RO-like efficiency while passing a meaningfully larger share of monovalent ions, letting the system run at NF’s lower feed pressure and energy draw rather than requiring full RO. (For converting between feed conductivity units, and between the psig/MPa/bar pressure figures common on NF data sheets, see this site’s Unit Converter.) Readers comparing this NF-based softening approach against conventional ion-exchange softening — sizing resin vessel volume and regeneration salt consumption — can use this site’s Ion Exchange Resin & CEDI Designer tool.
6. Case Study: Landfill Leachate Treatment
A 100 t/d leachate project used the following treatment train: raw leachate → equalization tank → neutralization tank → UASB → cartridge filter → nitrification tank → feed pump → cartridge filter → tubular MBR → disinfection tank → NF system → RO system → compliant discharge.
| Stage | Conductivity (µS/cm) | COD (mg/L) | BOD₅ (mg/L) |
|---|---|---|---|
| Raw leachate | 22,500 | 5,400 | 3,400 |
| MBR effluent | 22,460 | 1,400 | 440 |
| NF effluent | 20,600 | 180 | 50 |
| RO effluent | 2,720 | 28 | 5 |
The NF stage alone accounted for roughly 87% of the remaining COD reduction after MBR pretreatment (1,400 → 180 mg/L), running at a feed pressure of approximately 0.47 MPa and requiring daily chemical cleaning over more than two years of stable operation. That last detail is a genuine departure from conventional NF/RO practice: a clean-water system built around a 3-month CIP interval as the default assumption will fail against a leachate-grade feed. High COD fouls the membrane fast enough that daily cleaning has to be designed in as routine operation from day one, not treated as an exception-handling procedure.
7. Design Guidance for High-Fouling Streams
Two design changes recur across documented high-COD NF installations, both aimed at slowing fouling rather than removing it entirely:
Single-stage, high-recirculation array design. Rather than the two-stage tapered array common in clean-water NF/RO systems, high-fouling duties are more often built as a single stage with a dedicated recirculation pump positioned between the high-pressure feed pump and the membrane bank. This increases concentrate-side crossflow velocity for a given net production rate, which scours organic and colloidal foulants off the membrane surface faster than they can deposit, at the cost of running the system at a lower per-pass recovery than a clean-water design would use. A commonly cited design basis is roughly 10 m³/h of feed flow per pressure vessel, sized to guarantee adequate flush velocity even at the last element in the vessel.
Conservative design flux. Flux decline under high-COD service is fast and needs to be designed for rather than discovered in commissioning: one documented field case showed a mid-rejection (NF70-class) element losing roughly 30% of its flux within four hours of operation against a feed above 1,400 mg/L COD. The resulting design guidance for that duty was a target flux of 15–20 L/m²·h (LMH) — well below what the same element class would be run at against a clean feed. (For converting this design flux target into GFD, or checking a specific vessel’s feed-flow figure against the recirculation-rate guidance above, see this site’s Unit Converter — LMH and GFD are the two flux units that show up on essentially every NF/RO manufacturer data sheet.)
Antiscalant dosing ahead of the NF stage follows the same continuous-feed and day-tank batch-mixing math as an RO pretreatment program; see this site’s Chemical Dosing & Dilution Calculator for working through consumption and metering-pump sizing. For working NF array mass balance directly — with independent monovalent (salt) and divalent (hardness) rejection tracking across up to three-stage arrays, plus a solver for the bypass blend needed to hit a target finished hardness — see this site’s dedicated NF System Designer tool.
8. Case Study: Specialty Concentration — Herbicide Intermediate Recovery
A concentration trial on a glyphosate-family herbicide intermediate (molecular weight approximately 169 Da) targeted raising a 0.1% feed solution (pH 6–8) to a 2.5%+ concentrate, using a storage tank → booster pump → filter → high-pressure pump → NF element → concentrate tank process train.
| NF Element | Feed Concentration | Run Time (h, intermittent) | Operating Pressure (MPa) | Final Concentrate |
|---|---|---|---|---|
| Loose NF (NF70-class, 2540 format) | 0.1% | 60 | 2.5–3.0 | 2.6% |
| Loose NF (NF50-class, 2540 format) | 0.1% | 60 | 2.0–2.5 | 3.0% |
Both elements exceeded the 2.5% target, reaching a concentration factor of roughly 260× on the better-performing run — a result that depends on the target organic (169 Da) sitting comfortably above the element’s MWCO while chloride and other small dissociated ions pass through largely unhindered, keeping the required operating pressure modest relative to what a similar concentration factor would demand from RO.
9. Case Study: High-Salinity Fermentation Broth Desalting
A high-salt oligosaccharide fermentation broth — feed COD around 23,000 mg/L, conductivity around 40,000 µS/cm, with target organics in the 100–200 Da range — was processed in recirculating (batch-concentration) mode to strip salt while retaining as much of the organic fraction as possible, concentrating the original volume by 3× or more down to a final 80 L batch, via storage tank → dual cartridge filters → high-pressure pump → NF element → separated-liquor tank.
| NF Element | Batch Volume | Run Time | Avg. Permeate Rate | Operating Pressure (MPa) | Organic Retention |
|---|---|---|---|---|---|
| Mid NF (NF70-class, 4040 format) | 260 L | 2 h | 80 L/h | 0.6 | >80% |
| Loose NF (NF40-class, 4040 format) | 260 L | 1.5 h | 180 L/h | 0.55 | >75% |
The tradeoff between the two runs is direct: the looser element processes the batch in three-quarters of the time at a lower operating pressure, at the cost of roughly 5+ percentage points of organic retention — a real design choice between throughput/energy cost and product yield rather than a case of one element being simply “better.” (For converting the batch flow figures above into other flow units, or the feed COD/conductivity figures into different concentration bases, this site’s Unit Converter covers the common mg/L, ppm, and g/L conversions.)
10. Practical Implications for Design
Across all four cases above, the pattern is consistent: NF’s value comes from tuning rejection class and operating pressure to the target separation — divalent hardness removal at RO-like efficiency and RO-like energy savings for drinking water, moderate organic/COD rejection at sharply lower fouling-driven CIP intervals for high-strength leachate, and pressure/throughput tradeoffs against small-molecule organic retention for specialty concentration and desalting duties. The Donnan-effect interaction between co-existing ions means single-salt manufacturer data should be treated as a starting point, not a design guarantee, whenever the real feed water carries a genuinely mixed ionic background — which, outside of a bench test, is essentially always.