Zero Liquid Discharge Is a Sequencing Problem: Why the Membrane-to-Thermal Handoff Point Is the Real Design Decision

Most ZLD comparisons pit MVR against MEE against membrane distillation as if picking one settles the design. The bigger lever is upstream: every extra point of recovery a membrane train squeezes out before handoff removes disproportionate capital and energy from whatever thermal step comes after it.

Zero liquid discharge gets discussed like a technology choice — multi-effect evaporation versus mechanical vapor recompression versus membrane distillation, pick the winner. That framing skips the decision that actually determines cost: how much of the concentration work gets done by membranes before the stream is handed off to a thermal or crystallization step at all. A ZLD train that pushes recovery from 80% to 92% on the membrane side before handoff can shrink the downstream evaporator’s feed volume by more than half — and evaporation capital and energy scale with volume, not with cleverness. The membrane-versus-thermal technology debate matters less than getting that handoff point right for the feedwater actually in front of you.

This piece works through the building blocks in the order they usually sit in a real train — membrane concentration, then thermal crystallization — what stops each stage from going further on its own, and where a well-known industrial case (a four-stage UF+NF+RO+MVR system treating power-plant recirculating cooling water and flue-gas desulfurization wastewater) actually landed. It also builds on the RO/MED/MSF thermal comparison published on this site: the same osmotic-pressure ceiling and cascade-evaporation thermodynamics discussed there for seawater desalination apply directly to brine concentration and crystallization, just against a much more concentrated starting stream.

Where membranes stop: the osmotic-pressure wall

Reverse osmosis is the natural first concentration step for most ZLD feeds, for the same physical reason it dominates seawater desalination: as covered in this site’s RO vs. MED/MSF comparison, water flux through an RO membrane follows the net effective pressure differential, $J_w = A(\Delta P - \Delta \Pi)$ — and as the retained stream concentrates, its osmotic pressure $\Pi$ climbs until it approaches the membrane’s maximum practical operating pressure. Above roughly 70 g/L TDS, most commercial RO elements simply run out of usable driving force; pushing pressure higher to compensate runs into mechanical limits and accelerating membrane compaction. That ceiling, not chemistry, is why RO alone cannot reach zero discharge on a genuinely saline feed — it can concentrate a stream a long way, but something else has to finish the job. Readers sizing this concentration stage — recovery ratio, concentrate TDS, and osmotic-pressure checks across a multi-stage array — can work the mass balance with this site’s RO System Designer.

Electrodialysis (ED) sidesteps that specific limit because it doesn’t rely on pressure at all: an applied electric field drives ion migration across ion-selective membranes, so concentration isn’t capped by osmotic back-pressure the way RO is. ED can push a stream well beyond RO’s practical ceiling, and it’s particularly effective when the target is a specific, separable ion pair (Cl⁻/Na⁺ concentration for a relatively clean brine, for instance). The tradeoff is that ED’s power draw rises with the stream’s electrical resistance — meaning it gets more expensive per unit of additional concentration precisely as it does its job, and ion-exchange membrane and electrode wear adds a real maintenance line RO doesn’t have. In practice ED is deployed as a second concentration stage on RO’s reject, not as a standalone front end.

Nanofiltration (NF) plays a different role in this sequence: it’s usually placed before RO or ED, not after, specifically to strip divalent hardness ions (Ca²⁺, Mg²⁺, SO₄²⁻) out of the feed via charge-based rejection. Removing those ions early lowers the calcium sulfate and similar scaling risk in every downstream stage — RO, ED, and the evaporator alike — letting each one run safely at a higher limiting recovery than the raw feed chemistry would otherwise allow. NF also has a standalone use later in the train: selective salt-splitting of a mixed-salt reject stream into separate monovalent-rich and divalent-rich draws, which matters when the plant wants to recover distinct, sellable salt products rather than one mixed sludge. Readers modeling this stage — independent monovalent/divalent rejection tracking across a multi-stage array — can work the mass balance with this site’s NF System Designer.

Ahead of all of it sits ultrafiltration or microfiltration (UF/MF), whose job isn’t concentration at all — it’s protecting everything downstream from suspended solids and colloids that would otherwise foul a spiral-wound RO element or an ED stack within days. Sizing that pretreatment stage against the actual design flow, including the recovery lost to backwash water, can be worked through this site’s UF/MF System Designer.

Where thermal picks up: three ways to finish the job

Once a membrane train has pushed a feed as far as it economically can — commonly somewhere past 90% overall recovery — what’s left is a small-volume, high-salinity reject that has to go to dryness by some combination of evaporation and crystallization.

Multi-effect evaporation (MEE) is mechanically the same cascade principle covered in this site’s MED discussion for seawater desalination: motive steam condenses in a first vessel, boiling the feed and generating secondary vapor that’s slightly cooler than the motive steam, which is used to boil the feed in a second vessel held at correspondingly lower pressure, and so on down a cascade of typically 3-8 effects. Each unit of input steam energy does its evaporation work multiple times as it cascades, which is the entire efficiency argument — a well-designed multi-effect train cuts steam consumption 50-70% versus a single-effect evaporator running the same duty. MEE scales well to large continuous flows and is the most mechanically mature option here, but it needs a real steam supply, tolerates scale-forming chemistry poorly at the higher-concentration end, and its heat-transfer surfaces are the first thing to foul if upstream softening wasn’t thorough enough.

Mechanical vapor recompression (MVR) removes the external steam dependency entirely: a compressor takes the vapor boiled off the feed, compresses it (raising both its pressure and its condensing temperature), and feeds that same vapor back to the same vessel’s heating side as its own heat source. Only a small addition of steam is needed at startup or to cover heat losses; essentially all of the process’s thermal energy is continuously recycled by the compressor. That trades steam for electricity — a compressor-driven system reduces total energy consumption by roughly 60-80% relative to a single-effect evaporator, and cooling-water demand drops by more than 90% since there’s no final condenser rejecting a full vapor load to a cooling tower. The real constraint is the compressor itself: it needs stable, adequately-sized electrical supply, tolerates entrained non-condensable gas poorly, and its maintenance cost and vibration/noise footprint are a genuine operating consideration at scale that a simple gravity-fed evaporator doesn’t have.

Because MEE trades in steam and MVR trades in electricity, comparing them honestly on energy terms runs into the same problem this site’s RO/MED/MSF piece worked through for seawater thermal desalination — high-grade electrical work and low-grade thermal energy aren’t directly comparable kWh-for-kWh, and the Standard Primary Energy framework introduced there (converting both back to the primary fuel actually consumed to produce them) applies here without modification. A site with genuinely cheap waste steam available should weight MEE more heavily than the raw energy numbers alone would suggest; a site paying full retail for both power and steam should run the same primary-energy conversion before assuming MVR’s efficiency framing settles the comparison.

Membrane distillation (MD) is the outlier of the three: it’s not really competing on energy efficiency, it’s competing on where it can still function. MD uses a hydrophobic microporous membrane that passes water vapor but rejects the liquid phase, driven by a temperature difference across the membrane (typically just 50-80°C on each side) rather than by hydraulic pressure. Because rejection is governed by the membrane’s vapor-liquid interface rather than by osmotic pressure, MD’s salt rejection stays above 99.9% essentially regardless of how concentrated the feed already is — it keeps working on near-saturated brine that would stall RO and strain ED. That property, plus its tolerance for low-grade heat (waste heat, solar thermal, geothermal — anything hot enough to sustain the temperature differential), makes it a natural fit for the last stretch of concentration before crystallization, where the feed is too saline for anything pressure-driven. The tradeoff is throughput: vapor flux per unit membrane area is modest compared to RO’s water flux, so MD needs a large membrane area for a given flow, and that area is also where fouling and pore wetting show up first — it needs real cleaning discipline and has a shorter practical membrane life than the equipment-heavy alternatives. MD is also the one process in this comparison genuinely sensitive to volatile organics in the feed: since it transports vapor, not just water, volatile organic compounds cross the membrane along with the water unless they’re stripped out upstream, which is a real design consideration whenever the feed source is chemical- or solvent-adjacent.

What actually breaks a ZLD train

Across every stage — RO, ED, NF, and both evaporation routes — the two failure modes that actually shut a plant down are scale formation and organic/biological fouling, not a fundamental limit of the chosen technology. Continuous antiscalant dosing (phosphonate or polyacrylate chemistry, raising the nucleation barrier for sparingly-soluble salts before they crystallize on a membrane or heat-transfer surface) and periodic pH adjustment are standard across the whole train; readers sizing these dosing rates can use this site’s Chemical Dosing Calculator. For plants where a cooling-tower loop feeds into the same zero-discharge boundary — a common configuration in power and heavy-industrial sites, and specifically what the universal-process-zld scenario on this site covers — scale risk in that loop can be checked directly with the Cooling Tower Water Balance & LSI tool.

Heavy metals and high organic loading need to be dealt with before the feed reaches the membrane or thermal stages at all, not managed within them. Chemical precipitation, electrocoagulation, or ion exchange ahead of the main train pulls recoverable metals out as a separate, more concentrated stream and prevents them from co-precipitating into the final salt product (where they’d contaminate an otherwise sellable output) or fouling a membrane surface. Readers sizing ion-exchange vessels and regeneration chemical demand for this pretreatment role can use this site’s Ion Exchange Resin & CEDI Designer. High organic loading gets handled similarly — advanced oxidation or activated carbon adsorption ahead of the membrane train, both to protect membrane surfaces and to keep organics from ending up in a nominally “pure” salt product downstream.

Feedwater type, technology fit

Feed characteristic	Recommended handling	Why
High-salinity, inorganic-dominated (low COD)	NF softening → RO or ED concentration → MVR or MEE crystallization	Membranes handle the bulk concentration cheaply; thermal stage only needs to finish a pre-concentrated, pre-softened stream
High organic loading	Advanced oxidation or activated carbon pretreatment → RO or MD	Protects membrane surfaces and prevents organics from co-precipitating into the salt product
Heavy metals present	Precipitation, electrocoagulation, or ion exchange first, then standard ZLD train on the metal-free remainder	Recovers metals as a separable stream and keeps them out of the final salt and off membrane surfaces
Volatile organics present	MD (vacuum/air-gap configuration) or low-temperature vacuum evaporation with condensate recovery, not open thermal evaporation	Prevents stripping VOCs to atmosphere, which most air-emissions permits treat as a separate violation from the water discharge itself
Mixed/complex feed with several of the above	Staged pretreatment (metals, then organics) → membrane concentration → thermal finishing	Each contaminant class is removed at the stage best suited to it rather than forcing one technology to handle everything

A four-stage system at scale: recirculating cooling water and FGD wastewater

A documented industrial case worth citing directly: a power-plant site combining recirculating cooling water blowdown with flue-gas desulfurization (FGD) wastewater — a feed running 15,000-25,000 mg/L TDS, well past what a single technology handles cleanly — used a four-stage UF → NF → RO → MVR train to close the loop. UF protected the NF/RO stages from suspended solids; NF softened the stream ahead of RO to control scale-forming divalent ions; RO did the bulk of the volume reduction within its pressure ceiling; and MVR crystallized the RO reject to dryness, sized against a feed volume the upstream membrane stages had already cut substantially. The design logic tracks the thesis above directly: each membrane stage did as much of the concentration work as it could economically justify, which minimized the volume — and therefore the capital and energy — that the MVR crystallizer actually had to process. FGD wastewater specifically is a useful stress case because it combines high sulfate and chloride loading with trace heavy metals scrubbed out of the flue gas, meaning the pretreatment sequencing (metals removal before the main membrane train) matters as much as the concentration technology choice itself.

The regulatory case for getting this right

None of the major jurisdictions actually mandate zero liquid discharge by name, but the direction of travel makes ZLD-capable design the safer long-term bet. China’s Water Pollution Prevention and Control Law (Article 52) requires discharge to meet permit standards before release, and the newer Technical Specification for Zero Discharge of Industrial Circulating Cooling Water (GB/T 43328-2024) formalizes recirculation and near-zero-discharge practice specifically for industrial cooling systems — directly relevant to the cooling-tower-integrated ZLD scenario covered above. China’s current five-year water-conservation planning target calls for over 95% water-recycling technology penetration in high-water-use industries and over 30% near-zero-discharge adoption across industrial parks.

The EU’s Industrial Emissions Directive (2010/75/EU) requires facilities to operate to Best Available Techniques, and BAT reference documents (BREFs) for several sectors explicitly name wastewater minimization — including zero liquid discharge — as a technique to be considered. The Water Framework Directive (2000/60/EC) sets a 2027 target for good ecological status across EU surface and groundwater, which pushes discharge permitting in the same direction independent of any single ZLD mandate. In the US, the Clean Water Act’s NPDES permitting system and EPA’s sector-specific Effluent Limitations Guidelines don’t require ZLD outright, but several water-stressed states and specific sectors (electronics manufacturing, textile finishing) increasingly use permit conditions that are difficult to meet without something close to it.

The one factor cutting against ZLD adoption broadly is energy: crystallization-based ZLD is energy-intensive by nature, which matters directly for facilities also tracking a decarbonization target. Coupling the thermal stage to waste heat, solar thermal, or renewable electricity — rather than treating ZLD energy demand as a fixed cost — is where the regulatory and climate cases actually reinforce each other instead of trading off.

What this means for a real design decision

The technology-selection question (“MVR or MEE or MD?”) is real but secondary. The question that actually sets the project’s capital and operating cost is how far upstream membrane concentration should be pushed before handoff — and that answer is set by feedwater chemistry, not by a general preference for membranes over thermal equipment. A clean, largely inorganic high-salinity stream can be pushed hard on NF/RO/ED before handoff, minimizing what the thermal stage has to process. A stream with heavy organic loading, metals, or volatile compounds needs targeted pretreatment before it even reaches the concentration stage, and — for the volatile-compound case specifically — needs a thermal technology (MD, vacuum evaporation) chosen for its ability to contain vapor-phase losses rather than for its energy efficiency alone. Sequencing the train around the feedwater’s actual constraints, rather than starting from “which evaporator technology is best,” is what keeps a ZLD system’s capital cost proportional to the problem it’s actually solving.

Municipal & Utilities

Heavy Industrial & Energy

Advanced Tech & Manufacturing

Life Sciences & Commercial