Membrane separations

In short

A membrane separation uses a thin selective barrier that lets some species through faster than others under a driving force — pressure, concentration, partial pressure or an electric field — so the feed splits into a permeate and a retentate (concentrate) without any phase change in most cases. Because it needs no latent heat, it competes directly with distillation and evaporation on energy, and now dominates desalination, water reuse, dairy concentration and industrial gas separation. In Indian curricula it appears in the Mass Transfer-II / Separation Processes paper and in GATE-CH under "membrane separations" within mass transfer and in plant-design questions on ZLD. Employers who test it: VA Tech Wabag, Thermax, Ion Exchange (India), Permionics, Praj, Doshion, L&T Water, Aditya Birla, Reliance, Amul/dairy plants and the pharma-water groups at Sun Pharma and Cipla.

Explained

What a membrane is and how it separates

• Membrane: a selective interphase between two bulk phases. Separation is by size (sieving), solubility-diffusion, charge (Donnan exclusion) or differential volatility, depending on the process.
• Classification by structure: *porous* (micro/ultrafiltration — sieving), *dense* (RO, gas permeation, pervaporation — solution-diffusion), *ion-exchange* (electrodialysis), *liquid membranes* (supported/emulsion).
• Classification by morphology: symmetric (uniform) and asymmetric — a 0.1–1 µm dense skin on a 100–200 µm porous support, so resistance sits in the skin while strength comes from the support. Thin-film composite (TFC) RO membranes have a cross-linked aromatic polyamide skin on polysulfone on polyester fabric.
• Materials: cellulose acetate (chlorine-tolerant, pH 4–6 only), polyamide/TFC (pH 2–11, chlorine-intolerant — needs SMBS dechlorination), PVDF, PES, polysulfone, polypropylene (MF/UF); ceramic (alumina, zirconia, titania) for high temperature, solvents and steam sterilisation; palladium and zeolite for H₂; silicone rubber (PDMS) for pervaporation of organics; PVA for ethanol dehydration.
• Terms: flux J (L m⁻² h⁻¹, "LMH", or kg m⁻² h⁻¹); permeance/permeability; rejection R; recovery (conversion) Y = permeate/feed; selectivity α; MWCO — molecular weight at which 90% rejection occurs; transmembrane pressure TMP.

The pressure-driven family (MF, UF, NF, RO)

• Osmosis and RO: solvent moves from dilute to concentrated across a semi-permeable membrane until the osmotic pressure π is balanced. RO applies ΔP > Δπ to drive water the *other* way. Net driving force = (ΔP − Δπ); operating pressure for seawater (35,000 ppm, π ≈ 25–28 bar) is 55–70 bar, for brackish water 10–20 bar.
• Solution-diffusion model: the solute/solvent dissolves in the dense skin, diffuses down a concentration (activity) gradient and desorbs. Water flux depends on (ΔP − Δπ), salt flux on Δc only — which is why rejection improves as pressure rises and worsens as recovery (hence feed concentration) rises.
• Nanofiltration ("loose RO"/softening membrane): rejects 90–98% of Ca²⁺, Mg²⁺, SO₄²⁻ but only 20–60% of NaCl, via Donnan exclusion by its fixed negative charge. Used for hardness removal, dye desalting, textile effluent colour removal and in ZLD to split monovalent from divalent salts.
• Pore-flow (MF/UF) flux: laminar flow through pores, described by Hagen-Poiseuille/Darcy; flux ∝ TMP initially, then plateaus at a limiting flux set by the cake/polarisation layer, not by pressure.

Concentration polarisation and fouling — the two things exams always ask

• Concentration polarisation: rejected solute accumulates at the membrane wall, so wall concentration c_w > bulk c_b. It is reversible and instantaneous; it raises local osmotic pressure (cutting net driving force), increases salt passage and promotes scaling. Quantified by the film model: J = k ln[(c_w − c_p)/(c_b − c_p)]; polarisation modulus = c_w/c_b.
• Remedies: raise cross-flow velocity (higher mass-transfer coefficient k), use feed spacers/turbulence promoters, pulsed or reverse flow, lower flux, raise temperature.
• Fouling: largely irreversible flux decline from (i) particulate/colloidal (silt — monitor by SDI, keep SDI₁₅ < 3 and turbidity < 1 NTU for RO), (ii) organic (humics, oil, proteins), (iii) biofouling (biofilm — the usual killer in Indian surface-water plants), (iv) scaling (CaCO₃, CaSO₄, BaSO₄, SiO₂ — controlled by LSI, antiscalant dosing and recovery limits).
• Pretreatment train: screening → coagulation/clarification or DAF → multigrade/dual-media filter → activated carbon or SMBS dechlorination → 5 µm cartridge → high-pressure pump → RO. UF is increasingly used as RO pretreatment in place of media filters.
• Cleaning (CIP): alkaline (pH 10–12, NaOH + EDTA/surfactant) for organics and biofilm; acid (citric, HCl, pH 2–3) for scale; backflush and air scour for MF/UF; forward flush; chemically enhanced backwash. Normalised flux and normalised salt passage trigger a clean (typically 10–15% flux loss or 10% ΔP rise).
• Modes: dead-end (all feed becomes permeate; filtration only), cross-flow (feed sweeps the surface — standard in RO/NF/UF), diafiltration (wash water added to purify a retentate).

Modules, arrays and staging

• Plate-and-frame: easy to clean, low packing density (~100–400 m²/m³); viscous and fouling feeds.
• Tubular (12–25 mm): most fouling-tolerant, lowest packing density (~20–200 m²/m³), high pumping cost; ceramic tubes for oily and hot effluents.
• Spiral-wound: leaves of membrane + permeate spacer wound round a central permeate tube; ~300–1000 m²/m³; the industry standard for RO/NF — standard element 8 in × 40 in, ~37 m² (400 ft²).
• Hollow fibre: 10,000–30,000 m²/m³, cheapest per m², but needs very clean feed; shell-side feed for seawater RO, lumen-side for UF/dialysis; used in gas separation and haemodialysis.
• Staging vs. staging of permeate: a multi-stage (tapered) array — e.g. 2:1 with pressure vessels in series (4 elements each) — raises recovery; two-pass puts permeate of pass 1 through a second RO for very low TDS (used for pharma/boiler water). Energy recovery devices (pressure exchanger, Pelton turbine) recover brine energy in seawater RO, cutting specific energy from ~8 to 2.5–4 kWh/m³.
• Flow patterns for calculations: co-current, counter-current, cross-flow, perfect mixing — counter-current gives the best separation in gas permeation.

The other driving forces

• Dialysis: concentration gradient; solute diffuses to a dialysate. Haemodialysis is the largest application (cellulose/polysulfone hollow fibres); also acid recovery.
• Electrodialysis (ED): DC field moves ions through alternating cation- and anion-exchange membranes into concentrate compartments; energy ∝ salt removed, so it suits brackish water (1,000–5,000 ppm), not seawater. EDR reverses polarity periodically to control scaling. EDI (electrodeionisation) replaces mixed-bed polishing after RO and needs no regenerant chemicals.
• Gas permeation: dense polymer; separation by permeability = diffusivity × solubility, selectivity α = P_A/P_B. Glassy polymers (polysulfone, polyimide, cellulose acetate) separate by size — H₂, He and CO₂ are "fast" gases, CH₄ and N₂ "slow". Uses: H₂ recovery from refinery purge, CO₂ removal from biogas/natural gas, N₂ generation from air (O₂ is faster, N₂ is the retentate product), air dehydration. The Robeson plot shows the permeability-selectivity trade-off.
• Pervaporation: liquid feed, vapour permeate pulled off under vacuum; selectivity comes from sorption + diffusion, so it breaks azeotropes — PVA membranes dehydrate ethanol past the 95.6% azeotrope, and PDMS removes traces of VOC/aroma from water. Energy = latent heat of the small permeate only.
• Membrane distillation: hydrophobic porous membrane (PTFE/PVDF), vapour crosses under a temperature-driven vapour-pressure difference; liquid water cannot wet the pores. Runs on waste heat/solar; used in ZLD brine concentration (pilot scale in India).
• Forward osmosis: a draw solution supplies the osmotic driving force; no hydraulic pressure, but the draw must be regenerated.
• Membrane bioreactor (MBR): submerged UF/MF replaces the secondary clarifier in activated sludge, giving a compact plant and permeate fit for reuse — now common in Indian STPs, hotels and townships.

Indian practice, standards and regulation

• Drinking water: IS 10500:2012 (with amendments) sets TDS 500 mg/L desirable (2,000 mg/L permissible if no alternative), fluoride 1.0 mg/L, nitrate 45 mg/L, arsenic 0.01 mg/L — these limits are what RO plants are designed to meet. IS 16240 covers household RO systems. BIS certification applies to packaged drinking water (IS 14543) and packaged natural mineral water (IS 13428).
• Pharmaceutical water: IP (Indian Pharmacopoeia) Purified Water and Water for Injection — typically RO + EDI or RO + distillation; conductivity < 1.3 µS/cm at 25 °C and TOC limits; validated per Schedule M of the Drugs and Cosmetics Rules (revised Schedule M is being enforced from 2026 for small units, with staggered dates — check the current notification).
• Effluent and ZLD: CPCB/SPCB consent under the Water (Prevention and Control of Pollution) Act, 1974 and the Environment (Protection) Act, 1986. ZLD is mandated by CPCB for highly polluting sectors — textile dyeing, tanneries and distilleries in several states — and the standard train is UF → RO → NF salt splitting → MEE/ATFD. Textile clusters in Tirupur and the Pali/Jodhpur CETPs are the standard Indian case studies.
• Large plants: Chennai's Nemmeli and Minjur seawater RO plants (100 MLD each) with a 400 MLD plant at Perur under construction; Jamnagar (Reliance) and Dahej desalination; Tirupur CETPs; Delhi Jal Board and Bengaluru reuse schemes. Note that plant capacities and commissioning dates change — verify current figures before quoting.
• A membrane is not a pressure vessel in the code sense, but the RO skid's high-pressure piping and vessels fall under relevant pressure-vessel design and testing practice; FRP pressure vessels for 8-inch elements are supplied to ASME RTP/Section X-type standards.

Key points

• Membrane - a selective barrier splitting feed into permeate and retentate; no latent heat needed, so it is energy-cheap.
• Driving forces - pressure (MF/UF/NF/RO), concentration (dialysis, FO), electrical (ED/EDI), partial pressure (gas permeation, pervaporation, MD).
• Asymmetric/TFC structure - thin dense skin gives selectivity and high flux; porous support gives strength.
• MWCO - molecular weight rejected 90%; used to specify UF/NF, meaningless for dense RO.
• Rejection R = 1 − c_p/c_f; RO gives 97–99.7% NaCl rejection, NF only 20–60% for NaCl.
• Recovery Y = Q_p/Q_f; higher recovery means higher brine concentration, higher π, lower flux and worse permeate quality.
• Net driving pressure NDP = ΔP − Δπ − ΔP_permeate; flux goes to zero when ΔP ≤ Δπ.
• Osmotic pressure rule of thumb - about 0.7–0.8 bar per 1,000 mg/L TDS for NaCl-type water.
• Concentration polarisation - reversible wall build-up; cured by hydrodynamics (cross-flow velocity), not by cleaning.
• Fouling - irreversible flux decline; four types (particulate, organic, biological, scaling); controlled by pretreatment, antiscalant and CIP.
• SDI₁₅ < 3 - the standard feed-quality target for spiral-wound RO; < 5 is the absolute maximum.
• Chlorine kills polyamide membranes - dechlorinate with SMBS or carbon; cellulose acetate tolerates ~1 mg/L chlorine.
• Spiral-wound is the default RO module (8" × 40", ~37 m²); tubular/ceramic for dirty feeds; hollow fibre for clean feeds and gases.
• Energy recovery device cuts seawater RO energy to ~2.5–4 kWh/m³ against ~10+ kWh/m³ for thermal desalination.
• Electrodialysis energy scales with salt removed, so it beats RO only at low-to-moderate TDS.
• Pervaporation breaks azeotropes - selectivity is sorption-based, not vapour-liquid equilibrium based.
• Temperature correction - flux rises roughly 2.5–3% per °C; performance is normalised to 25 °C before judging a membrane.
• ZLD - UF → RO → NF → MEE/ATFD; mandated by CPCB for textile, tannery and distillery sectors in several states.

Formulas and examples

Example 1 — osmotic pressure and minimum pressure. Brackish water with 3,000 mg/L NaCl at 25 °C.

• Molar concentration C = 3.0 kg/m³ ÷ 58.5 kg/kmol = 0.0513 kmol/m³.
• π = i C R T = 2 × 0.0513 × 0.08314 × 298 = 2.54 bar.
• At 75% recovery the brine is concentrated ~4× → c_r ≈ 12,000 mg/L, π_brine ≈ 10.2 bar. So the pump must deliver well above 10.2 bar at the tail element; a typical design pressure is 15–17 bar. (Check: 0.85 bar per 1,000 mg/L — consistent with the rule of thumb.)

Example 2 — flux, area and element count. A membrane has A = 3.0 LMH/bar. Feed 2,000 mg/L (π = 1.7 bar), operated at ΔP = 14 bar; average brine-side osmotic pressure 3.0 bar; permeate π negligible.

• NDP = 14 − 3.0 = 11.0 bar → J_w = 3.0 × 11.0 = 33 LMH.
• For 50 m³/h permeate: area = 50,000 L/h ÷ 33 LMH = 1,515 m².
• With 37 m² per 8"×40" element → 1,515/37 = 41 elements → 42 elements in 7 vessels of 6 (a 4:3 tapered array).

Example 3 — recovery, rejection and permeate quality. Feed 100 m³/h at 5,000 mg/L TDS, recovery 75%, membrane rejection 98%.

• Q_p = 75 m³/h, Q_r = 25 m³/h.
• Salt balance: 100 × 5,000 = 500,000 g/h in. Permeate c_p = (1 − R) × average feed-side concentration. Using the standard array formula c_p ≈ c_f (1 − R)·CF-weighted value; take the simple balance with c_p from rejection on the *feed*: c_p ≈ 0.02 × 5,000 = 100 mg/L first-pass.
• Brine: c_r = (100 × 5,000 − 75 × 100)/25 = (500,000 − 7,500)/25 = 19,700 mg/L.
• Concentration factor = 19,700/5,000 = 3.94 (close to 1/(1−Y) = 4.0, as expected for high rejection). Check against the CaSO₄/BaSO₄ solubility limits — antiscalant will be needed.

Example 4 — concentration polarisation. J = 36 LMH, k = 2.0 × 10⁻⁵ m/s, c_p ≈ 0.

• Convert J: 36 L m⁻² h⁻¹ = 0.036 m³ m⁻² h⁻¹ ÷ 3600 = 1.0 × 10⁻⁵ m/s.
• c_w/c_b = exp(J/k) = exp(1.0×10⁻⁵/2.0×10⁻⁵) = exp(0.5) = 1.65.
• So the wall sees 65% more salt than the bulk: osmotic pressure at the wall is 65% higher and salt passage rises in proportion. Doubling cross-flow velocity raises k by about 2^0.8 = 1.74× → J/k = 0.287 → modulus 1.33.

Example 5 — gas permeation, nitrogen from air. A polysulfone fibre has O₂ permeance 2.0 GPU and α(O₂/N₂) = 6, so N₂ permeance = 0.333 GPU. Feed air at 8 bar, permeate at 1 bar, area 100 m².

• Using Q = permeance × A × Δp with 1 GPU = 2.7 × 10⁻³ m³(STP) m⁻² h⁻¹ bar⁻¹ (approximate conversion).
• O₂: Δp ≈ 8 × 0.21 − 1 × y_O₂. For a low stage cut, y_O₂ ≈ 0.6 → Δp = 1.68 − 0.60 = 1.08 bar → Q_O₂ = 2.0 × 2.7e-3 × 100 × 1.08 = 0.58 m³(STP)/h.
• N₂: Δp = 8 × 0.79 − 1 × 0.40 = 6.32 − 0.40 = 5.92 bar → Q_N₂ = 0.333 × 2.7e-3 × 100 × 5.92 = 0.53 m³(STP)/h.
• Permeate O₂ fraction = 0.58/(0.58+0.53) = 52% — so a single stage enriches O₂ only modestly; the *useful* product is the retentate, which becomes N₂-rich. Raising α or using two stages is needed for 99.5% N₂.

Where it is used

• Water and wastewater: municipal seawater/brackish RO (Chennai Nemmeli and Minjur), township and industrial DM plants, MBR-based STPs for reuse in IT parks and hotels, UF-based community plants under Jal Jeevan Mission for fluoride and arsenic belts in Rajasthan, West Bengal and Bihar.
• Textile, tannery and distillery ZLD: UF → RO → NF salt splitting → MEE/ATFD at Tirupur, Pali and Jodhpur CETPs, driven by CPCB/SPCB consent conditions. Roles: process engineer, O&M engineer, membrane application engineer.
• Dairy and food: UF for whey protein concentrate and standardising milk protein (Amul, Mother Dairy, Hatsun), RO for milk and whey pre-concentration, NF for lactose-reduced products, MF for cold sterilisation of milk and clarification of fruit juice, beer and vinegar.
• Pharma and biotech: sterile MF (0.22 µm) filtration, UF/diafiltration for protein buffer exchange and virus removal, RO+EDI for IP Purified Water and WFI loops at Sun Pharma, Cipla, Dr Reddy's, Biocon, Serum Institute.
• Refining, petrochemicals and gas: H₂ recovery from hydrotreater purge, CO₂ removal from natural gas and biogas upgrading to CNG-grade bio-CNG under the SATAT programme, on-site N₂ generators, dehydration of compressed air, and organic-solvent nanofiltration for catalyst recovery.
• Healthcare and electronics: haemodialysis cartridges (a large Indian manufacturing and service business), and ultrapure water (RO + EDI + polishing) for semiconductor and solar-cell fabs.

Interview questions

• Q: What exactly drives water through an RO membrane? A: The net driving pressure — applied pressure minus the osmotic pressure difference across the membrane. If applied pressure only equals the osmotic pressure, flux is zero.
• Q: Why does RO permeate quality worsen as recovery increases? A: Higher recovery concentrates the brine, so the feed-side concentration and hence the salt-flux driving force (c_w − c_p) rise while water flux falls — salt passage goes up on both counts.
• Q: Difference between concentration polarisation and fouling? A: Polarisation is a reversible concentration gradient in the boundary layer that disappears when flow stops, fixed by hydrodynamics; fouling is material deposited or adsorbed on or in the membrane and needs chemical cleaning, sometimes permanently.
• Q: Why is NF chosen over RO for a textile dye effluent? A: NF retains the large dye molecules and divalent salts while letting most NaCl pass, so it runs at lower pressure and energy and yields a reusable salt stream — full desalination is unnecessary.
• Q: Why must chlorine be removed before a polyamide RO membrane? A: Free chlorine oxidises the aromatic polyamide skin, irreversibly destroying rejection; SMBS dosing or activated carbon is used. Cellulose acetate, by contrast, needs a little chlorine to stay biologically clean.
• Q: What is SDI and what value do you design for? A: Silt Density Index from a 15-minute, 0.45 µm filter test, measuring colloidal fouling potential; spiral-wound RO feed should be below 3 (5 maximum).
• Q: Why does pervaporation break an azeotrope when distillation cannot? A: Its selectivity comes from sorption and diffusion in the membrane, not from vapour-liquid equilibrium, so the composition limit imposed by the azeotrope does not apply.
• Q: When would you pick electrodialysis over RO? A: For low-to-moderate TDS feeds, because ED energy is proportional to the salt removed, and when a high water recovery or a high-silica feed makes RO scaling-limited; also where the product need not be nearly salt-free.
• Q: In a membrane nitrogen generator, is nitrogen the permeate or the retentate? A: The retentate — oxygen and water vapour permeate faster through the glassy polymer, so the slow gas, nitrogen, leaves on the high-pressure side.
• Q: Why are spiral-wound modules standard for water but unsuitable for a sludge? A: Their narrow feed spacers give high area per volume but plug easily; fouling or high-solids feeds need tubular or ceramic modules.
• Q: What does an energy recovery device do in seawater RO? A: It transfers the pressure of the high-pressure brine to the incoming feed (isobaric pressure exchanger) or to a turbine, cutting specific energy from roughly 8 to 2.5–4 kWh/m³.
• Q: Flux has dropped 20% in six months but differential pressure is unchanged and salt passage has risen. What do you suspect? A: Scaling or compaction/oxidative damage rather than particulate fouling — plugging would show as rising ΔP. Normalise the data to 25 °C first, then check LSI, antiscalant dosing and chlorine history, and do an acid clean.

Common mistakes

• Using applied pressure instead of net driving pressure (ΔP − Δπ) in the flux equation — the commonest numerical error in exams.
• Assuming salt flux rises with pressure; it does not — J_s depends on concentration difference, which is why rejection improves at higher pressure.
• Quoting MWCO for an RO membrane, or treating a dense RO skin as having pores that sieve.
• Mixing up permeate and retentate as the "product" — in gas separation and in NF salt splitting the valuable stream is often the retentate.
• Comparing fluxes measured at different temperatures without normalising to 25 °C (flux changes ~2.5–3% per °C through viscosity).
• Trying to cure concentration polarisation by chemical cleaning, or trying to cure fouling by raising pressure — raising pressure above the limiting flux just thickens the cake.
• Unit slips: LMH versus m/s, bar versus kPa, mg/L versus kmol/m³, and forgetting the factor i in van't Hoff for an electrolyte.

Related topics

Mass transfer: diffusion · Adsorption and ion exchange · Design of trays and packed columns · Distillation · Fluid mechanics: flow through porous media and pumps · Environmental engineering: effluent treatment and ZLD