Why Coal Chemical Wastewater Needs a Nanofiltration Step
China's coal-to-ethylene glycol capacity reached 30.142 million tons per year in 2024, accounting for 49.24% of global production, and the coal-to-glycol route alone delivers 35% of China's domestic ethylene glycol output (per S4, ScienceDirect 2025). Every ton of glycol produced generates a parallel stream of high-salinity, refractory wastewater that the standard biological + RO train cannot fully resolve. The RO step recovers clean permeate for reuse but leaves behind a concentrate loaded with mixed NaCl, Na₂SO₄, and dissolved organic matter (DOM) that resists further membrane separation.
That concentrate is the operational pain point. The conventional end-of-pipe solution is thermal evaporation followed by crystallization, but the resulting mixed-salt cake is classified as solid waste and sent to landfill — a line item that can exceed 30% of the plant's total wastewater OPEX, with zero byproduct credit. Landfill capacity for inorganic salt waste in Inner Mongolia and South Asia is also tightening under new solid-waste rules, raising the disposal liability.
A properly designed nanofiltration step breaks the problem in two. By splitting the monovalent NaCl from the polyvalent Na₂SO₄ upstream of crystallization, each salt exits the evaporator as a single-component crystal that meets industrial-grade sale specifications rather than a mixed waste. The economic case is no longer "treat brine and pay for landfill" — it becomes "recover two salable products and avoid most of the evaporation duty on the NF permeate stream."
How Nanofiltration Separates Salts in Coal Chemical Brine
Nanofiltration sits between ultrafiltration and reverse osmosis in both pore size and selectivity. The defining mechanical range is 0.5–2 nm equivalent pore size, which corresponds to a molecular weight cut-off (MWCO) of 200–400 Da for the polyamide thin-film composite elements used in industrial coal chemical service (per S4, 2025). The two commercial reference elements in the literature are MNF-200 at 200 Da and MNF-400 at 400 Da, both negatively surface-charged.
Two coupled mechanisms govern salt rejection on these membranes. Steric pore-size sieving rejects species whose hydrated radius is larger than the membrane's effective pore opening — which already excludes most polyvalent salts and most DOM. The second mechanism, the Donnan effect, comes from the negative fixed charge on the polyamide surface: co-ions (anions) are electrostatically repelled, while counter-ions (cations) are partially held back to maintain electroneutrality on the permeate side. A divalent anion like SO₄²⁻ carries twice the charge of Cl⁻ and is rejected much more strongly as a result.
The economic basis of an NF salt-splitting train rests on a counterintuitive data point from S4: in mixed NaCl/Na₂SO₄ systems, monovalent salt rejection by NF can drop below 0% because the polyvalent anion screens the membrane's surface charge and the chloride ion is actually pulled through with the water. That negative rejection is the entire reason the salt split is so clean — Na₂SO₄ is concentrated on one side of the membrane and NaCl is enriched on the other. RO cannot do this; it rejects both salts above 99% indiscriminately, which leaves the crystallizer with the same mixed-salt problem the evaporator was already failing on.
Membrane Selection: 200 Da vs 400 Da for Coal Chemical Streams
The benchmark full-scale data point comes from a coal-to-ethylene glycol plant in Anhui Province, China, where an MNF-200 NF train processed approximately 4,000 m³/d of RO concentrate continuously for three months, sustaining greater than 98% Na₂SO₄ rejection and less than 10% NaCl rejection throughout the run (per S4, ScienceDirect 2025). That is the performance a buyer should anchor a specification to.
MNF-400 offers higher pure-water flux at the same pressure because its larger pores offer less hydraulic resistance, but the monovalent/polyvalent selectivity gap narrows. For a coal-to-glycol brine where the downstream crystallizer needs a sharp split to produce a single-component Na₂SO₄ crystal, the conservative choice is MNF-200. The selection logic reduces to three criteria that the Anhui team used to screen five candidates down to two finalists: surface charge (negative is required for Donnan exclusion of SO₄²⁻), MWCO (200 vs 400 Da), and structural stability under high salinity (per S4, 2025). Surface charge and structural stability win before flux does — high flux on a membrane that swells or loses charge after 30 days is not a real performance number.
| Parameter | MNF-200 (200 Da) | MNF-400 (400 Da) |
|---|---|---|
| Na₂SO₄ rejection (mixed brine, full-scale) | >98% | ~90–95% (typical polyamide range) |
| NaCl rejection (mixed brine, full-scale) | <10% | ~20–35% (typical polyamide range) |
| Pure-water flux (relative) | Baseline | ~1.5–2× baseline |
| Operating pressure envelope | 10–30 bar | 10–25 bar |
| Best-fit application | Sharp NaCl/Na₂SO₄ split before crystallization | Higher-flux polishing, lower-purity salt split |
| Documented full-scale run length | 3 months continuous (Anhui case, S4) | Not documented at full scale for this service |
Decision rule: if the crystallizer feed purity target for either salt stream is above 95% on a single-pass basis, specify 200 Da. If the plant can tolerate a recycle loop on the Na₂SO₄ side to push purity up, 400 Da becomes defensible and earns back flux.
Fouling and DOM Behavior in the NF Concentrate
Plug-and-play NF on raw coal chemical RO concentrate will collapse within weeks because the dissolved organic matter is not generic — it is chemically specific. Molecular-level characterization of a full-scale coal chemical ZLD train (per S2, Water Research 2026) showed that the NF concentrate accumulates high-molecular-weight, sulfur-rich DOM together with tryptophan-like biogenic metabolites that survive biological treatment. By contrast, the ultra-high-pressure RO concentrate downstream of NF accumulates low-molecular-weight halogenated N-heterocycles. The two concentrates foul differently and require different mitigations.
The sulfur-rich, high-MW fraction in the NF concentrate drives organic fouling and creates a secondary risk: any free chlorine dosed upstream will attack the sulfur centers and release aggressive oxidants that damage the polyamide layer. Tryptophan-like fluorescence signals incomplete biological treatment — if the upstream MBR is underperforming, BIX and UV254 will rise before flux does, giving the operator a 24–72 hour warning. Both UV254 and BIX are usable as online surrogates for fouling propensity on a real plant (per S2, 2026), and either is cheap to instrument compared to a membrane replacement.
The operating consequence is that the pre-treatment chain ahead of NF is not optional, the CIP recipe is not generic, and the upstream biology must be in specification before NF ever sees the water. Treat these as a single coupled design problem rather than three independent unit operations.
Pre-Treatment and System Integration Around the NF Skid
NF does not stand alone. A defensible train on coal chemical RO concentrate requires, in order: a DAF unit ahead of UF and NF for suspended solids, oil, and grease; a PVDF ultrafiltration system as NF guard at 0.03 µm nominal pore size to strip colloids and biomass; a PLC-controlled anti-scalant and pH dosing skid to keep calcium sulfate and silica below their saturation indices in the NF concentrate; and the NF rack itself. Skipping the DAF step collapses NF flux within 4–6 weeks on real coal chemical feed; skipping UF risks colloidal fouling that no CIP cycle fully reverses.
NF is positioned on RO concentrate, not on raw wastewater. The full hydraulic train is biological treatment (A/O or MBR) → industrial RO system upstream of NF → NF → evaporative crystallization, with the NF permeate (NaCl-rich) and NF concentrate (Na₂SO₄-rich) routed to two separate crystallizers. Electrodialysis is uncommon on coal chemical brine because the divalent content fouls ion-exchange membranes, but it can polish the recovered Na₂SO₄ or NaCl streams for sale into higher-purity markets if the revenue justifies the added stage. UF flowrates for a 4,000 m³/d NF feed typically fall in the 2,000–40,000 L/h envelope depending on recovery and feed quality; sizing is set by the upstream MBR effluent SDI and the target silt density index into the NF rack.
For broader ZLD design context, the same hybrid DAF-RO-MBR ZLD train engineering specs used in heavy-metal service carry over to coal chemical applications with different chemical dosing, and the ZLD sizing methodology for mineral concentrators provides a transferable mass-balance framework for sizing the NF stage against the crystallizer heat duty.
Operating Parameters, CIP and Membrane Life Expectation
Polyamide NF on coal chemical RO concentrate operates comfortably inside a feed pressure envelope of 10–30 bar, a system recovery of 50–75%, a pH window of 2–11 during normal operation (2–12 during CIP), and a continuous-feed temperature below 45 °C. These are design ranges consistent with standard polyamide chemistry, not vendor-specific — a supplier claiming 40 bar continuous feed on a 200 Da element is overspec'ing a standard part.
The CIP schedule follows the DOM fingerprint. Trigger a cleaning cycle at 10–15% normalized flux decline at fixed temperature and pressure, or on a scheduled basis (weekly alkaline + monthly acidic) — whichever comes first. The sulfur-rich, protein-like DOM documented in the NF concentrate (per S2, Water Research 2026) responds best to alkaline surfactant CIP at pH 10–11 with a non-ionic surfactant, followed by an acidic stage at pH 2–3 to dissolve any mineral scale that codeposited with the organic fouling layer. Free chlorine above approximately 0.1 ppm on the feed side damages the polyamide cross-link; use non-oxidizing biocides (DBNPA or isothiazolone) upstream of the NF rack for biological control instead.
Membrane life expectation on well-pretreated coal chemical RO concentrate is 3–5 years, with the Anhui MNF-200 train demonstrating the three-month continuous operating floor that any pilot should be required to match before procurement signs off. A pilot run shorter than 90 days is not a basis for a 5-year membrane life warranty.
Process Flow: From Coal Chemical Effluent to Recovered Salts
The complete train runs in four steps. Step 1 is biological treatment — typically anoxic/oxic (A/O) followed by MBR — which strips biodegradable COD and ammonia but, per the S2 characterization (2026), leaves tryptophan-like biogenic metabolites and a fraction of saturated/reduced DOM that the bacteria cannot mineralize. Step 2 is RO, which produces a clean permeate for reuse and a concentrate containing the bulk of the dissolved salts plus the refractory DOM. Step 3 is the NF salt split using an MNF-200 class element: the NF permeate, enriched in NaCl because monovalent chloride passes the membrane, goes to a NaCl crystallizer; the NF concentrate, enriched in Na₂SO₄ because the divalent sulfate is rejected, goes to a separate Na₂SO₄ crystallizer. Step 4 is evaporative crystallization of each separated stream, yielding two industrial-grade salt products; the crystallizer condensate returns to the RO feed tank to close the water loop.
The economic shift is the point. The plant replaces one mixed-salt landfill waste stream with two saleable byproduct streams and reduces the thermal evaporation duty because the NF permeate (NaCl-rich water) is far less refractory than the original mixed brine. The result is near-zero liquid discharge in the regulatory sense plus a partial revenue offset against CAPEX — defensible numbers a procurement engineer can put in front of management, anchored to the Anhui full-scale performance rather than a vendor brochure.
Frequently Asked Questions
Why use NF instead of RO alone on coal chemical wastewater?
RO rejects both NaCl and Na₂SO₄ above 99% and produces a mixed-salt concentrate that still has to be evaporated into a mixed-salt cake classified as solid waste. NF exploits the monovalent/polyvalent selectivity gap — greater than 98% Na₂SO₄ rejection with less than 10% NaCl rejection (per S4, 2025) — so each salt exits the crystallizer as a single-component product that can be sold rather than landfilled.
What MWCO is correct for coal-to-glycol brine?
200 Da (MNF-200 class) is the documented full-scale choice for coal-to-ethylene glycol RO concentrate, demonstrated at approximately 4,000 m³/d over a three-month continuous run in Anhui Province, China (per S4, ScienceDirect 2025). 400 Da is acceptable only if the plant can tolerate lower selectivity and recycle the Na₂SO₄ stream to push purity up.
What pre-treatment does an NF skid need on coal chemical streams?
A DAF or lamella clarifier for oil and suspended solids, a 0.03 µm PVDF ultrafiltration guard, and chemical dosing for pH adjustment and anti-scalant. Without all three, NF flux collapses within weeks on real coal chemical feed.
How long do NF membranes last on this service?
Three to five years is realistic for polyamide NF on well-pretreated coal chemical RO concentrate. A pilot demonstration of at least 90 days continuous stable operation should be a minimum procurement condition; the Anhui case provides that reference point (per S4, 2025).
Can NF achieve true zero liquid discharge on its own?
No. NF performs the salt split that makes ZLD economic; evaporative or mechanical crystallization of the two separated brine streams is still required to reach near-zero liquid discharge. NF is the enabling step, not a replacement for the thermal stage.
