An ultrafiltration membrane system separates water from suspended solids, colloids, bacteria and macromolecules using membranes with a pore size of roughly 0.01–0.1 µm, rated by molecular weight cut-off (MWCO). Design centres on selecting a sustainable design flux, sizing the membrane area to the peak flow, and controlling transmembrane pressure through backwash and cleaning cycles.
What does ultrafiltration remove, and how does it compare to MF, NF and RO?
Ultrafiltration (UF) is a pressure-driven membrane process that retains particles by size exclusion. With pores of about 0.01–0.1 µm and a molecular weight cut-off (MWCO) typically between 10 and 300 kDa, UF removes suspended solids, colloids, most bacteria and viruses, and high-molecular-weight organics, while passing dissolved salts and low-MW solutes. It sits between microfiltration (MF) and nanofiltration (NF) on the separation spectrum.
| Process | Nominal pore size | MWCO | Retains | Typical pressure |
|---|---|---|---|---|
| Microfiltration (MF) | 0.1–1 µm | > 500 kDa | Suspended solids, protozoa, some bacteria | 0.1–2 bar |
| Ultrafiltration (UF) | 0.01–0.1 µm | 10–300 kDa | Colloids, bacteria, viruses, macromolecules | 0.5–5 bar |
| Nanofiltration (NF) | ~0.001 µm | 0.2–1 kDa | Divalent ions, small organics, hardness | 3–20 bar |
| Reverse osmosis (RO) | < 0.001 µm (dense) | < 0.1 kDa | Monovalent ions, virtually all solutes | 10–80 bar |
Because UF rejects colloids and microorganisms almost completely but not dissolved salts, it is widely used as pre-treatment ahead of RO and NF membrane trains, as a direct barrier for potable and process water, and as the separation stage in membrane bioreactors. For desalination duties where salt rejection is required, see our companion guide on reverse osmosis system design.
How does flow through a UF membrane actually work?
Convective flow through the near-cylindrical pores of a UF membrane is laminar and can be described by the Hagen–Poiseuille pore-flow model. For an ideal array of straight capillaries, the permeate flux is proportional to the transmembrane pressure and the fourth power of pore radius, and inversely proportional to viscosity and membrane thickness:
where J = permeate flux (m/s), ε = surface porosity, r = pore radius (m), ΔP = transmembrane pressure (Pa), μ = permeate dynamic viscosity (Pa·s), τ = pore tortuosity and δ = membrane active-layer thickness (m). Flux scales with r2 per unit area but with r4 per pore, so pore size dominates permeability.
Two practical consequences follow. First, flux is strongly temperature-dependent because water viscosity falls by roughly 2–3% per °C; permeability is therefore always normalised to 20 °C. Second, small increases in the nominal pore radius sharply raise permeability but coarsen the MWCO, so membrane selection trades throughput against rejection.
What are transmembrane pressure, flux and permeability?
Three interlocking quantities describe UF hydraulics. Flux (J) is the permeate flow rate per unit membrane area, almost universally expressed in litres per m² per hour (LMH). Transmembrane pressure (TMP) is the net pressure driving permeate across the membrane, and permeability (Lp) is flux divided by TMP, in LMH/bar — the key indicator of membrane condition.
where Pfeed, Pconc and Pperm are the feed, concentrate and permeate pressures (bar), J is flux (LMH) and Lp is permeability (LMH/bar). Clean-water permeability for a new UF membrane is typically 200–1,000 LMH/bar at 20 °C; a sustained fall in Lp at constant flux signals fouling.
Systems run in one of two control modes. In constant-flux operation (the norm for hollow-fibre UF) the permeate pump holds J fixed and TMP is allowed to rise as fouling develops; the TMP rise rate is the primary fouling metric. In constant-pressure operation TMP is held and flux declines. Constant flux is preferred because it makes downstream flow predictable and gives an unambiguous cleaning trigger.
How is fouling modelled? The resistance-in-series approach
Real permeate flux is lower than the clean-membrane value because several resistances act in series. The resistance-in-series model expresses flux as TMP divided by the total hydraulic resistance:
where Rm = intrinsic membrane resistance, Rc = reversible cake/gel-layer resistance (removed by backwash), Rf = irreversible fouling resistance (pore blocking and adsorption, removed only by chemical cleaning), all in m−1, μ = viscosity (Pa·s) and ΔP = TMP (Pa).
The value of the model is diagnostic: Rc is recovered by hydraulic backwash, whereas a growing Rf that survives backwash indicates the membrane needs a chemically enhanced backwash (CEB) or a clean-in-place (CIP). Tracking how Rtotal partitions between reversible and irreversible components over a run tells the operator whether the cleaning strategy is keeping pace with fouling.
What is concentration polarisation, the limiting flux and critical flux?
As solvent passes through the membrane, retained solutes accumulate at the surface faster than they diffuse back into the bulk. This concentration polarisation raises the wall concentration and, once it reaches the gel concentration Cg, a gel layer forms and flux stops responding to further increases in TMP. Film theory gives the resulting limiting flux:
where Jlim = limiting (gel-controlled) flux (m/s), k = mass-transfer coefficient (m/s, from the Sherwood correlation k = Sh·D/dh), Cg = gel concentration at the surface and Cb = bulk solute concentration. Raising crossflow velocity increases k and hence Jlim.
Related but distinct is the critical flux: the highest flux at which TMP stays essentially stable with time, i.e. below which irreversible fouling is negligible. Sound UF design sets the sustainable operating flux below the critical flux — commonly at 50–80% of it — so that the TMP rise between cleans is gentle and predictable rather than running into rapid, gel-controlled fouling.
Dead-end or crossflow? Backwash, CEB and CIP
UF is operated in two hydraulic configurations. In dead-end (direct) filtration all feed passes through the membrane and rejected solids build up on the surface; it is energy-efficient and standard for low-turbidity feeds, but requires frequent backwashing. In crossflow filtration feed sweeps tangentially across the membrane, shearing away the cake and sustaining higher flux on concentrated or fouling feeds at the cost of higher pumping energy.
Cleaning is a hierarchy matched to the resistance being removed:
- Backwash (BW) — permeate reversed through the membrane every 20–60 min to lift the reversible cake (recovers Rc).
- Chemically enhanced backwash (CEB) — backwash dosed with hypochlorite, caustic or acid, run several times daily to reverse early irreversible fouling.
- Clean-in-place (CIP) — a full recirculating chemical soak every few weeks to months, restoring permeability to near-clean values.
The fraction of permeate consumed by backwash and CEB determines the net recovery, typically 90–98% for hollow-fibre UF. In wastewater duties, UF membranes packaged inside a bioreactor form the barrier stage of an MBR; the trade-offs against attached-growth processes are covered in our MBR vs MBBR comparison.
Worked example: sizing UF membrane area for a process-water duty
Take a duty that must deliver a net (product) permeate flow of 100 m³/h to downstream RO. The membrane is rated for a sustainable design flux of 60 LMH at 20 °C, and the plant recovery after backwash and CEB is 92%. Size the membrane area and the gross feed.
- Gross permeate required: because 8% of production is used for backwash, the membranes must produce Qgross = 100 / 0.92 = 108.7 m³/h.
- Convert flux to consistent units: 60 LMH = 60 L/m²·h = 0.060 m³/m²·h.
- Membrane area: A = Qgross / J = 108.7 / 0.060 = 1,812 m².
- Module count: with modules of 50 m² each, N = 1,812 / 50 = 36.2, so 37 modules (round up), giving 1,850 m² installed and a small design margin.
- Temperature check: if the coldest feed is 10 °C, viscosity is ~30% higher than at 20 °C, so achievable flux falls to roughly 60 / 1.3 ≈ 46 LMH. Re-sizing at 46 LMH needs 108.7 / 0.046 = 2,363 m² (48 modules) — the winter case governs.
So the membrane area is set not by the nominal 20 °C rating but by the worst-case cold-water flux: about 2,360 m², or 48 modules of 50 m². Always size UF area to the lowest design temperature and the peak instantaneous flow, then confirm the design flux against the critical flux measured on the actual feed.
What are the common UF design and operating mistakes?
- Sizing at 20 °C only. Cold water raises viscosity and cuts flux; size to the minimum feed temperature or the plant under-delivers in winter.
- Operating above the critical flux. Chasing a high flux to save membrane area drives rapid irreversible fouling and frequent CIPs — the whole-life cost rises.
- Weak pre-treatment. UF tolerates turbidity but not oil, grease or oxidants above the membrane's tolerance; screen and condition the feed first.
- Ignoring backwash recovery. Net product flow, not gross flux, sets the plant duty; forgetting the backwash fraction undersizes the feed pumps and pre-treatment.
- No critical-flux testing. Design flux should be verified on the real feed, not taken from a datasheet, because feed chemistry controls the sustainable value.
Specified and operated correctly, a UF stage is one of the most robust barriers in water treatment — a near-absolute barrier to particles and microorganisms and a stable RO pre-treatment.
How to size an ultrafiltration membrane system
- Define the duty and feed. Fix the net product flow, peak factor, feed water quality (turbidity, TSS, organics, oil) and the minimum and maximum feed temperature.
- Select the membrane and MWCO. Choose a UF membrane whose MWCO and pore size (0.01–0.1 µm) meet the rejection target while giving adequate permeability.
- Set a sustainable design flux. Pick a design flux below the critical flux (commonly 50–80% of it), confirmed by bench or pilot testing on the actual feed.
- Correct flux for temperature. Normalise the design flux to the coldest feed temperature using viscosity correction; the winter case usually governs the area.
- Calculate membrane area and modules. Divide the gross permeate flow (net flow / recovery) by the corrected flux to get area, then round up to whole modules.
- Design the cleaning regime. Set backwash interval, CEB frequency and CIP protocol so the TMP rise between cleans is controlled and recovery meets target.
Frequently asked questions
What is the pore size and MWCO of an ultrafiltration membrane?
Ultrafiltration membranes have a nominal pore size of about 0.01–0.1 µm and a molecular weight cut-off (MWCO) typically between 10 and 300 kDa. This retains colloids, bacteria, viruses and macromolecules while passing dissolved salts and small organics, placing UF between microfiltration and nanofiltration on the separation spectrum.
What is a typical design flux for a UF system?
Sustainable design flux for hollow-fibre UF is commonly 40–80 LMH (litres per m² per hour) at 20 °C, chosen below the critical flux for the feed. Cleaner feeds and crossflow operation allow the higher end; fouling waters and cold temperatures require a lower flux and therefore more membrane area.
What is transmembrane pressure (TMP) in ultrafiltration?
Transmembrane pressure is the net pressure driving permeate across the membrane, equal to the mean feed/concentrate pressure minus the permeate pressure, in bar. At constant flux, TMP rises as fouling develops, so the TMP trend is the main fouling indicator. Permeability (flux divided by TMP, in LMH/bar) tracks membrane condition over time.
What is the difference between dead-end and crossflow UF?
In dead-end filtration all feed passes through the membrane and solids accumulate on the surface, which is energy-efficient but needs frequent backwashing — suited to low-turbidity feeds. In crossflow filtration feed sweeps tangentially, shearing away the cake to sustain higher flux on concentrated or fouling feeds, at the cost of higher pumping energy.
How does backwashing restore UF membrane flux?
Backwashing reverses permeate through the membrane to lift the reversible cake layer (the Rc term in the resistance-in-series model), recovering most lost flux. Irreversible fouling that survives backwash is tackled by chemically enhanced backwash (CEB) and periodic clean-in-place (CIP), which restore permeability toward the clean-membrane value.
Why is UF used as pre-treatment before reverse osmosis?
UF removes colloids, suspended solids and microorganisms almost completely, delivering a low, stable silt density index (SDI) to the RO membranes. This protects the downstream RO from particulate and biological fouling, stabilises its flux and extends membrane life, making UF a preferred pre-treatment barrier ahead of RO and NF trains.