What Is Diafiltration?
Diafiltration is a tangential flow filtration (TFF) technique that washes small molecules out of a protein solution by continuously replacing the original buffer with a new one. It is the standard method for buffer exchange, desalting, and small-molecule impurity removal in biopharmaceutical downstream processing.
During diafiltration, fresh exchange buffer is added to the retentate reservoir while permeate is removed through the membrane. The protein is fully retained by the ultrafiltration membrane (typically 10 to 30 kDa MWCO), while salts, sugars, detergents, and other small molecules pass through into the permeate. The volume of exchange buffer added relative to the retentate volume is measured in diavolumes (DV). One diavolume equals one retentate volume of buffer added and removed.
Understanding how to calculate the number of diavolumes needed for a target clearance is essential for sizing the UF/DF step, estimating buffer consumption, and predicting process time. The calculation is straightforward for freely permeable solutes but requires adjustment when the solute is partially retained by the membrane.
The Exponential Dilution Formula
Impurity concentration drops exponentially with each diavolume when the sieving coefficient is constant. The governing equation for constant volume diafiltration is:
C / C0 = e−σN
Where C = final concentration, C0 = initial concentration, σ = sieving coefficient (0 to 1), N = number of diavolumes
This equation assumes ideal mixing in the retentate reservoir and a constant sieving coefficient throughout the process. For a freely permeable solute (σ = 1), each diavolume reduces the impurity concentration by a factor of e−1 = 0.368, meaning roughly 63.2% is removed per diavolume.
To find how many diavolumes are needed for a target clearance, rearrange the formula:
N = −ln(C / C0) / σ
| Diavolumes (N) | C/C0 | % Removed | Log Reduction |
|---|---|---|---|
| 1 | 0.368 | 63.2% | 0.43 |
| 2 | 0.135 | 86.5% | 0.87 |
| 3 | 0.050 | 95.0% | 1.30 |
| 4 | 0.018 | 98.2% | 1.74 |
| 5 | 0.0067 | 99.3% | 2.17 |
| 6 | 0.0025 | 99.75% | 2.60 |
| 7 | 0.00091 | 99.91% | 3.04 |
| 8 | 0.00034 | 99.97% | 3.47 |
| 10 | 0.000045 | 99.995% | 4.34 |
The exponential relationship means the first few diavolumes achieve the greatest absolute removal. Going from 5 to 7 DV only improves clearance from 99.3% to 99.9%, a marginal gain that consumes 40% more buffer. This is why targeting the minimum number of diavolumes that meets the specification is important for buffer cost and process time.
Sieving Coefficient and Its Effect on Clearance
The sieving coefficient (σ) determines how freely a solute passes through the membrane, and it is the single most important parameter controlling diafiltration volume calculation accuracy. A sieving coefficient of 1 means the solute is completely unretained by the membrane (permeate concentration equals retentate concentration). A coefficient of 0 means the solute is fully retained.
In practice, most species targeted for removal during diafiltration have sieving coefficients between 0.8 and 1.0. Common buffer salts like NaCl, Tris, and phosphate are freely permeable (σ ≈ 1.0) through a 30 kDa membrane. However, some larger excipients, detergents, or partially bound impurities have lower sieving coefficients that require more diavolumes.
| Solute | MW (Da) | Typical σ | DV for 99% Removal |
|---|---|---|---|
| NaCl | 58 | ~1.0 | 4.6 |
| Tris | 121 | ~1.0 | 4.6 |
| Sucrose | 342 | 0.95-1.0 | 4.8-5.0 |
| Polysorbate 80 (monomeric) | 1,310 | 0.7-0.9 | 5.1-6.6 |
| Polysorbate 80 (micellar, above CMC) | ~80,000 | 0.1-0.4 | 11.5-46 |
| PEG 4000 | 4,000 | 0.3-0.6 | 7.7-15.3 |
| Antifoam (silicone) | >10,000 | 0.0-0.2 | >23 |
| Protein-bound impurity | Variable | 0.0-0.5 | >9.2 |
When impurities are partially bound to the retained protein, the effective sieving coefficient is lower than the free solute would suggest. Shao and Zydney (2004) demonstrated that even weak binding (Kd in the mM range) can dramatically increase the required diavolumes. If you observe that impurity clearance plateaus before reaching the target, suspect protein binding and consider adjusting pH or ionic strength during DF to reduce the association.
Constant Volume vs Variable Volume Diafiltration
Constant volume diafiltration (CVD) is the most common mode in biopharmaceutical manufacturing because it provides predictable, exponential impurity removal with simple process control. However, variable volume diafiltration (VVD) offers advantages in certain situations and is worth understanding for process optimization.
Constant volume diafiltration (CVD) adds exchange buffer at the same rate as permeate removal. Retentate volume stays fixed, protein concentration stays constant, and impurity clearance follows the exponential equation directly. This mode is easy to control (buffer addition is linked to a load cell or flow meter) and consumes the minimum buffer for a given number of diavolumes at a fixed retentate concentration.
Variable volume diafiltration (VVD), also called discontinuous diafiltration, alternates between concentration and dilution steps. The retentate is first concentrated (removing permeate without adding buffer), then diluted back to the original volume with exchange buffer. This cycle is repeated until the target clearance is reached.
| Parameter | CVD (Constant Volume) | VVD (Variable Volume) |
|---|---|---|
| Buffer efficiency | Higher (minimum buffer per log reduction) | Lower (20-40% more buffer for same clearance) |
| Average flux | Constant (set by protein concentration) | Higher average (dilution steps restore flux) |
| Process control | Simpler (one flow balance) | More complex (alternating modes) |
| Membrane area needed | More at high concentrations | Less (higher average flux) |
| Fouling behavior | Steady-state fouling | Dilution can partially reverse fouling |
| Typical use | Standard mAb UF/DF | High-concentration formulation, fouling-prone feeds |
For VVD, the impurity removal per cycle depends on the concentration factor X (ratio of initial to concentrated volume). Each cycle removes a fraction (1 − 1/X) of the impurity present at the start of that cycle. After n cycles with concentration factor X, the residual fraction is (1/X)n. Converting to an equivalent CVD diavolume count: Nequivalent = n × ln(X). For example, 5 cycles with X = 3 is equivalent to N = 5 × ln(3) = 5.5 diavolumes, but consumes 5 × (3 − 1) = 10 retentate volumes of buffer (compared to 5.5 volumes for CVD), making it roughly 82% less efficient on buffer.
Filtration & TFF Calculator
Size your TFF membrane, estimate flux, and calculate process time for UF/DF operations.
Finding the Optimal Concentration for DF
The optimal protein concentration for diafiltration minimizes total process time by balancing two competing effects: higher protein concentration reduces the retentate volume (and therefore the total buffer volume) but also reduces membrane flux due to concentration polarization and osmotic pressure.
The key metric is the product of concentration and flux: C × J(C). At low concentrations, flux is high but the retentate volume is large, so total filtration volume is large. At high concentrations, the retentate volume is small but flux drops sharply. The optimum sits where C × J is at its maximum.
For the gel polarization model, where flux follows J = k × ln(Cgel/C), the optimum concentration can be found analytically:
Copt = Cgel / e ≈ 0.37 × Cgel
Where Cgel is the gel concentration (typically 200-350 g/L for mAbs) and e is Euler's number
For a typical mAb with Cgel around 250 g/L, the optimal diafiltration concentration is approximately 0.37 × 250 = 92 g/L. In practice, most processes diafilter at somewhat lower concentrations (20 to 50 g/L) to maintain a safety margin against gel formation, reduce viscosity-related pump issues, and avoid protein aggregation at high concentrations.
Key practical considerations when choosing the diafiltration concentration:
- Viscosity: Solutions above 80 to 100 g/L become highly viscous (>10 cP), increasing pump requirements and pressure drop across the membrane module.
- Aggregation risk: High protein concentrations increase protein-protein interactions and shear-induced aggregation during TFF recirculation.
- Excipient sieving: Some excipients (e.g., polysorbate, sucrose) exhibit concentration-dependent sieving that changes as protein concentration increases.
- Gel layer: Operating too close to Cgel risks irreversible membrane fouling that is difficult to clean even with NaOH washes.
Worked Example: mAb Buffer Exchange
This worked example demonstrates a complete UF/DF calculation for a monoclonal antibody process switching from protein A elution buffer to the final formulation buffer.
Worked Example: mAb UF/DF Buffer Exchange
Starting conditions:
- Volume after Protein A elution: 20 L
- Protein concentration: 5 g/L (total protein: 100 g)
- Current buffer: 50 mM sodium acetate, 150 mM NaCl, pH 3.5
- Target buffer: 25 mM histidine, 150 mM NaCl, pH 6.0
- Target residual acetate: <0.5 mM (from 50 mM, so 99% removal needed)
- Membrane: 30 kDa Pellicon 3 cassette, 0.11 m2
- Flux at 5 g/L: 80 LMH. Flux at 25 g/L: 45 LMH.
Step 1: Determine concentration factor before DF.
Target DF concentration: 25 g/L (practical optimum for this system)
Concentration factor: 25 / 5 = 5×
Retentate volume after UF: 20 / 5 = 4 L
Protein mass check: 4 L × 25 g/L = 100 g ✓
Step 2: Calculate diavolumes needed.
Target: reduce acetate from 50 mM to <0.5 mM
C/C0 = 0.5 / 50 = 0.01
σ for sodium acetate (MW 82) through 30 kDa = 1.0
N = −ln(0.01) / 1.0 = 4.605 / 1.0 = 4.6 DV
Round up to 5 DV for safety margin
Step 3: Calculate total buffer volume.
Buffer volume = N × VR = 5 × 4 L = 20 L of histidine buffer
Without pre-concentration: 5 × 20 L = 100 L
Buffer savings from 5× concentration: 80 L (80% reduction)
Step 4: Estimate process time.
UF step (20 L → 4 L): permeate = 16 L
Average UF flux (5→25 g/L): ~60 LMH
UF time: 16 L / (60 LMH × 0.11 m2) = 2.4 h
DF step: permeate = 20 L (5 DV × 4 L)
DF flux at 25 g/L: 45 LMH
DF time: 20 L / (45 LMH × 0.11 m2) = 4.0 h
Total UF/DF time: 2.4 + 4.0 = 6.4 h
Step 5: Verify final buffer composition.
Residual acetate: 50 × e−1.0 × 5 = 50 × 0.0067 = 0.34 mM ✓ (<0.5 mM)
Final volume: 4 L at 25 g/L in histidine buffer
Product recovery (typical for UF/DF): 95-98%
Buffer Preparation Calculator
Calculate buffer recipes, pH adjustments, and dilution volumes for your downstream buffers.
Scale-Up Considerations for DF
Diafiltration volume calculation scales linearly with retentate volume, but several practical factors change as you move from development to manufacturing scale. The number of diavolumes stays the same (it is a dimensionless ratio), but membrane area, pump capacity, and buffer preparation volumes all increase proportionally.
The primary scale-up principle for TFF is to maintain constant membrane loading (L/m2). If your development process uses 4 L retentate on 0.11 m2 (36 L/m2 loading), a 400 L manufacturing-scale retentate needs 11 m2 of membrane area to maintain the same loading and equivalent flux.
| Parameter | Development (0.11 m2) | Pilot (1.1 m2) | Manufacturing (11 m2) |
|---|---|---|---|
| Retentate volume | 4 L | 40 L | 400 L |
| Membrane loading | 36 L/m2 | 36 L/m2 | 36 L/m2 |
| Diavolumes (N) | 5 | 5 | 5 |
| Buffer volume | 20 L | 200 L | 2,000 L |
| DF flux | 45 LMH | 45 LMH | 45 LMH |
| DF process time | 4.0 h | 4.0 h | 4.0 h |
| Cross-flow rate | 0.7 L/min | 7 L/min | 70 L/min |
Key factors that can differ between scales:
- Mixing time: Larger retentate tanks take longer to achieve uniform composition. Ensure mixing is adequate so that the well-mixed assumption of the exponential model holds.
- Holdup volume: Tubing, pump heads, and membrane channels contain product that is not in the retentate tank. At manufacturing scale, holdup can be 5 to 15% of the retentate volume, reducing effective clearance. Add 0.5 to 1 extra diavolume to compensate.
- Buffer preparation: 2,000 L of formulation buffer requires dedicated vessels, WFI, and QC release testing. Factor buffer prep lead time into the batch schedule.
- Integrity testing: Pre-use and post-use membrane integrity testing (e.g., air diffusion test) is required at GMP scale and adds 30 to 60 minutes per test.
Chromatography Calculator
Calculate column dimensions, resin volumes, and buffer requirements for your purification train.
Frequently Asked Questions
How many diafiltration volumes are needed for complete buffer exchange?
For a freely permeable solute (sieving coefficient = 1), 5 diavolumes remove 99.3% of the original buffer and 7 diavolumes remove 99.9%. Most industrial processes use 6 to 8 diavolumes to achieve greater than 99.5% exchange. Partially retained solutes require proportionally more diavolumes based on their sieving coefficient.
What is the difference between constant volume and variable volume diafiltration?
Constant volume diafiltration (CVD) adds fresh buffer at the same rate as permeate removal, keeping retentate volume fixed. Variable volume diafiltration (VVD) alternates concentration and dilution cycles. CVD is more buffer-efficient per unit of impurity removed but VVD can achieve higher instantaneous flux because concentration is periodically lowered.
What is a sieving coefficient and how does it affect diafiltration?
The sieving coefficient (σ) is the ratio of solute concentration in the permeate to that in the retentate, ranging from 0 (fully retained) to 1 (freely permeable). Higher sieving coefficients mean faster impurity removal. For a solute with σ = 0.5, you need approximately twice as many diavolumes as for a fully permeable solute to reach the same clearance.
Should I concentrate before or after diafiltration?
Concentrating before diafiltration reduces the absolute volume of buffer needed because the retentate volume is smaller. A 5-fold concentration before DF cuts buffer consumption by approximately 5-fold. However, concentrating too much reduces membrane flux. The optimum is typically where the product of concentration and flux (C × J) is maximized, often around 20 to 50 g/L for monoclonal antibodies.
How do I calculate the total buffer volume needed for diafiltration?
Total buffer volume equals the number of diavolumes multiplied by the retentate volume. For example, 7 diavolumes on a 10 L retentate requires 70 L of exchange buffer. The number of diavolumes is calculated from N = −ln(C/C0) / σ, where C/C0 is the target fractional residual concentration and σ is the sieving coefficient of the solute being removed.
What is the optimal protein concentration for diafiltration?
The optimal concentration minimizes total process time by balancing reduced buffer volume against lower membrane flux at higher concentrations. Plot flux versus log concentration and find the point where the product C × J is maximized. For typical mAb processes this optimum falls between 20 and 50 g/L, corresponding to an initial 5 to 10-fold concentration before DF.
Related Tools
- Filtration & TFF Calculator — Size membranes, estimate flux and pressure, and calculate filtration process time.
- Buffer Preparation Calculator — Calculate buffer recipes, pH adjustment, and dilution volumes for downstream processing.
- Chromatography Calculator — Design column dimensions, estimate resin volumes, and plan buffer usage for purification steps.
References
- Shao J. & Zydney A.L. (2004). Optimization of ultrafiltration/diafiltration processes for partially bound impurities. Biotechnology and Bioengineering, 87(3), 286-292. doi:10.1002/bit.20113
- van Reis R. & Zydney A. (2007). Bioprocess membrane technology. Journal of Membrane Science, 297(1-2), 16-50. doi:10.1016/j.memsci.2007.02.045
- Kovács Z., Discacciati M. & Samhaber W. (2009). Modeling of batch and semi-batch membrane filtration processes. Journal of Membrane Science, 327(1-2), 164-173. doi:10.1016/j.memsci.2008.11.024
- Baek Y., Singh N., Arunkumar A., Borys M., Li Z.J. & Zydney A.L. (2017). Ultrafiltration behavior of monoclonal antibodies and Fc-fusion proteins: effects of physical properties. Biotechnology and Bioengineering, 114(9), 2057-2065. doi:10.1002/bit.26326
- Nambiar A.M.K., Li Y. & Zydney A.L. (2018). Countercurrent staged diafiltration for formulation of high value proteins. Biotechnology and Bioengineering, 115(1), 139-144. doi:10.1002/bit.26441