Engineering Guide · Vendor-Neutral

UF vs DF: Ultrafiltration vs Diafiltration for Concentration and Buffer Exchange

Published June 30, 2026 · 14 min read
Ultrafiltration vs diafiltration: two operational modes of the same TFF cassette side-by-side UF Ultrafiltration · concentrate retentate shrinks feed cassette recycle loop permeate water + small solutes Volume ↓ · Cprotein ↑ no buffer added Ultrafiltration (UF) Volume reduction · concentration Composition unchanged VS DF Diafiltration · exchange buffer fresh DF buffer retentate constant V same cassette recycle loop permeate old buffer + salts wash out Volume = · Buffer swap N diavolumes of fresh buffer Diafiltration (DF) Constant volume · buffer exchange Composition shifts
Figure 1: Ultrafiltration (left) pulls permeate out of the same TFF cassette without replacing it, so the retentate volume shrinks and the protein concentration rises while composition stays the same. Diafiltration (right) adds fresh buffer to the retentate at exactly the same volumetric rate as the permeate leaves, so the retentate volume and protein concentration stay constant while small solutes wash through the membrane and the buffer composition is swapped over N diavolumes. The hardware is identical; only the buffer-feed pump distinguishes the two modes.
Quick Verdict

UF and DF are two operational modes of the same TFF cassette, not competing technologies. UF concentrates by pulling permeate out without replacement; DF exchanges buffer by adding fresh buffer at the permeate-out rate. The biopharma default for a final-formulation step is the UF→DF→UF (UFDF) sequence: pre-concentrate 5-10x to shrink the diafiltration volume, exchange buffer with 7-10 diavolumes in constant-volume mode, then concentrate to the dose-strength target (typically 50-200 g/L for high-concentration subcutaneous mAb). Skipping the pre-concentration UF wastes diafiltration buffer linearly with starting volume; skipping the final UF leaves you at the wrong protein concentration for the next step.

Key differences at a glance

Side-by-side comparison

FactorUltrafiltration (UF)Diafiltration (DF)
Primary purpose Volume reduction; raise protein concentration Buffer exchange; remove small solutes
Retentate volume during operation Decreases (5-20x reduction typical) Constant (continuous mode)
Buffer fed into retentate? No (only permeate leaves) Yes (matches permeate flow rate)
Protein concentration Rises in inverse proportion to volume Constant (constant-volume DF)
Small-solute composition Unchanged (just concentrated) Exchanged exponentially per diavolume
Buffer consumption None (no DF buffer required) 7-10 retentate volumes (typical UFDF target)
Process time scaling Proportional to permeate volume removed Proportional to diavolumes × retentate volume
Membrane (MWCO) Same cassette as DF (10-300 kDa typical) Same cassette as UF
Typical use position in DSP Pre-column concentration, harvest hold, intermediate concentration Final formulation, post-column buffer swap, denaturant removal

Hardware values reflect typical cassette / hollow-fiber TFF systems from Merck Millipore (Pellicon), Sartorius (Sartocon), Cytiva (Hydrosart), Repligen (KrosFlo), and Pall (Minimate). Your cassette vendor's datasheet takes precedence.

Ultrafiltration in detail

Ultrafiltration is the simpler of the two modes in operational terms: harvest broth, chromatography eluate, or any other dilute protein solution is fed across the surface of a semi-permeable membrane while a transmembrane pressure (TMP) of 0.5-1.5 bar drives water and small solutes through the membrane into the permeate. The protein, retained by the membrane pore size, stays in the retentate. As permeate accumulates and no fresh buffer is added back, the retentate volume falls and the protein concentration rises in inverse proportion. A 10 L feed at 1 g/L concentrated to 1 L sits at 10 g/L; concentrated to 0.5 L it sits at 20 g/L. The mass balance is conserved (no protein is lost through a tight-MWCO membrane), so the only loss is from holdup volume in the cassette, recycle loop, and dead space at the end of the run.

How it works

Cassettes and hollow fibres provide the surface area; the same hardware runs UF or DF depending on whether you connect the diafiltration buffer pump. For a typical CHO mAb feed, a 30 kDa MWCO regenerated-cellulose cassette retains the 150 kDa antibody at sigma < 0.001 while passing water, salts, sugars, and the small free amino acids. The recirculation pump generates the cross-flow that sweeps the membrane surface and keeps the concentration-polarization layer thin; a separate retentate-control valve sets the TMP, which is the pressure differential pushing solvent through the membrane. Three flux regimes apply: linear with TMP at low pressure, the "knee" where flux becomes mass-transfer-limited (usually 0.5-1.5 bar for most proteins), and a polarization-limited regime above the knee where increased pressure only worsens fouling without raising permeate flux. The mid-knee operating point is the standard target. The TFF flux calculation guide covers the sizing math in full.

When UF wins

Pure UF (no diafiltration) is the right answer for any step where you only need to change concentration, not composition. The canonical examples are: pre-concentrating a Protein A eluate from 2-5 g/L up to 30-50 g/L before a polishing ion-exchange column to fit the column load volume; reducing the volume of a clarified harvest from 1,000 L down to 100 L for cold storage hold; concentrating a perfusion bleed stream from 0.5 g/L up to 10 g/L before downstream processing. UF is also the first and last step of every UFDF formulation sequence, where the pre-concentration step shrinks the diafiltration buffer volume and the post-concentration step hits the dose-strength target. The advantage in these contexts is buffer-consumption efficiency: pure UF uses zero diafiltration buffer, only the small volume of buffer you would feed for a final formulation rinse and recovery step.

Diafiltration in detail

Diafiltration is the mode that solves the buffer-exchange problem. Where UF only changes concentration, DF holds protein concentration constant and instead shifts the composition of the small-solute pool surrounding the protein. In constant-volume (continuous) mode, the operator runs the recirculation loop the same as for UF but adds a second pump that feeds fresh diafiltration buffer into the retentate at the same volumetric rate as permeate leaves through the membrane. Volume in equals volume out; retentate volume holds constant; protein concentration holds constant; small solutes wash through the membrane and are gradually replaced by the new buffer composition. The total volume of fresh buffer added, divided by the retentate volume, is the number of diavolumes (N). Each diavolume exchanges roughly 63% of the remaining small-solute concentration for a fully permeable solute, following the exponential C/C0 = exp(-N).

How it works

The mathematical model is straightforward. For a small solute with sieving coefficient sigma (1 means fully permeable, 0 means fully retained), the residual concentration after N diavolumes of constant-volume diafiltration is C/C0 = exp(-N(1-sigma)). For salts, free amino acids, sucrose, and most small-molecule excipients, sigma is close to 1 (small relative to a 30 kDa MWCO membrane), so the simplified C/C0 = exp(-N) holds. Five diavolumes removes 99.3%; seven removes 99.9%; nine removes 99.99%. The Pall Diafiltration Application Note uses the same 5-diavolume / 99% rule. For partially retained excipients (large polymeric stabilizers, charged compounds in high-concentration formulations), the (1-sigma) term shrinks the effective washout rate and requires extra diavolumes. The classic Saksena and Zydney (2004) work on protein-solute interactions documents how solute partitioning into the protein-rich retentate phase changes the apparent sigma at high protein concentrations.

When DF wins

Pure DF (no concentration step) is the right answer when you need to change composition without changing protein concentration. The classic examples are: simple desalting of a 1-5 g/L lab-scale sample where concentration cycling is not desired; removing a denaturant (urea, guanidine HCl) from a refolded protein where holding the protein at low concentration prevents re-aggregation during exchange; washing a shear-sensitive product (live cells, AAV, lentivirus) where the gentle hollow-fibre cross-flow plus constant-volume operation minimises mechanical stress. The combined UF/DF process for oncolytic measles virus described by Loewe et al. (2022) in Membranes is a good case study of UF-then-DF where the DF step is the operationally dominant one because virus stability constrains how aggressively the pre-concentration can be pushed.

Pros and cons

Ultrafiltration (UF)

Advantages

  • Zero diafiltration buffer consumption - no second buffer prep, no second hold tank.
  • Volumetric reduction up to 20-50x in a single step; shrinks downstream hold-tank size by the same factor.
  • Faster than DF for an equivalent endpoint (no buffer-feed flow constraint).
  • Required pre-step for any DF that wants to be buffer-efficient.
  • Hardware-identical to DF; cassette doubles up.

Disadvantages

  • Cannot change buffer composition; impurities are concentrated alongside the product.
  • Concentration polarization grows with falling retentate volume; flux drops as run progresses.
  • Over-concentration risk: aggregation and viscosity climb non-linearly above 100 g/L for many mAbs.
  • Increases the activity of all retained solutes including aggregates, HCP, and salts.
  • Cannot remove denaturants or excipients on its own.

Diafiltration (DF)

Advantages

  • Holds protein concentration constant; product sits at one well-defined operating point.
  • Removes salts, excipients, denaturants, and low-MW HCP exponentially with diavolumes.
  • 50-100x more buffer-efficient than dialysis for the same exchange efficiency.
  • Gentle on shear-sensitive products at constant volume and concentration.
  • Required for every formulation step that targets a defined final buffer.

Disadvantages

  • Buffer consumption scales linearly with retentate volume; expensive if you skip pre-concentration UF.
  • Cannot raise or lower protein concentration on its own.
  • Donnan effect at high protein concentration (above 50 g/L) inflates the diavolume requirement for oppositely-charged excipients.
  • Buffer-feed pump and flow-control loop add a failure mode that UF does not have.
  • Slower per unit of permeate processed (the buffer-feed rate caps throughput).

Which mode should you run, and when?

The choice is usually not UF or DF alone; it is the proportion of UF to DF and the sequence in which they are combined. Pick based on what the upstream context delivers and what the downstream step requires.

Pre-column concentration of a chromatography eluate

You have 200 L of Protein A pool at 5 g/L and you need to load 20 L onto a polishing ion-exchange column. Pure UF: concentrate 10x in the cassette, hold the elution buffer, no DF needed because the next column will exchange buffer itself.

Choose UF only

Buffer swap before storage hold at constant concentration

You need to move a 50 g/L intermediate pool from acetate elution buffer into a histidine storage buffer for a 30-day frozen hold, but the concentration is already where you want it. Pure DF in constant-volume mode at 50 g/L for 7 diavolumes; no UF needed.

Choose DF only

Final formulation of a high-concentration mAb drug substance

You have a 1,000 L polishing pool at 4 g/L and you need to deliver a 100 g/L mAb in histidine-sucrose-polysorbate formulation buffer for a subcutaneous fill. UFDF sequence: UF to ~50 g/L, DF for 8-10 diavolumes in formulation buffer, final UF to 100 g/L with Donnan correction.

Choose UF→DF→UF

Refold pool with urea or guanidine that cannot tolerate concentration

You have a refolded inclusion-body protein in 2 M urea that aggregates above 5 g/L. Pure DF at the refold concentration to wash out the denaturant for 10-12 diavolumes; final UF only after the protein is in benign buffer.

Choose DF then UF

Real-world use cases

Typical mode combinations and the bioprocess context where each has settled in.

CHO mAb, 2000 L final formulation
UFDF to subcutaneous-grade dose strength

UF from 4 g/L to 50 g/L (12.5x volume reduction). Constant-volume DF at 50 g/L for 8-10 diavolumes into histidine-sucrose-polysorbate formulation buffer. Final UF to 140-160 g/L drug substance for a 100 mg/mL fill, with extra diavolumes added if the Donnan effect inflates excipient retention. Cassette: 30 kDa Pellicon or Sartocon.

E. coli refolded protein
DF-dominant denaturant washout

Pure DF at 2-3 g/L for 10-12 diavolumes to wash 2-8 M urea or guanidine out of a refolded inclusion-body pool, holding protein concentration low to prevent re-aggregation. Only after the urea is below 0.1 M does a final UF step concentrate to the polishing column load. Hollow fibre preferred for gentle shear profile.

AAV viral vector
Combined UF/DF for vector concentration and formulation

UF to concentrate dilute clarified harvest 20-30x, then DF for 6-8 diavolumes into final storage buffer (PBS + sucrose or proprietary cryoprotectant). 100-300 kDa MWCO hollow fibre is the standard format because the cassette shear can damage the capsid. Loewe et al. (2022) documents the same UF/DF combination for measles virus, where pre-concentration aggressiveness is bounded by particle stability.

Perfusion-stream UF
UF-only volume reduction at the bleed

Continuous perfusion delivers 1-2 reactor volumes per day of low-titer harvest (0.3-1 g/L). Pure UF on the bleed stream concentrates 10-20x to shrink the hold-tank volume before the capture column. No DF needed because the downstream Protein A or affinity column will exchange buffer itself. Single-pass tangential flow filtration (SPTFF) is the modern variant for continuous integration.

Need to size the UF/DF cassette area and diafiltration buffer volume for your batch?

The Filtration Calculator sizes TFF cassette area, calculates diavolumes for a target washout percentage, and estimates total buffer volume for any UF/DF sequence. Drop in feed volume, target concentration, sieving coefficient, and dose-strength target; it returns sizing, buffer cost, and run-time estimate in seconds.

Open the Filtration Calculator

Cost and lifecycle considerations

UF and DF share hardware - the cost split is buffer and time, not capex

Cassettes, hollow fibres, holders, transmembrane pressure controllers, and recirculation pumps are identical between UF and DF runs. The DF mode adds a buffer-feed pump and a flow-control loop, both of which are part of every modern skid (Sartorius FlexAct, Repligen TangenX, Cytiva AKTA flux, Pall Cadence). Cost differences between pure UF and UF/DF combined sequences are dominated by diafiltration buffer consumption, hold-tank size, and incremental run time.

For a 1,000 L feed at 4 g/L mAb (4 kg of product), a UF-only step that reduces to 50 g/L (80 L final) consumes essentially no diafiltration buffer, takes 4-6 hours, and needs an 80 L collection vessel. A pure DF at the starting 4 g/L for 7 diavolumes consumes 7,000 L of formulation buffer and takes 14-20 hours - a buffer cost of 14,000-35,000 USD at typical 2-5 USD/L formulation buffer pricing (citrate or histidine with excipients).

The UFDF sequence is the cost-efficient compromise: pre-UF from 1,000 L to 80 L cuts the diafiltration retentate volume by 12.5x. The constant-volume DF then needs only 80 L × 7 = 560 L of buffer, a 12.5x cost reduction over the pure-DF case. A final UF to 100 g/L takes another 1-2 hours and produces 40 L of drug substance at the formulation dose strength. The total UFDF buffer cost lands around 1,100-2,800 USD instead of 14,000-35,000 USD, and the total cycle time drops to 8-12 hours.

Cost component (1,000 L feed at 4 g/L mAb) UF-only (no buffer exchange) UFDF (UF→DF→UF)
Diafiltration buffer volume0 L~560 L (80 L retentate × 7 DV)
Diafiltration buffer cost (2-5 USD/L)0 USD~1,100-2,800 USD
Cycle time4-6 h8-12 h
Cassette consumable (single-use)~1.5-2.5 m² cassette, 8-15k USD~1.5-2.5 m² cassette, 8-15k USD
Hold-tank size required~100 L collection vessel~100 L collection vessel
3-year run cost at 20 batches/year (consumables + buffer)~500-900k USD~570-1,070k USD

Pure DF at the starting volume of 1,000 L would consume 7,000 L of formulation buffer per batch (14-35k USD/batch in buffer alone, 280-700k USD/year on buffer). Pre-concentration UF removes that penalty without losing the buffer-exchange benefit.

Vendor landscape

UF and DF are mode choices rather than vendor choices, but the cassette / hollow-fibre membrane format and skid automation are where vendors differentiate. Major TFF system vendors are listed below; each supports both modes from the same hardware.

Cassette / flat-sheet membrane vendors

Hollow-fibre membrane vendors

Automated UFDF skid vendors

Frequently asked questions

What is the difference between UF and DF?
Ultrafiltration (UF) and diafiltration (DF) use the same tangential flow filtration membrane but in two different operational modes. UF reduces the retentate volume by pulling permeate out without replacing it, which concentrates the product. DF holds the retentate volume constant by adding fresh buffer at the same rate the permeate is pulled out, which washes small solutes through the membrane and exchanges the surrounding buffer. UF changes concentration; DF changes composition. Most biopharma TFF steps run UF then DF then UF (UFDF) in sequence to hit both a target concentration and a target buffer in a single unit operation.
When should I use UF vs DF?
Use UF whenever the goal is to reduce volume or raise protein concentration; you have a dilute pool out of a chromatography eluate, a clarified harvest, or a low-titer perfusion stream, and you want to drop the volume 5-20 fold before the next step. Use DF whenever the goal is to change the buffer composition or remove a small-molecule contaminant without changing protein concentration; you need to swap an elution buffer for a formulation buffer, drop conductivity before a polishing column load, or wash out an excipient or denaturant. Almost every final-formulation step uses both: a UF pre-concentration to reduce diafiltration buffer consumption, a DF in constant-volume mode to exchange buffer with 7-10 diavolumes, then a final UF to hit the dose-strength target concentration.
How many diavolumes do I need for buffer exchange?
For a fully permeable small solute (sieving coefficient sigma = 1) in constant-volume diafiltration, the residual concentration follows C/C0 = exp(-N) where N is the number of diavolumes. So 5 diavolumes removes 99.3% of the original buffer, 7 diavolumes removes 99.9%, and 9 diavolumes removes 99.99%. The Pall Diafiltration Application Note gives the same 5-diavolume rule for 99% salt removal. For partially retained excipients, the equation becomes C/C0 = exp(-N(1-sigma)), so a sigma of 0.2 (20% retention) means 9 diavolumes only achieves 95% removal instead of 99.99%. Charged excipients in high-concentration UFDF need extra diavolumes to overcome the Donnan effect. A typical final-formulation UFDF step targets 8-10 diavolumes for clearance of salts and excipients while keeping diafiltration buffer volume reasonable.
Is diafiltration the same as buffer exchange?
Yes - in bioprocessing, diafiltration is the technique used to perform buffer exchange. Diafiltration uses the same UF membrane as the concentration step, but adds fresh diafiltration buffer to the retentate at the same rate as permeate leaves through the membrane. Small solutes (salts, free amino acids, denaturants like urea or guanidine, low-MW excipients) wash through the membrane and are replaced by the new buffer composition, while the retained protein stays in the retentate. Diafiltration is about 50-100x more buffer-efficient than dialysis for the same exchange efficiency; the Pall application note compares 5 diavolumes for 99% exchange via diafiltration against 200-500x sample volume for the same exchange via dialysis.
What is the difference between continuous and discontinuous diafiltration?
Continuous (constant-volume) diafiltration adds fresh diafiltration buffer to the retentate tank at exactly the same volumetric rate as the permeate flows out, so the retentate volume and protein concentration stay constant throughout. Discontinuous (sequential dilution) diafiltration dilutes the retentate with a slug of fresh buffer to a higher volume, then concentrates back down to the starting volume, then repeats. For the same buffer-exchange efficiency, continuous diafiltration uses fewer total diavolumes and is gentler on the product (no concentration cycling). Discontinuous mode is sometimes used at lab scale where flow-rate matching is hard, or in early UF/DF screening where the volume swings simplify sampling, but constant-volume is the GMP biopharma default.
What sequence should I run UF and DF in?
The standard biopharma sequence is UF then DF then UF (UFDF). The first UF step concentrates the dilute feed (typically 5-10x) to reduce the protein-mass-corrected volume that the diafiltration buffer must wash through, cutting buffer consumption by the same factor. The DF step then runs in constant-volume mode at this lower retentate volume, typically for 7-10 diavolumes, exchanging the chromatography elution buffer for the final formulation buffer. The second UF step then concentrates the product to the final dose-strength target (often 50-200 g/L for high-concentration subcutaneous mAb formulations). The choice of how much to pre-concentrate before DF is an economic optimization: pre-concentrate too little and you waste diafiltration buffer, pre-concentrate too much and you risk concentration-polarization fouling that increases process time.
What is the Donnan effect in UF/DF?
The Donnan effect is an electrostatic partitioning that becomes significant at high protein concentrations (above about 50 g/L), where the charged protein retains an unequal distribution of charged excipients across the membrane. Steele and Arias (2014) showed that excipients with the opposite charge to the protein experience attractive Donnan forces and partition into the retentate, requiring additional diavolumes for their removal; excipients with the same charge are repelled and are removed faster than the ideal C/C0 = exp(-N) equation predicts. For high-concentration UFDF (140-160 g/L final mAb formulations), ignoring the Donnan effect typically underestimates the required diavolumes by 20-50% for oppositely-charged excipients. Modeling the effect with a modified sigma that scales with protein concentration improves diavolume prediction accuracy.
Can you run UF without DF or DF without UF?
Yes - both modes can run alone if the upstream context calls for it. Pure UF (no DF) is used in any concentration-only step: pre-concentrating a chromatography eluate before a polishing column, concentrating a clarified harvest before storage hold, or volume reduction of a low-titer perfusion bleed. Pure DF (no UF pre-concentration) is occasionally used at lab scale for simple desalting of a sample that is already at a reasonable concentration, or when concentration cycling would damage a shear-sensitive product. In biopharma manufacturing, however, almost every final-formulation TFF step combines both modes in the UFDF sequence because the diafiltration buffer cost saving from pre-concentration is too large to ignore.

Resources and references