What Wafer Cleaning Wastewater Actually Contains
Fluoride is the design-limiting species in spent wafer-cleaning effluent, with composite concentrations typically ranging from 50 to 2,000 mg/L depending on whether DHF, Piranha, or a blended RCA stream dominates the day's flow. The other parameter that drives every downstream decision is pH swing: Piranha exits at pH <1, DHF at pH 2–4, SC1 at pH 10–11, and SC2 at pH 1–3, so a single combined drain line can move 12 pH units between shift changes. Treat the composite stream as a high-fluoride, low-to-moderate COD, low-TSS, heavy-metal-bearing industrial wastewater — not as a municipal sewage analog.
Four chemistries generate roughly 90% of fab cleaning wastewater. Piranha (H2SO4:H2O2 at 3:1–4:1, 120–150°C) strips photoresist and organics, contributing the bulk of COD and sulfate load. SC1 (NH4OH:H2O2:H2O ≈ 1:1:5 at 65–80°C) removes particles and organics, contributing the bulk of NH3-N. SC2 (HCl:H2O2:H2O ≈ 1:1:6 at 70–80°C) dissolves residual metals, contributing the bulk of chloride and trace Cu/Fe/Ni. DHF (1–2% HF, 20–25°C) strips native oxide and is the dominant fluoride source. Modutek's ozone/DI rinse is a low-chemical alternative but still produces a dilute fluoride-bearing rinseate that must be treated before discharge.
Per-fab flow ranges from 200 m³/day at a pilot line to 2,000 m³/day at a 300 mm advanced-node fab, with peak shift-up flows running 1.2–1.5× the daily average. Equalization for 8–24 h HRT is therefore non-negotiable; without it, the morning Piranha batch will slug the fluoride removal reactor with a stoichiometric overdose demand and the afternoon SC1 batch will drive NH3-N above the biological polishing window. The composite numbers below are what an equalized blend typically looks like at the head of the treatment train.
| Parameter | Unit | Range after equalization | Design basis |
|---|---|---|---|
| F⁻ (fluoride) | mg/L | 50–2,000 | drives Ca dose in Stage 2 |
| COD | mg/L | 30–300 | controls Stage 5 sizing |
| TSS | mg/L | 10–100 | settles in Stage 4 |
| NH3-N | mg/L | 5–80 | toxic to MBR above ~50 mg/L |
| Cu / Fe / Ni (each) | mg/L | 0.5–20 | polished in Stage 3 |
| pH | — | 1–13 swing; 6–9 after EQ | online probe required |
| Total flow | m³/day | 200–2,000 | per fab |
The 2026 Treatment Train: Stage by Stage
A defensible 2026 treatment train runs six unit processes in series: equalization, fluoride precipitation, heavy-metal precipitation, clarification, organic polishing, and RO reuse. Skipping or reordering stages is the most common cause of permit excursions — fluoride must be removed before biological polishing because free F⁻ above 10–20 mg/L inhibits nitrifiers and most heterotrophs.
Stage 1 — Equalization. A lined concrete or FRP basin sized for 8–24 h HRT with mechanical mixers (G ≈ 50–100 s⁻¹ for blending, lower for settling avoidance) and online pH plus F⁻ probes tied to a PLC-controlled chemical dosing system for NaOH/H2SO4 trim. Volume is sized for 1.2–1.5× peak shift flow, not daily average.
Stage 2 — Fluoride removal. Calcium precipitation with CaCl2 at 2–5× stoichiometric dose (roughly 4.4–11 kg CaCl2 per kg of F⁻ removed as CaF2), pH 7–9, HRT 30–60 min, drives F⁻ from 50–2,000 mg/L down to 5–15 mg/L. For residuals under 2 mg/L, augment with alumina or calcium phosphate coagulation, or polish on activated alumina in a separate contactor. The F⁻ residual is governed by Ca²⁺ solubility, so under-stoichiometric dosing is the most common operator error.
Stage 3 — Heavy-metal precipitation. Raise pH to 9–10 with NaOH or Ca(OH)2 to crash Cu/Ni/Zn as hydroxides, then dose Na2S or a dithiocarbamate (TMT-15 class) for tighter Cu polishing to <0.5 mg/L. Staged chemistry (hydroxide first, sulfide second) is cheaper than a single sulfide dose and produces a more stable sludge.
Stage 4 — Clarification. A lamella clarifier runs at 20–40 m²/h surface loading for hydroxide floc, while a dissolved air flotation (DAF) clarifier at 4–25 m²/h is preferred when light sulfide or biological floc dominates. Sludge is routed to a plate-and-frame filter press for dewatering to 15–30% dry solids.
Stage 5 — Organic polishing. Either a submerged MBR system (0.1–0.4 μm PVDF, HRT 6–12 h) for COD 80–300 mg/L down to <50 mg/L, or a Fenton/AOP step (H2O2/Fe²⁺ at 0.3–1.0× stoichiometric) when the stream is non-biodegradable. For guidance on Fenton reactor sizing, see the Fenton oxidation design guide and the AOP process guide.
Stage 6 — Polishing and reuse. A multi-media filter brings SDI below 3, then a two-pass industrial RO system at 70–95% recovery feeds the UPW make-up loop. RO concentrate is routed to a brine evaporator or crystallizer for ZLD on the fluoride stream.
| Stage | Unit process | Key reagents / media | HRT | Influent → effluent | Removal |
|---|---|---|---|---|---|
| 1 | Equalization + pH trim | NaOH / H2SO4 | 8–24 h | pH 1–13 → 6–9 | — |
| 2 | Ca precipitation (+ alumina polish) | CaCl2, NaOH | 30–60 min | F⁻ 50–2,000 → 5–15 mg/L | 97–99% |
| 3 | Hydroxide + sulfide precipitation | NaOH, Na2S / TMT | 20–40 min | Cu/Ni 0.5–20 → <0.5 mg/L | >95% |
| 4 | Lamella or DAF clarification | Polymer 1–5 mg/L | 1–2 h | TSS 100 → <10 mg/L | 90%+ |
| 5 | MBR or Fenton/AOP | H2O2/Fe²⁺ or PVDF membranes | 6–12 h | COD 80–300 → <50 mg/L | 70–90% |
| 6 | MMF + two-pass RO | SDI <3 feed; CIP chemicals | — | TDS 500–2,000 → <20 mg/L | 70–95% recovery |
Selecting the Right Unit Processes for Each Fab Size
Right-sizing is a question of which unit processes earn their OPEX at a given flow envelope, not which stages are theoretically required. A 300 m³/day R&D line cannot justify an MBR and a two-pass RO; a 3,000 m³/day advanced-node fab cannot rely on cartridge filtration before reuse.
For a small fab or pilot line below 500 m³/day, the cost-effective package is equalization + Ca precipitation + DAF + cartridge filtration, going direct to RO without biology. For mid-scale operations between 500 and 2,000 m³/day, add a lamella clarifier, an MBR, and single-pass RO with CIP, reusing permeate to non-rinse service such as cooling-tower make-up or scrubber feed. For large or advanced-node fabs above 2,000 m³/day, the full train — MBR, AOP, two-pass RO, and evaporator brine system — is required to push reuse above 85% and achieve ZLD on the fluoride concentrate. Decision rule: select RO over ion exchange when feed TDS stays below 2,000 mg/L and reuse target exceeds 70%; reserve ion exchange for final trace-metal polishing before the UPW polisher. The economics and membrane-fouling constraints at this scale are detailed in the microelectronics ZLD plant benchmark.
| Fab class | Flow (m³/day) | Treatment package | Reuse target | CAPEX band (USD) |
|---|---|---|---|---|
| Small / R&D | <500 | EQ + Ca ppt + DAF + cartridge + single RO | 50–70% | 0.4–1.2 M |
| Mid-scale | 500–2,000 | EQ + Ca ppt + lamella + MBR + single RO + CIP | 70–85% | 1.5–4.5 M |
| Large / advanced node | >2,000 | Full train + AOP + two-pass RO + evaporator/ZLD | >85%, ZLD on F⁻ conc. | 6–18 M |
2026 Compliance Limits You Must Hit
Three regulatory documents govern fab wastewater design in the China and Taiwan markets in 2026, and the design must satisfy the strictest applicable limit at every parameter. Treat 50–70% of the numerical limit as the design target so that shock loads and analytical variance do not cause excursions.
China GB 39731-2020 (Electronic industry water pollutants) sets existing-source F⁻ at 20 mg/L, new-source F⁻ at 10 mg/L, COD at 100 mg/L, SS at 30 mg/L, total Cu at 0.5 mg/L, and pH 6–9. The Taiwan EPA semiconductor effluent standard (2024 revision, in force through 2026) sets F⁻ at 15 mg/L, COD at 100 mg/L, SS at 30 mg/L, Cu at 1.0 mg/L, Ni at 0.5 mg/L, and NH3-N at 30 mg/L. For the reuse side, reference SEMI E12/E61 fab water specifications target resistivity above 15 MΩ·cm, TOC below 10 ppb, and particle counts under 1/L at 50 nm. The train is the same regardless of which side of the strait the fab sits on — only the effluent targets change.
| Parameter | China GB 39731 (new source) | Taiwan EPA (2024 rev.) | SEMI E12/E61 reuse | Design target |
|---|---|---|---|---|
| F⁻ | 10 mg/L | 15 mg/L | — | 5–7 mg/L |
| COD | 100 mg/L | 100 mg/L | TOC <10 ppb | <50 mg/L |
| SS | 30 mg/L | 30 mg/L | <1/L at 50 nm | <10 mg/L |
| Cu | 0.5 mg/L | 1.0 mg/L | — | <0.3 mg/L |
| Ni | — | 0.5 mg/L | — | <0.2 mg/L |
| NH3-N | — | 30 mg/L | — | <15 mg/L |
| pH | 6–9 | 6–9 | — | 6.5–8.5 |
Costs, Sludge, and Common Engineering Mistakes
CAPEX for a packaged 2026 fab wastewater system runs USD 0.4–1.2 M for a small line, USD 1.5–4.5 M for a mid-scale train with MBR and RO, and USD 6–18 M for a large ZLD-enabled system; mega-fab ZLD plants reach USD 50 M at the upper bound (see the electronics wastewater treatment system guide). OPEX is dominated by chemical dosing — CaCl2, NaOH, polymer — and RO membrane replacement on a 3–5 year cycle; Fenton reagent alone adds USD 0.20–0.60 per m³, and total OPEX typically lands between USD 0.55 and 2.40 per m³ treated.
Sludge is mostly CaF2 from Stage 2 plus metal hydroxides and sulfides from Stage 3. CaF2 yield is stoichiometric at about 2.2 kg of dry cake per kg of fluoride removed, and the plate-and-frame press should be sized for 8–15% cake solids — below that, hauling costs dominate. The four recurring engineering mistakes on these projects are: omitting fluoride removal before the MBR, which kills the biomass; co-mingling HF and alkaline streams without equalization, which produces localized CaF2 scaling in transfer piping; under-sizing equalization for the morning shift-up Piranha batch; and using carbon-steel tanks in fluoride service, which fail within months. Detailed mass-balance and piping lessons appear in the wet etching wastewater treatment guide and the regional compliance context in the electroplating wastewater compliance guide.
Frequently Asked Questions
What fluoride limit must a 2026 fab hit in China? Under GB 39731-2020, new fabs must meet 10 mg/L F⁻, while existing sources face 20 mg/L; design to 5–7 mg/L for margin.
What is the copper effluent limit for a Taiwan semiconductor fab in 2026? The Taiwan EPA sets Cu at 1.0 mg/L and Ni at 0.5 mg/L; design to 0.3 mg/L Cu and 0.2 mg/L Ni to absorb shock loads.
What RO recovery is realistic for fab UPW make-up reuse? Two-pass RO on equalized, Ca-precipitated feed runs 70–95% recovery; mid-scale fabs target 70–85%, large fabs exceed 85% before the evaporator.
How much does a ZLD-capable fab wastewater plant cost in 2026? A large fab ZLD train runs USD 6–18 M, with mega-fab installations reaching the USD 5–50 M band documented for full microelectronics ZLD plants.
Can MBR handle fluoride-bearing fab wastewater without pre-treatment? No. Free F⁻ above 10–20 mg/L inhibits nitrifiers and heterotrophs; calcium fluoride precipitation must precede the MBR.
What chemical dose is needed to remove fluoride by Ca precipitation? CaCl2 at 2–5× stoichiometric (about 4.4–11 kg CaCl2 per kg F⁻ removed) at pH 7–9, with alumina or calcium phosphate coagulation for residuals under 2 mg/L.
