Written and maintained by CASRAI Editorial Board
Last updated
Short answer
Tangential flow filtration (TFF), also called cross-flow filtration, pumps a liquid across the surface of a membrane instead of straight into it. Part of the liquid passes through the membrane (permeate); the rest, with the molecules or particles the membrane holds back (retentate), keeps circulating. Because the flow runs along the surface, filter cake formation is avoided or reduced compared with dead-end filtration.
Definition and scope
What is tangential flow filtration?
TFF is a membrane operation in which fluid flows tangentially across the membrane surface. Busatto et al. (2018) contrast it with conventional dead-end filtration, where larger particles usually clog the pores and the separation drifts away from the membrane’s original pore size; with tangential flow, filter cake formation is avoided. Fernandez-Cerezo et al. (2019) use the same two terms: tangential (cross-flow) mode to reduce fouling of the membrane surface, and normal-flow (dead-end) mode, often used to remove trace particulates.
The same paper separates two membrane classes: membranes designed to retain macromolecules are generally called ultrafiltration, and membranes designed to retain particulates are called microfiltration. In bioprocessing, TFF steps appear for concentration and for diafiltration, either to prepare a feed for chromatography or for final formulation. In the lab, TFF sits alongside other separation and sample-preparation steps such as 0.22 micron sterilizing filtration and FPLC chromatography.
How TFF works in practice
The principle: pressure, flux and fouling
Loewe et al. (2019) describe membrane filtration as a pressure-driven process that generally follows Darcy’s law: flux depends on transmembrane pressure (TMP), fluid viscosity and total membrane resistance. They give TMP as the mean of the feed and retentate pressures minus the permeate pressure: TMP = (pfeed + pretentate) / 2 − ppermeate. Total resistance is the membrane’s own resistance plus a fouling resistance, for example from protein deposition. The same paper lists four dominant fouling mechanisms: cake filtration, intermediate blocking, pore blocking and complete blocking.
Concentration and diafiltration
Concentration mode removes permeate so the retentate volume falls. In diafiltration, Fernandez-Cerezo et al. fed diafiltration buffer to hold the feed volume constant, for a total of seven diafiltration volumes. Busatto et al. concentrated a sample to about 50 mL and then diafiltrated it six times into a sucrose buffer, finishing at 6–9 mL.
Types and related concepts
Membrane formats and materials
Fernandez-Cerezo et al. list three TFF membrane geometries: hollow fibre, spiral wound and flat-sheet. They state flat-sheet is the form typically found in bioprocessing for therapeutics. Flat sheets are commonly mounted in cassettes: compact units of several flat sheets layered with flow-channel spacers, which often include turbulence promoters to reduce fouling. Loewe et al. note that organic TFF membranes are typically made of cross-linked regenerated cellulose, polysulfone or polyether sulfone (PES), and that other materials such as polyvinylidene fluoride may also be suitable. Fernandez-Cerezo et al. list batch, fed-batch, single-pass and feed-and-bleed (continuous) as operating configurations.
Where it matters for labs and supply decisions
Three published TFF set-ups compared
Each row below comes from one open-access paper. The brand names are the equipment those authors used, listed so the identifiers are visible; they are not recommendations.
| Study | Feed | Format and material | MWCO or pore size | Scale and mode | Reported outcome |
|---|---|---|---|---|---|
| Busatto et al., Cells 2018 | Extracellular vesicles in cell culture media (0.2 L) and lipoaspirate-derived fluid (1 L) | Hollow fibre, polyethersulfone (Spectrum Labs KrosFlo Research 2i system) | 0.65 µm, then 500 kD | Input flow 80 mL/min to keep shear below 2000 s−1; concentrated to ~50 mL, six diafiltrations, final 6–9 mL | About 1 h for 200 mL of media; roughly five-fold higher vesicle yield from 100 mL than ultracentrifugation |
| Loewe et al., Membranes 2019 | Oncolytic measles virus in serum-containing and serum-free medium | Flat-sheet, 0.02 m2 (Sartorius Stedim Slice200): PES and cross-linked regenerated cellulose | PES 100 nm pore, 300 kDa, 100 kDa; cellulose 100 kDa | Concentration mode | Infectious virus retained with MWCO ≤ 300 kDa; impurities best depleted with MWCO ≥ 300 kDa; concentration factor at least 15 for three of four membranes (7.3 for the 100 nm PES) |
| Fernandez-Cerezo et al., Biotechnol Bioeng 2019 | Monoclonal antibody (~150 kDa), 12 g/L | Flat-sheet cassette, composite regenerated cellulose, 0.11 m2 (Pellicon 3 C-screen) | 30 kDa | 0.89 L feed, TMP 1.0 bar, seven diafiltration volumes | Protein concentration in solution fell 6–19% after diafiltration at pilot scale |
What a TFF methods section reports
Across these three papers, a TFF run is described by the same set of parameters. Use this as a checklist when reading or writing a methods paragraph:
- Membrane format (hollow fibre or flat-sheet cassette), material and membrane area.
- MWCO or pore size, and what each stage was meant to remove or retain.
- Feed composition and volume, including whether serum proteins were present.
- Flow rate or shear, and transmembrane pressure.
- Mode: concentration only, or diafiltration with the number of diafiltration volumes.
- How recovery was measured (for example infectious titre, protein concentration or particle counts).
Loewe et al. found that fouling severity depended on membrane material and that serum proteins reduced flux, so the material, MWCO and feed belong in any record of a run. For a related infectious-particle readout, see viral titer: what PFU, TCID50, FFU and genome copies measure; for the buffer a run is exchanged into, see what a biological buffer is.
Common misunderstandings
- “TFF is just fast filtration.” It is a different flow arrangement: fluid moves across the membrane surface rather than straight into it, which is the feature the papers credit with reducing fouling and filter cake formation.
- “A lower MWCO is always safer.” In the measles virus study, infectious virus was retained at ≤ 300 kDa, but impurities such as proteins and host-cell DNA were best depleted at ≥ 300 kDa, so recovery and clearance pulled in different directions.
- “Diafiltration loses nothing.” In the antibody study, protein concentration in solution fell by 6–19% after diafiltration at pilot scale.
- “Two membranes with the same MWCO behave alike.” The measles virus paper reports that fouling depended on the membrane material, with the cellulose-based membrane least susceptible.
- “TFF means one membrane format.” The three papers used hollow fibre, a flat-sheet membrane and a flat-sheet cassette.
FAQ
What is tangential flow filtration?
Tangential flow filtration (TFF), also called cross-flow filtration, is a membrane operation in which fluid flows across the membrane surface rather than straight into it. Busatto et al. (2018) note that this avoids filter cake formation, which is how it differs from conventional dead-end filtration.
What is the principle of tangential flow filtration?
Filtration is pressure-driven: flux depends on transmembrane pressure, fluid viscosity and total membrane resistance (Loewe et al., 2019). The part of the liquid that crosses the membrane is the permeate; the part held back is the retentate, which keeps circulating.
What is diafiltration in TFF?
In diafiltration, buffer is fed in while permeate is removed so that the feed volume stays constant. Fernandez-Cerezo et al. (2019) used a total of seven diafiltration volumes; Busatto et al. (2018) diafiltrated six times into a sucrose buffer.
What membrane formats and materials do TFF systems use?
Hollow fibre, spiral wound and flat-sheet geometries are described, with flat-sheet cassettes the form typically found in bioprocessing for therapeutics (Fernandez-Cerezo et al., 2019). Organic membranes are typically cross-linked regenerated cellulose, polysulfone or polyether sulfone (Loewe et al., 2019).
Does the MWCO change what is recovered?
In the measles virus study, infectious virus was retained with MWCO of 300 kDa or lower, while proteins and host-cell DNA were best depleted at 300 kDa or higher (Loewe et al., 2019).
Sources
- Busatto S, Vilanilam G, Ticer T, et al. Tangential Flow Filtration for Highly Efficient Concentration of Extracellular Vesicles from Large Volumes of Fluid. Cells 2018;7(12):273. doi:10.3390/cells7120273 (open access, PMC6315734). Checked 2026-11-07.
- Loewe D, Grein TA, Dieken H, Weidner T, Salzig D, Czermak P. Tangential Flow Filtration for the Concentration of Oncolytic Measles Virus: The Influence of Filter Properties and the Cell Culture Medium. Membranes 2019;9(12):160. doi:10.3390/membranes9120160 (open access, PMC6950090). Checked 2026-11-07.
- Fernandez-Cerezo L, Rayat ACME, Chatel A, Pollard JM, Lye GJ, Hoare M. An ultra scale-down method to investigate monoclonal antibody processing during tangential flow filtration using ultrafiltration membranes. Biotechnology and Bioengineering 2019;116:581–590. doi:10.1002/bit.26859 (open access, PMC6492246). Checked 2026-11-07.
