Membrane Distillation for High-Salinity Brine Concentration

What is membrane distillation and how does it work?

Membrane distillation is a hybrid thermal-membrane process. A hot, usually saline, feed flows on one side of a porous hydrophobic membrane. The membrane material — typically PTFE, PVDF or polypropylene with a pore size of roughly 0.1–0.5 µm — is not wetted by liquid water at ordinary transmembrane pressures, so the pores remain filled with air and vapour rather than liquid. A liquid–vapour interface therefore forms at the mouth of each pore on both faces. Water evaporates at the hot interface, the vapour diffuses across the gas-filled pore, and it condenses (or is swept away) on the cold permeate side. Dissolved, non-volatile species such as Na⁺, Cl⁻, SO₄²⁻ and dissolved organics cannot enter the gas phase, so — provided the membrane stays unwetted — the permeate is essentially pure distilled water. Rejection of non-volatile solutes approaches 100 %.

The distinguishing feature is the driving force. In reverse osmosis (RO), transport is driven by a hydraulic pressure applied in excess of the osmotic pressure. In MD, transport is driven by the difference in water-vapour partial pressure between the two membrane faces, which is created by keeping the two sides at different temperatures. Typical operating temperatures are modest — a hot side of 50–80 °C and a cold side of 20–40 °C — well below boiling, because it is the vapour-pressure gradient, not the boiling point, that matters. This is why MD can exploit solar heat, geothermal fluid, engine jacket water or industrial waste heat that would otherwise be dumped.

The wetting resistance is governed by the Laplace–Young relation: the liquid-entry pressure (LEP) is the hydraulic pressure at which liquid intrudes into the largest pore. It rises with surface tension and with the hydrophobicity (contact angle) of the membrane, and falls as pore size increases. Keeping the operating pressure well below LEP is the single most important condition for reliable MD operation, and it is the reason feeds with surfactants or low-surface-tension organics are challenging.

What drives the flux? The vapour-pressure equation

The vapour transported through the membrane per unit area per unit time — the permeate flux J — is proportional to the difference in water-vapour partial pressure between the feed-side and permeate-side membrane surfaces:

The vapour pressure of pure water rises steeply and non-linearly with temperature, and is well described by the Antoine equation:

For a saline feed the relevant vapour pressure is reduced by the water activity a_w (Raoult’s law, corrected by an activity coefficient): p_f = a_w · p^sat(T_f). For seawater a_w ≈ 0.98; even for a saturated NaCl brine a_w only falls to about 0.75. This is the crux of MD’s advantage: increasing salinity depresses the driving force by only a modest fraction, whereas in RO the osmotic pressure it must overcome rises without bound. The steep temperature dependence of p^sat also means a few extra degrees on the hot side buys disproportionately more flux — doubling the feed temperature from 40 to 60 °C roughly doubles p^sat and hence the driving force.

Worked example: vapour-pressure difference and flux

Consider a direct-contact MD module treating a concentrated brine. Feed-surface temperature T_f,m = 60 °C, permeate-surface temperature T_p,m = 40 °C, brine water activity a_w = 0.90, and a membrane coefficient C_m = 5.0 × 10⁻⁴ kg m⁻² h⁻¹ Pa⁻¹. Work through the driving force step by step.

• Saturation vapour pressure at 60 °C (Antoine): log₁₀ p = 8.07131 − 1730.63 / (233.426 + 60) = 8.07131 − 5.8981 = 2.1732, so p^sat = 149.0 mmHg = 19,860 Pa.
• Feed-surface vapour pressure (salinity-corrected): p_f,m = 0.90 × 19,860 = 17,874 Pa.
• Saturation vapour pressure at 40 °C (Antoine): log₁₀ p = 8.07131 − 1730.63 / (233.426 + 40) = 8.07131 − 6.3299 = 1.7414, so p^sat = 55.1 mmHg = 7,349 Pa. The permeate is pure water, so p_p,m = 7,349 Pa.
• Driving force: Δp = 17,874 − 7,349 = 10,525 Pa.
• Flux: J = C_m × Δp = 5.0 × 10⁻⁴ × 10,525 = 5.3 kg m⁻² h⁻¹ (about 5.3 L m⁻² h⁻¹, i.e. 5.3 LMH).

Note what happens if this were fresh water rather than brine: p_f,m would be the full 19,860 Pa, Δp = 12,511 Pa and J = 6.3 LMH. The salinity penalty in going from pure water all the way to a_w = 0.90 is only about 16 %. In RO, that same brine could already lie beyond the achievable pressure envelope. This near-independence of flux from salinity is what makes MD a natural brine concentrator.

Two corrections matter in practice. First, temperature polarisation: because heat is consumed by evaporation at the feed interface, the surface temperatures T_f,m and T_p,m are less extreme than the bulk temperatures, shrinking Δp. The temperature polarisation coefficient (interface ΔT over bulk ΔT) is often only 0.4–0.7, so the bulk temperatures must be more separated than the surface figures used above. Second, concentration polarisation raises salinity at the wall, lowering a_w further. Good module hydraulics — high cross-flow velocity, turbulence promoters — mitigate both.

What are the MD configurations (DCMD, AGMD, VMD, SGMD)?

MD variants differ only in how the vapour is condensed or removed on the permeate side. That choice trades flux against conductive heat loss and against ease of heat recovery.

For ZLD and brine concentration, AGMD and VMD dominate practical designs. AGMD is favoured where internal latent-heat recovery is wanted and a compact multi-effect stack can be built; VMD is favoured where maximum flux per unit area justifies the extra vacuum and condensing hardware. DCMD remains the workhorse of laboratory studies because of its mechanical simplicity, but its conductive heat loss makes it less attractive at scale. Whichever configuration is chosen, the same vapour-pressure driving force governs flux; the engineering differences are about heat management, not mass transfer physics.

Why is MD good for ZLD and brine concentration?

The economic wall in most brine-management schemes is the point where RO can no longer push against osmotic pressure. A seawater RO train typically stops around 70–75 g/L total dissolved solids (TDS) because the osmotic pressure of the reject then approaches the practical membrane and pump pressure limit of roughly 70–80 bar. Beyond that, conventional practice reaches for thermal evaporators (mechanical vapour recompression, brine concentrators) that are energy- and capital-intensive. MD slots into precisely this gap: it can take an RO reject at 70 g/L and drive it on toward saturation (260 g/L for NaCl, higher still for mixed brines) because — as the worked example showed — flux barely responds to salinity.

Two properties make MD attractive as the brine-concentration stage in a zero-liquid-discharge treatment train:

• Salinity-tolerant flux. The driving force is depressed only by water activity, which for real brines falls slowly. A process that keeps working at 200 g/L is worth a great deal when the alternative is a multi-stage flash or MVR evaporator.
• Low-grade heat. MD runs on 50–80 °C heat. Solar thermal, geothermal, condenser reject, flue-gas heat recovery or CHP jacket water can supply it, so the marginal energy cost can approach zero where waste heat exists.

Coupled with a crystalliser, MD becomes a membrane distillation crystalliser (MDC). As the retentate is concentrated past saturation, salts nucleate and are harvested as a solid product — sodium chloride, sodium sulphate or mixed salts — while distilled water leaves as permeate. This is the essence of resource recovery: the waste stream is split into clean water and a saleable or landfillable solid, with no liquid discharge. MD sits naturally alongside other emerging concentration methods such as high-recovery membrane processes and complements pressure-driven and osmotically driven steps described in our note on forward osmosis, which can pre-concentrate a feed before the thermal polish. Facilities pursuing higher water recovery in seawater desalination increasingly evaluate MD as the tail-end concentrator that turns brine disposal into product recovery.

Worked example: thermal energy and GOR per cubic metre

MD’s Achilles heel is thermal efficiency, so the energy bookkeeping deserves a worked figure. The unavoidable minimum is the latent heat of vaporisation of water, λ ≈ 2,320 kJ/kg at 60 °C. To evaporate 1 m³ (1,000 kg) of permeate:

• Latent heat only: 1,000 kg × 2,320 kJ/kg = 2,320,000 kJ = 2,320 MJ = 644 kWh_th per m³ in the ideal, single-pass, no-recovery case.
• With sensible-heat and polarisation losses: a single-pass DCMD module recovers little latent heat, so the specific thermal energy is often 1,000–3,000 kWh_th/m³ — far above the latent minimum.

Efficiency is expressed as the gained output ratio (GOR), the mass of distillate produced per unit mass of steam-equivalent heat input, or equivalently the latent heat of the product divided by the actual heat supplied:

Take a multi-stage AGMD unit at GOR = 5. The specific thermal energy becomes 644 kWh_th / 5 = 129 kWh_th per m³ of distillate. If that heat is genuine waste heat valued at near zero, the operating cost collapses to the electricity for pumps and any vacuum duty (typically 0.5–2 kWh_e/m³) — competitive with thermal evaporation and, in a waste-heat-rich site, cheaper. If, instead, the heat must be purchased as steam or gas at 40 £/MWh_th, that same 129 kWh_th/m³ costs about 5.2 £/m³ in energy alone — which is why MD is rarely economic without cheap or free heat. The design lesson is blunt: MD’s viability is decided by the price and availability of low-grade heat and by the GOR the module achieves, not by membrane cost.

Worked example: concentration factor versus the RO limit

How much further than RO can MD actually take a feed? Define the volumetric concentration factor CF = V_feed / V_concentrate, which for a fully rejected solute equals the ratio of final to initial TDS. Start from a seawater feed at 35 g/L TDS.

• RO stage: practical seawater RO concentrates to about 70 g/L before osmotic pressure (~55 bar at 70 g/L, rising sharply above) meets the pressure ceiling. CF_RO = 70 / 35 = 2.0, i.e. about 50 % water recovery.
• MD polishing stage: take the 70 g/L RO reject and concentrate toward the NaCl solubility limit, say 250 g/L before crystallisation. CF_MD = 250 / 70 = 3.6.
• Overall: CF_total = 250 / 35 = 7.1. Water recovery across the combined train is 1 − (35/250) = 86 %, versus 50 % for RO alone.

The remaining 14 % is a small, near-saturated brine ideal for a crystalliser, closing the loop to true ZLD. To see why RO cannot do this alone, estimate the osmotic pressure of a 250 g/L brine using the van’t Hoff approximation Π ≈ i M R T. For NaCl, i = 2, molarity M ≈ 4.3 mol/L, giving Π ≈ 2 × 4,300 mol/m³ × 8.314 J mol⁻¹ K⁻¹ × 313 K ≈ 2.24 × 10⁷ Pa ≈ 224 bar. No practical RO membrane or pump operates against 224 bar; MD is indifferent to it because its driving force is thermal, not hydraulic. That single number captures why MD, MDC and thermal concentration exist at all.

What are the limitations and failure modes of MD?

MD is not a universal answer, and honest engineering means naming its weaknesses:

• Low thermal efficiency. As shown, MD moves large quantities of latent heat. Without effective heat recovery (high GOR) or genuinely free waste heat, its energy cost is uncompetitive with MVR evaporation.
• Temperature polarisation. Evaporative cooling at the feed interface and warming at the permeate interface reduce the surface ΔT below the bulk ΔT, cutting the driving force by a factor that can approach two. It is an intrinsic penalty, mitigated only by better module hydraulics, never eliminated.
• Membrane wetting. If the operating pressure exceeds the liquid-entry pressure, or if surfactants, oils or low-surface-tension organics lower LEP, liquid intrudes into the pores. Wetting destroys rejection: salt passes and permeate quality collapses. Feed pre-treatment to remove oils and surfactants is essential.
• Fouling and scaling. At the high concentration factors MD is prized for, sparingly soluble salts (CaSO₄, CaCO₃, silica) scale the membrane, and organic/biofouling adds resistance and promotes wetting. Antiscalant dosing, careful staging and periodic cleaning are required.

The practical envelope for MD is therefore a feed that is clean of oils and surfactants, a site with abundant low-grade heat, and a duty that specifically needs concentration beyond the RO ceiling. Where all three hold — RO-reject polishing, produced-water and mining brines, resource recovery from spent process liquors — MD earns its place. Where they do not, a conventional evaporator or an RO-only scheme is usually the sounder choice.