Technical publication · Downstream processing

Diafiltration volume calculation for TFF buffer exchange

Abstract. This method note derives the constant-volume diafiltration calculation used in tangential flow filtration (TFF) and ultrafiltration/diafiltration (UF/DF). It connects impurity clearance, diavolumes, buffer consumption, product sieving loss, membrane area and process time without hiding the assumptions. A synthetic, downloadable dataset makes the screening calculation reproducible.

Acatian EngineeringPublished and reviewed 12 September 202623 min technical publication

Synthetic constant-volume diafiltration model. At S_i = 0.95, reducing an impurity from 120 to 1.0 mmol L⁻¹ requires 5.04 diavolumes; no customer or experimental Acatian data are shown.

Scope and model boundary

Calculate diafiltration demand from a component balance.

Technical question

For a TFF or UF/DF step, how much diafiltration buffer is needed to reduce a permeating impurity from C_i,0 to C_i,target, what product is lost by sieving, and what membrane area meets the process-time limit? The calculation is a screening mass balance, not a substitute for membrane characterization.

Constant-volume mode

Buffer enters the well-mixed retentate at the same volumetric rate that permeate leaves: Q_B = Q_P. The retained volume V_DF is therefore constant while permeable solutes wash out. Model-based optimal-control analyses show how feed policy can instead be optimized when flux and rejection vary with concentration [1] [2].

Evidence classes

Measure or qualify permeate flux, component sieving, product retention, feed and retentate composition, viscosity, membrane chemistry, cassette, crossflow, transmembrane pressure (TMP), temperature and hold-up. Keep target specifications and maximum processing time separate from experimental observations and calculated results.

Nomenclature

Use explicit volume, concentration and flux bases

On small screens, swipe the table horizontally to compare every column.

SymbolDefinitionUnit or basis

S_iObserved sieving coefficient of component i, C_P,i/C_R,i; rejection R_i = 1 − S_iDimensionless

N_DDiavolumes processed, V_B/V_DF = ∫Q_P(t)dt/V_DFDimensionless

V_DFConstant retentate volume during diafiltrationL

JPermeate flux, Q_P/AL m⁻² h⁻¹ (LMH)

TMPMean transmembrane pressure, commonly (P_in + P_out)/2 − P_P; state the skid conventionbar or Pa

AInstalled effective membrane aream²

Y_pRetained product mass fraction over the stated stepDimensionless or %

02 · Interactive engineering calculator

Calculate diavolumes, buffer volume, product retention and membrane area

Screening model · constant volume · ideal mixing · constant observed sieving and phase-specific flux

Process volumes Initial feed volumeL Diafiltration retentate volumeL Maximum processing timeh

Exchange target Initial impurity concentrationrelative Target impurity concentrationsame unit Impurity in incoming buffersame unit Impurity sieving coefficient, S_i0–1

Membrane performance Product S during concentration0–1 Product S during diafiltration0–1 Concentration-step fluxLMH Diafiltration-step fluxLMH Selected membrane aream²

Download clearance curves Download design basis

Required wash5.04 DVN_D from the component target Buffer volume504 LBefore line, flush and overage demand Total permeate904 LConcentration plus diafiltration DF product retention99.25%Membrane sieving loss only Combined product yield99.01%Concentration and DF sieving only Minimum membrane area6.03 m²At phase-specific flux and maximum time Time at selected area5.56 hExcludes non-processing allowances

General component balanceV_DFdC_i/dt = Q_P(C_B,i − S_iC_i)

The incoming buffer may itself contain component i at C_B,i. With constant V_DF, C_B,i and S_i, C_i = C_B,i/S_i + (C_i,0 − C_B,i/S_i)exp(−S_iN_D). The familiar C_i/C_i,0 = exp(−S_iN_D) is only the zero-impurity-buffer case C_B,i = 0.

Calculation workflow

Size diavolumes, then expose the assumptions that move them.

1. Define the specification

State a component concentration or mass target and where it is sampled. Do not insert pH, conductivity or osmolality directly as C_i in the exponential equation: model individual species, electroneutrality and equilibria, or validate an empirical surrogate. At high protein concentration, retentate composition and pH need not equal the diafiltration buffer because electrostatic partitioning and exclusion affect the final state [3] [4].

2. Measure component sieving

Calculate S_i = C_P,i/C_R,i from paired, time-aligned samples at relevant conditions. Do not assume S = 1 for every small solute or S = 0 for the product. Experiments with bovine serum albumin show that pH, ionic strength, protein conformation and electrostatic membrane interactions can change observed protein sieving [5].

3. Solve and challenge

For C_B,i = 0, calculate N_D = ln(C_i,0/C_i,target)/S_i and V_buffer = N_DV_DF. For nonzero C_B,i, solve the general expression and confirm the target lies above its asymptote C_B,i/S_i. Design to the largest N_D,i across all component constraints.

04 · Synthetic worked exampleN_D = ln(120/1.0)/0.95 = 5.04 DV
V_buffer = 5.04 × 100 L = 504 L

A 500 L feed is first concentrated to V_DF = 100 L. The impurity starts diafiltration at 120 mmol L⁻¹, its target is 1.0 mmol L⁻¹, and its assumed screening sieving coefficient is 0.95. Download the clearance curves and complete design basis to reproduce every displayed result and substitute measured coefficients.

Worked results

Track buffer, product and time on one basis

On small screens, swipe the table horizontally to compare every column.

QuantityWorked calculationEvidence class

Concentration factorVCF = V₀/V_DF = 500/100 = 5.0×Synthetic process basis

Required diavolumesln(120/1.0)/0.95 = 5.039 ≈ 5.04 DVCalculated screening result

Buffer volume5.039 × 100 = 503.9 ≈ 504 LCalculated screening result; excludes line and flush demand

DF product retentionexp(−S_p,DFN_D) = exp(−0.0015 × 5.039) = 99.25%Synthetic S_p,DF assumption; replace experimentally

Concentration-step retentionVCF^−S_p,UF = (100/500)^0.0015 = 99.76%Ideal constant-sieving calculation

Combined product yieldVCF^−S_p,UFexp(−S_p,DFN_D) = 99.01%Calculated membrane-sieving yield only

Low · nominal · high screen

The sieving assumption controls buffer demand

On small screens, swipe the table horizontally to compare every column.

Impurity S_iRequired diavolumesBuffer at V_DF = 100 L

1.004.787 DV478.7 L

0.95 · nominal5.039 DV503.9 L

0.805.984 DV598.4 L

For several clearance constraints, calculate each N_D,i on its own measured sieving basis and design to N_D,design = max(N_D,i). A single conductivity endpoint does not prove every component has met specification.

Membrane area and process time

Use a qualified design flux, not a brochure maximum.

Total permeate volume

The concentration step removes 500 − 100 = 400 L. Diafiltration removes another 504 L while buffer addition holds retentate volume constant. Total permeate volume is therefore 904 L. Add measured skid, line, sample, pre-use, flush and recovery volumes separately.

Phase-specific area

At J_UF = J_DF = 25 LMH and a six-hour limit, A_min = [(400/25) + (503.9/25)]/6 = 6.03 m². Selecting 6.5 m² gives 2.46 h for concentration plus 3.10 h for diafiltration: 5.56 h before non-processing allowances.

Scale-up evidence

Flux can change with TMP, crossflow, concentration, viscosity, polarization, fouling and cassette architecture. A 2026 vendor-specific AAV8 study increased feed volume ten-fold and membrane area 15.6-fold across different cassette and system formats; it supports process-specific scale-up evidence, not a universal linear rule [6]. Qualified high-throughput studies can bracket TMP, crossflow and diafiltration concentration before confirmation at pilot and manufacturing scale [7].

05B

Diavolumes and operating modes

One diavolume has a precise meaning only inside the stated model.

Buffer-exchange efficiency

For a fully permeating solute, zero-solute buffer and S = 1, one diavolume leaves e⁻¹ = 36.8% of the starting concentration: 63.2% removal. The ideal diavolumes for 90%, 95%, 99% and 99.9% removal are 2.303, 2.996, 4.605 and 6.908. At S < 1, divide each value by S.

Continuous versus discontinuous DF

Constant-volume diafiltration adds buffer continuously as permeate leaves. Discontinuous diafiltration alternates dilution and reconcentration, so its time, fouling response and buffer demand require a cyclewise balance. An experimental measles-virus study compared both modes and found process-specific differences in permeate flow and product recovery [8].

Single-pass and continuous TFF

Single-pass diafiltration distributes exchange across staged membrane geometry rather than a recirculating tank, so the batch exponential cannot be copied unchanged. A 2026 dual-membrane study tested continuous concentration and diafiltration with BSA and antibodies and reported strong dependence on feed concentration, flow and diavolumes [9].

Phase-specific capacity screenA_min = [(V₀ − V_DF)/J_UF + N_DV_DF/J_DF]/t_max

Equivalently, t = (V₀ − V_DF)/(J_UFA) + N_DV_DF/(J_DFA). The steady-flux equation is an initial screen; a decision model should integrate measured flux over concentration, TMP, crossflow, temperature and time and include setup, equilibration, sampling, recovery, cleaning and integrity testing.

Limitations

What the ideal exponential model does not prove

On small screens, swipe the table horizontally to compare every column.

Failure modeBias or riskRequired control

S_i assumed constantChanging ionic strength or concentration moves impurity clearance.Measure paired permeate/retentate samples across the run.

S_p constant or underestimatedSmall fractional passage becomes material over concentration and many diavolumes.Estimate S_p,UF and S_p,DF separately and close the product mass balance.

Perfect mixing assumedSkid and tank hold-up delay washout relative to the ideal model.Characterize recirculation, heel, dead volume and sampling location.

Flux held constantArea is undersized when concentration polarization or fouling lowers flux.Use representative, phase-specific flux-decay and pressure-limit data.

Buffer equals product stateFinal pH or excipient concentration misses the formulation target.Use high-concentration composition data or a validated mechanistic model.

Nominal area usedActual cassette area, manufacturing tolerance or unavailable size changes time.Select available hardware and recalculate the complete cycle.

Practical UF/DF checklist

Minimum record for a reviewable TFF design

Material and target

Feed and diafiltration volumes; concentration factor; component concentrations; product titer; target pH, conductivity, osmolality and excipients; temperature; viscosity; product and impurity assays; allowable yield loss.

Membrane operation

Membrane chemistry and MWCO; cassette and lot; effective area; feed flow; crossflow; inlet, retentate and permeate pressure; TMP definition; flux history; load per area; concentration polarization; fouling and recovery.

Model and decision

Mass-balance equations; S_i and S_p evidence; variability; hold-up and heel; buffer overage; process-time allowance; pump and shear limits; sampling plan; scale-up rule; version, reviewer and acceptance criteria.

Primary technical sources

References

Paulen, R.; Fikar, M.; Foley, G.; Kovács, Z.; Czermak, P. “Optimal feeding strategy of diafiltration buffer in batch membrane processes.” Journal of Membrane Science 411–412, 160–172 (2012). doi:10.1016/j.memsci.2012.04.028.
Fikar, M.; Kovács, Z.; Czermak, P. “Dynamic optimization of batch diafiltration processes.” Journal of Membrane Science 355(1–2), 168–174 (2010). doi:10.1016/j.memsci.2010.03.019.
Ambrožič, R.; Arzenšek, D.; Podgornik, A. “Designing scalable ultrafiltration/diafiltration process of monoclonal antibodies via mathematical modeling by coupling mass balances and Poisson–Boltzmann equation.” Biotechnology and Bioengineering 118, 633–646 (2021). doi:10.1002/bit.27598.
Teeters, M.; Bezila, D.; Benner, T.; Alfonso, P.; Alred, P. “Predicting diafiltration solution compositions for final ultrafiltration/diafiltration steps of monoclonal antibodies.” Biotechnology and Bioengineering 108(6), 1338–1346 (2011). doi:10.1002/bit.23067.
Mohammadzadehmarandi, A.; Zydney, A. L. “Buffer effects on protein sieving losses in ultrafiltration and their relationship to biophysical properties.” Biotechnology Progress 40(6), e3481 (2024). doi:10.1002/btpr.3481.
Cardoso, S.; Bollmann, F.; Tappe, A. “Optimization of Tangential Flow Filtration for High-Yield, Scalable Downstream Processing of Adeno-Associated Virus.” Membranes 16(2), 73 (2026). doi:10.3390/membranes16020073.
Fernandez-Cerezo, L.; Benner, S. W.; Pollard, J. M. “Streamlining process characterization efforts using the high throughput ambr® crossflow system for ultrafiltration and diafiltration processing of monoclonal antibodies.” Biotechnology Progress 37(3), e3118 (2021). doi:10.1002/btpr.3118.
Loewe, D.; Dieken, H.; Grein, T. A.; Salzig, D.; Czermak, P. “A Combined Ultrafiltration/Diafiltration Process for the Purification of Oncolytic Measles Virus.” Membranes 12(2), 105 (2022). doi:10.3390/membranes12020105.
Helling, A.; Borujeni, E. E.; Leuthold, M.; Tindal, S.; Fernandez-Cerezo, L.; Brower, M. “Continuous concentration and diafiltration tangential flow filtration with scalable dual membrane technology.” Biotechnology Progress, e88501 (2026). doi:10.1002/btpr.88501.

FAQ

Technical FAQ

Questions engineers ask before fixing a UF/DF recipe

How many diavolumes are required for buffer exchange?

For constant-volume diafiltration with a well-mixed retentate, constant impurity sieving coefficient S and impurity-free incoming buffer, use N_D = ln(C_initial/C_target)/S. A 120-fold reduction with S = 0.95 requires 5.04 diavolumes. For a nonzero incoming-buffer concentration, use the general balance above and check its asymptote.

How is diafiltration buffer volume calculated?

Multiply the required number of diavolumes by the constant diafiltration retentate volume: V_buffer = N_D × V_DF. Five diavolumes at a 100 L retentate volume require 500 L of buffer, before overage and line volumes.

What does one diavolume mean?

One diavolume is a buffer volume equal to the constant retentate volume during diafiltration. In the ideal S = 1, impurity-free-buffer model, one diavolume removes 63.2% and leaves 36.8% of the starting concentration.

Can conductivity, pH or osmolality be entered as the component concentration?

Not directly. The exponential balance applies to a defined component concentration. Conductivity and osmolality combine several species, while pH depends on equilibria and electrostatic partitioning. Use a validated empirical relationship or an appropriate multi-species model.

How should membrane area and product loss be screened?

Use phase-specific qualified fluxes for concentration and diafiltration. Divide each phase permeate volume by its flux, add the area-time terms, and divide by the allowed process time. Screen yield as VCF^−S_p,UFexp(−S_p,DFN_D), then replace the assumptions with measured values.