Scale-Down And Scale-Up Theory For Tangential Flow Filtration (TFF) Systems
In conventional dead-end filtration, the direction of feed flow is aligned with the direction of filtration, making it highly susceptible to the formation of a filter cake or gel layer on the membrane surface. As filtration continues, the deposited layer gradually thickens, resulting in a decrease in filtration flow rate. Tangential flow filtration (TFF) refers to a filtration process in which the fluid flow direction is perpendicular to the direction of filtration. During TFF, the fluid flows tangentially along the membrane surface, generating shear forces that help reduce solute deposition on the membrane and mitigate concentration polarization. TFF is widely used in the industrial production of biopharmaceuticals. Typically, a TFF system consists of a feed reservoir, a peristaltic pump or centrifugal pump, an inlet pressure gauge, a membrane cassette, an outlet pressure gauge, a backpressure valve, and associated tubing and piping. A schematic diagram of the system is shown below.
Permeate flux and sieving coefficient are two important parameters in membrane filtration processes. Permeate flux is defined as the volume of liquid passing through a membrane per unit membrane area per unit time. The sieving coefficient is the ratio of the concentration of the target product in the permeate to that in the retentate. When comparing tangential flow filtration (TFF) systems of different scales, experimental variables must be controlled to minimize interference. The membrane cassettes should have consistent membrane materials, molecular weight cut-offs (MWCOs)/pore sizes, manufacturers, and channel screen structures within the same product series. The feed solutions must have identical compositions and concentrations, and experiments should be conducted under uniform initial temperature conditions. Once these initial conditions are satisfied, the scale-up process should follow three fundamental principles: maintaining the same feed volume per unit membrane area, the same membrane surface shear rate, and the same initial operating flux. In addition, dynamic concentration–diafiltration processes, experimentally verified fluid dynamics, and system hardware deviations must be taken into account. These measures help ensure that the permeate flux and sieving coefficient of filtration systems at different scales remain as consistent as possible at every time point.
Principle 1: Maintain the Same Feed Volume Processed per Unit Membrane Area
Permeate flux and the sieving coefficient are closely related to the intrinsic properties of the membrane. Provided that the membrane material, pore size, and channel screen structure remain consistent, achieving comparable permeate flux and sieving performance across filters of different sizes requires the membrane surface conditions (including cake deposition and the degree of concentration polarization) to remain essentially consistent at all times. To achieve comparable levels of membrane fouling and concentration polarization, the volume of feed solution processed per unit membrane area must be the same across filtration systems of different scales, assuming identical feed composition and concentration. Therefore, the ratio of feed volume to membrane area must be maintained across different filter sizes, ensuring a consistent feed volume processed per unit membrane area (L/m²).
Tangential flow filtration (TFF) in biopharmaceutical manufacturing typically involves both ultrafiltration (UF) for concentration and diafiltration (DF) for buffer exchange. Matching only the initial feed volume per unit membrane area is therefore insufficient. As concentration proceeds, the feed concentration and viscosity increase continuously, affecting the degree of concentration polarization and the rate of membrane fouling. Consequently, the volumetric concentration factor (VCF) must be kept consistent between small-scale and large-scale trials at every time point. During diafiltration, the number of diafiltration volumes (DV) must be exactly the same, as it directly determines the removal efficiency of low-molecular-weight impurities. In addition, all process endpoint criteria-including retentate volume, target protein concentration, maximum DV, and TMP alarm thresholds-must be fully transferred from the scale-down model to full-scale manufacturing.
Principle 2: Maintain the Same Membrane Surface Shear Rate
In the biopharmaceutical industry, most feed solutions requiring filtration contain biologically active substances. The shear forces exerted by the fluid flowing across the membrane surface affect not only cake formation and concentration polarization, but also the biological activity of the product. Therefore, the membrane surface wall shear rate must be maintained consistently between small-scale and large-scale tangential flow filtration (TFF) systems during experimental operation.
In general, large-scale membrane cassettes feature rectangular flow channels with integrated woven support screens. In contrast, the flow channel geometry and screen configuration of small-scale laboratory membrane cassettes differ significantly from those of production-scale cassettes.
Relying solely on computational fluid dynamics (CFD) simulations to determine wall shear rates can introduce errors due to model assumptions. In practice, CFD simulations should be validated against the experimentally measured pressure drop (ΔP) between the membrane cassette inlet and outlet. The pressure drop across the retentate-side flow channel is a direct engineering indicator of shear conditions. For membrane cassettes within the same product series, the channel pressure drop (ΔP) should be matched between small-scale and large-scale systems. The recirculation flow rate should be adjusted accordingly to bring the wall shear rates of the membrane cassettes at different scales as close as possible. However, consistent membrane surface shear rates do not necessarily mean consistent overall shear exposure of the process fluid. Small-scale laboratory systems generally use peristaltic pumps, whereas industrial production systems predominantly use centrifugal pumps. The pumps themselves can introduce additional shear stress that may damage proteins, viruses, and other biological materials. During scale-up, the risks of product aggregation and loss of biological activity caused by pump-induced shear must therefore be carefully evaluated.
Principle 3: Maintain the Same Initial Operating Flux
Under consistent feed solution properties, membrane surface characteristics, and operating conditions, if the initial permeate flux is the same under filtration conditions and the membrane surface evolves in a similar manner, the subsequent filtration flux will generally remain consistent. Although membrane cassettes from the same manufacturer, with identical membrane materials and pore sizes, are used before and after scale-up, differences in cassette flow channel geometry mean that the overall filtration resistance of the systems may not be equal. When filtration resistance differs between filters of different sizes, different transmembrane pressure (TMP) settings are required to achieve the same permeate flux.
TMP is generally defined as the difference between the average of the inlet and outlet pressures and the permeate-side pressure. Since the permeate outlets of both small-scale and large-scale filters are generally open to atmospheric pressure, the permeate pressure is 0 Pa (gauge pressure). Therefore, TMP is calculated as the average of the inlet and outlet pressures. This value represents the overall average TMP across the membrane cassette. However, TMP varies spatially along the flow channel, meaning that the local TMP may exceed the average value. Filtration processes generally operate in one of two control modes: constant TMP or constant flux. Constant-TMP operation is commonly used for protein ultrafiltration. During scale-up, the TMP setting from small-scale trials should not be directly replicated; instead, the flux–TMP inflection point should be used as a reference. Constant-flux operation is commonly used in microfiltration and viral vector processing. During scale-up, the permeate flux should be prioritized as the controlled parameter, while an upper TMP alarm limit should also be established.
When all three principles are satisfied, together with consistent initial conditions, the permeate flux of small-scale and large-scale filters should theoretically be the same at any given time after filtration begins. If the permeate flux profiles of the two systems are identical at all times, their total filtration times will also be equal.
The three scale-up principles outlined above are derived under ideal filtration conditions. In actual process scale-up, however, non-ideal deviations arising from differences in equipment hardware must also be considered. These primarily include system hold-up volume, differences in recirculation loop configuration and material residence time, and heat generation caused by fluid recirculation and shear. Furthermore, this scaling theory is applicable only to membrane cassettes of the same configuration and product series. Significant deviations from linear scale-up relationships may occur when the membrane module configuration changes or when processing high-viscosity, high-solids-content, or non-Newtonian fluids.
Because actual operating conditions can rarely satisfy all theoretical assumptions perfectly, theoretical equations alone are insufficient to directly achieve process transfer. A comprehensive set of criteria must therefore be established to assess the validity of the scale-down model. By comparing flux–time profiles, sieving coefficients, product yields, impurity removal performance, and critical quality attributes (CQAs)-including aggregate content, particle levels, and biological activity-across filtration systems of different scales, the scale-down model can be considered representative and reliable for process scale-up only when consistency is demonstrated across all key performance indicators.
