Why a Pulp Mill Expansion Is a Different ZLD Question Than Petrochem or Pharma
UPM's 2026 pulp mill expansion does not, on its own, trigger full zero liquid discharge — it triggers minimum liquid discharge (MLD) with high-recovery reverse osmosis. Because kraft and sulfite effluents carry high BOD, color, lignin, silica, and scaling ions, conventional RO at 50–80% recovery cannot close the loop. A hybrid MLD train using fluidized-bed crystallization plus cyclic RO can reach >90% recovery and route only 5–10% of feed to a polishing thermal stage, which is the technically and economically defensible path for a pulp expansion in 2026.
The "ZLD or RO" framing is a false binary for pulp and paper. ZLD, as defined in the 2026 wcponline process-engineering review, means no liquid waste leaves the site — all water is recovered and only solid residues remain. MLD sits between conventional treatment and full evaporation: by combining mechanical, membrane, and selective thermal steps, MLD shrinks the residual brine stream by 60–90% before any thermal polish is needed (per wcponline 2026-01). For most kraft-mill expansions, that 60–90% brine reduction is enough to meet effluent limits and water-reuse targets without paying the CAPEX and energy penalty of a full evaporator train.
Pulp-mill effluent is uniquely hostile to high-recovery RO for three reasons. First, the organic load (COD 800–2,500 mg/L, BOD 250–600 mg/L) fouls membranes quickly and starves biological stages if not handled upstream. Second, color bodies and lignin derivatives create irreversible adsorption fouling that no cleaning cycle fully reverses. Third, and most decisive for recovery ceiling, the dissolved ion profile — silica 20–80 mg/L as SiO₂, calcium 80–200 mg/L, sulfate 200–600 mg/L — caps conventional RO recovery around 70% before silica and calcium sulfate scaling become unmanageable. Petrochem and pharma effluents rarely hit this combination of organic load and silica-limited recovery simultaneously, which is why the same ZLD/RO logic does not transfer cleanly from those verticals. The same MLD vs. ZLD logic has been applied to the Pfizer API expansion and the Unilever FMCG expansion, but the influent character is materially different, and so is the recommended train.
Pulp Mill Effluent Profile: The Influent Parameters That Drive RO Recovery
You cannot size a high-recovery RO train without a defensible influent envelope, and the recovery ceiling is set by the worst-case scaling ion, not the average. The table below summarizes the post-secondary-treated kraft effluent envelope a UPM-scale expansion must be designed against; UPM site-specific values are not publicly disclosed, so these are the engineering ranges a process engineer should expect from a modern European kraft line.
| Parameter | Typical post-secondary kraft effluent | RO scaling implication |
|---|---|---|
| COD | 800–2,500 mg/L | Requires biological pre-treatment; RO feed target <25 mg/L COD |
| BOD | 250–600 mg/L | Same; also drives MBR sizing |
| TSS | 50–200 mg/L | DAF or filter polish to <5 mg/L before RO |
| Color | 500–2,000 Pt-Co | Color bodies foul RO; DAF + biological stage typically removes 60–85% |
| pH | 6.5–8.5 | Adjust to 7.0–7.5 for RO feed; silica solubility is pH-sensitive |
| Silica (SiO₂) | 20–80 mg/L | Recovery-limiting species; conventional RO caps at ~70% recovery to stay below 120–150 mg/L saturation in concentrate |
| Calcium | 80–200 mg/L | Drives CaSO₄ and CaCO₃ scaling; lime softening or FBCR purge required above 80% recovery |
| Sulfate | 200–600 mg/L | Combines with Ca to form gypsum scale at high recovery |
| Temperature | 25–40 °C | Higher flux but lower silica solubility; affects RO recovery math |
The recovery-limiting species in this envelope is silica, not hardness. At 40 mg/L SiO₂ in the feed and a 70% recovery target, the concentrate already sits near 130 mg/L — uncomfortably close to the 120–150 mg/L amorphous silica saturation window where scale formation becomes autocatalytic. Pushing recovery to 90% without selective ion removal forces concentrate silica above 360 mg/L, which no antiscalant can hold reliably over a membrane's operating life. A dissolved air flotation (DAF) system for pulp-mill pre-treatment is the standard first step: DAF reliably removes 80–95% of TSS and 60–85% of color, and it strips a meaningful fraction of colloidal silica and suspended lignin before the biological stage sees the stream. Getting DAF, biological, and filtration right is what determines whether the downstream RO can even attempt >90% recovery.
The High-Recovery RO Toolkit: Four Paths, One Decision Logic
Once the influent envelope is set, the technology choice collapses to four high-recovery paths the engineer can mix and match. The wcponline 2026 process review groups them into a single decision framework, and they apply cleanly to a kraft-mill retrofit when sized against the table above.
| Path | Mechanism | Typical recovery | Best fit for pulp effluent |
|---|---|---|---|
| 1. Fluidized-bed crystallization reactor (FBCR) | Controlled precipitation on seed particles purges silica, calcium, and barium before RO sees saturation | >90% | High-silica kraft/sulfite streams where silica is the recovery ceiling |
| 2. Cyclic / pulsed-flow RO | Alternating production and high-velocity flush phases shear fouling layers off the membrane | 85–92% | High-organics streams where biofilm and color-body fouling dominate |
| 3. Integrated membrane–thermal hybrid (MLD template) | RO handles 90–95% of feed; only 5–10% of feed reaches evaporator/crystallizer | >95% system recovery | The default MLD train for pulp; minimizes thermal CAPEX |
| 4. Digital process control & predictive maintenance | AI-based scaling-index monitoring, flux trending, and cleaning-cycle optimization | Adds 2–5% recovery headroom | Layered on top of Paths 1–3; not a stand-alone solution |
Path 1 deserves close attention because it directly addresses the silica ceiling. In a fluidized-bed reactor, brine circulates through a bed of seed particles; sparingly soluble salts nucleate on those seeds rather than on the membrane surface, and the loaded seeds are periodically removed and regenerated. The effect is that the RO downstream operates at recoveries often exceeding 90% because the scaling ions have already been purged (per wcponline 2026-01). Path 2 is the kinetic answer to organic fouling: instead of letting fouling layers accumulate at steady state, the system periodically spikes crossflow velocity to dislodge them, which extends membrane life and reduces cleaning frequency. Path 3 is the architectural answer — and it is the one that makes MLD economically defensible for pulp. By accepting that 5–10% of feed must leave as a concentrated brine, the design routes the rest through a high-recovery industrial RO system at membrane-scale energy cost (typically 0.5–1.5 kWh/m³) and only pays thermal energy on the residual fraction. Path 4 is the control layer that keeps any of the first three near the recovery limit without unplanned downtime — particularly valuable on a brownfield retrofit where feed quality drifts seasonally.
Applying the Toolkit to a UPM-Scale Expansion: A Decision Framework
Translating the four paths into a defensible configuration for a kraft expansion requires a decision tree, not a generic "high-recovery is good" recommendation. The logic below is what an EPC would walk a client through in a process selection meeting.
- Is post-biological BOD still > 25 mg/L? If yes, add a polishing MBR or sand filter ahead of RO. A MBR membrane bioreactor for biological pulp pre-treatment typically achieves BOD < 5 mg/L and TSS < 1 mg/L in a single step, which protects downstream RO from organic fouling.
- Is feed silica > 40 mg/L as SiO₂? If yes, mandate a fluidized-bed crystallization reactor (Path 1) or lime softening ahead of RO. Below 40 mg/L, antiscalant chemistry alone can hold conventional RO at 80% recovery; above 40 mg/L, selective removal is non-negotiable.
- Is feed calcium > 150 mg/L or sulfate > 400 mg/L? If yes, FBCR or lime softening is required to prevent CaSO₄ scaling at the concentrate end of the RO train.
- Can the final brine volume be reduced to < 5% of feed by RO + mechanical concentration alone? If yes, full ZLD is justified. If no, route the residual 5–10% to a brine concentrator and small crystallizer — the MLD template.
This is not a theoretical framework. A Chilean power plant documented in the wcponline 2026 review combined a fluidized-bed reactor with cyclic RO on a high-scaling cooling tower blowdown and achieved more than 93% water recovery, with the thermal stage reduced to a minimal polishing step. The influent character is not identical to a kraft line, but the scaling mechanism is — and the field result is the strongest published evidence that the MLD path works at industrial scale on a silica-limited feed. For a UPM-scale kraft expansion, the likely configuration is therefore: DAF → biological (MBR or activated sludge) → multi-media filter as RO protection → high-recovery RO (with fluidized-bed crystallization on the concentrate loop) → brine concentrator → small crystallizer handling 5–10% of feed. The same hybrid logic has been validated for hybrid systems for high-strength organic wastewater in other verticals, which is useful prior art when defending the selection to a project reviewer.
CAPEX, OPEX and Footprint: Why Full ZLD Is Over-Engineered for This Expansion
The board-level question is not "can we do full ZLD?" — it is "does full ZLD pay back compared to MLD + RO for this expansion?" The directional answer is no, and the gap is large enough to defend in a steering meeting. Full ZLD CAPEX is dominated by the evaporation and crystallization stages; routing only 5–10% of feed to thermal equipment cuts the thermal-stage CAPEX by roughly an order of magnitude (Zhongsheng field data, 2026). On the OPEX side, energy is the dominant line item. High-recovery RO with modern energy-recovery devices typically operates at 0.5–1.5 kWh/m³ of permeate, while mechanical vapor recompression — the workhorse of a full ZLD plant — runs 15–25 kWh/m³ of distillate. When the thermal stage is only polishing 5–10% of feed, the annualized energy penalty of going to full ZLD is in the order of millions of euros per year for a UPM-scale flow, before accounting for waste-heat integration opportunities that a hybrid train can exploit. Footprint matters on a brownfield retrofit: a fluidized-bed reactor plus RO skid occupies a small fraction of the area of an equivalent evaporator train, which preserves space for fiber-line expansion rather than consuming it in the water-treatment yard. Treat these as order-of-magnitude guidance — site-specific numbers require a mass-balance and a vendor quote — but the directional argument is consistent: for a 2026 kraft-mill expansion, MLD + high-recovery RO is the engineering-defensible choice, and full ZLD is over-engineering.
Frequently Asked Questions
Does the UPM expansion automatically trigger a regulatory ZLD requirement?
No. EU industrial discharge rules under the IED and the related BAT conclusions govern pulp and paper effluent quality, not water volume. ZLD only becomes a hard requirement if site-specific discharge permits, local water-stress classification, or corporate water-positive commitments force zero liquid effluent — none of which are automatically triggered by a capacity expansion alone (per wcponline 2026-01).
What is the practical difference between MLD and ZLD for a pulp mill?
ZLD means no liquid leaves the site at all; all water is recovered and only solids are managed. MLD uses membrane and mechanical steps to shrink the brine stream by 60–90% before any thermal polishing, so only a small residual fraction (typically 5–10% of feed) reaches an evaporator or crystallizer (per wcponline 2026-01).
What is the realistic RO recovery ceiling for a kraft effluent without selective ion removal?
Conventional RO caps at roughly 70% recovery on a typical kraft feed because silica and calcium sulfate scaling become unmanageable above that point. With a fluidized-bed crystallization reactor or lime softening upstream, recoveries above 90% are achievable, as demonstrated at >93% on a Chilean power-plant blowdown (per wcponline 2026-01).
How is the concentrated brine sludge handled in an MLD train?
The brine concentrator produces a saturated salt stream that feeds a small crystallizer, which yields a solid salt cake for offsite disposal or, in some cases, reuse in the pulp chemical recovery loop. Volume is typically less than 1% of the original feed, which is what makes the MLD configuration economically defensible versus full evaporation.
Does the recommendation change for a greenfield pulp mill versus a brownfield expansion?
The technology selection is the same, but a greenfield site can co-locate the DAF, biological, RO, and thermal stages on a green pad with optimized hydraulic gradients, which can lower installed CAPEX by 10–20%. A brownfield retrofit is constrained by existing piping, footprint, and tie-in windows, which is why the MLD + RO skid approach — with a small crystallizer rather than a full evaporator train — is even more strongly favored on retrofits.
