Optimization & Process Simulation for Ultrafiltration

---

Before you implement an ultrafiltration step, define these key operation parameters:

• Crossflow rate
• Transmembrane pressure (TMP)
• Filtrate control
• Membrane area
• Diafiltration design

In addition, see below for characterization of performance and how to test your process.

Crossflow rate*

The crossflow rate will vary based on your choice of module and feed channel turbulence promoter. The higher the crossflow rate, the higher the flux at equal TMP. A higher crossflow rate increases the sweeping action across the membrane, which reduces the concentration gradient towards the membrane surface.

However, higher crossflow rates cause the product to experience more passes through the pump in a given amount of time, which can lead to product quality degradation. Also, higher crossflow rates require larger pumps and larger-diameter piping, which will increase the system holdup volume and could increase product losses due to unrecoverable holdup.

* Crossflow can be defined in a number of different ways; we are referring to the simplest definition, the feed flow rate (L/min per m2).

Best Practice: During optimization trials, it is important to choose an appropriate combination of feed flow rate and TMP to maximize flux. This will minimize process time and/or membrane area, allowing for optimal membrane area and pump sizing (minimal holdup due to pump sizing as well as pump passes).

Transmembrane Pressure (TMP)

The relationship between flux and transmembrane pressure can be understood by generating a TMP versus a flux curve:

• The first part of the curve is where the flux increases with increasing pressure (TMP) and is referred to as the pressure-dependent regime.
• The level part of the curve (knee of the curve) is known as the pressure-independent reqime; at this point, there is no gain in flux made by increasing pressure.
• When flux starts to plateau at the knee of the curve, the optimum process TMP operating point has been identified.

If the process is run with a TMP setpoint in the pressure-independent regime, you can achieve maximum flux and minimize the required membrane area. However, at this point, the protein wall concentration is high and could exceed a solubility limitation, leading to yield losses. Additionally, running at these conditions could cause fouling and eventual decrease in flux. On the other hand, if you choose a TMP setpoint in the pressure-dependent regime, fluxes are lower and more membrane area is required.

Ideally, for a standard UF/DF process, the optimum TMP to run a process is at the knee of the curve. This is the point where nearly the highest flux is achieved without exerting excessive pressure or reaching exceedingly high protein wall concentrations (polarization) or fouling.

Transmembrane Pressure: The average applied pressure from the feed to the filtrate side of the membrane. TMP [bar] = [(PF + PR)/2] - Pf

For additional information, see Protein Concentration and Diafiltration by Tangential Flow Filtration -- An Overview.

Filtrate Control

When using very open ultrafiltration membranes (any membrane above 100 kD NMWL), permeability is so high that nearly all of the feed flow is converted to filtrate with very little applied TMP. This results in high fluxes, similar to a normal flow filtration (NFF) mode. As a result, you lose the benefits of tangential flow filtration. In this situation, very high protein concentrations at the membrane surface and high membrane fouling may occur, especially during the process start-up. You must control the filtrate flow to reduce the filtrate rate and control the TMP at the low values required for robust tangential flow filtration (TFF) operations.

In a controlled flow filtrate operation, a pump or valve on the filtrate line restricts filtrate flow to a set value. In addition to reducing the filtrate flow to maintain adequate tangential flow, the pump creates pressure in the filtrate line to reduce the TMP while the feed and retentate pressures remain fixed.

• Normal Flow Filtration (NFF):
  100% of the fluid is convected directly toward the membrane under an applied pressure. Particulates too large to pass through the membrane accumulate at the membrane surface; smaller molecules pass through to the downstream side.
  
  
• Tangential Flow Filtration: Fluid is pumped tangentially along the surface of the membrane. An applied pressure forces a portion of the fluid through the membrane to the filtrate side and the remainder recirculates on the upstream side of the filter. Particulates too large to pass through the membrane pores are retained on the upstream side. However, these components do not build up on the surface of the membrane, such as in NFF applications, making TFF an ideal process for finer sized-based separations.

Diafiltration Design

Choose Diafiltration Control Mode
If you need to include a diafiltration step in your process, first choose the mode of diafiltration control:

• In a batch process mode, a large volume of diafiltration buffer is added to the recycle tank, then the retentate is concentrated. When a certain retentate volume is reached, another volume of buffer is added. This cycle continues until you reach the desired total volume of DF buffer. The benefit is that no level control is required. However, the buffer exchange is not as efficient and the product concentration cannot be optimized because protein concentration changes as the buffer is added and concentrated.
  
  
• Constant-volume DF mode is used more often than batch process mode. In this process, buffer is added to the recycle tank at the same rate filtrate is removed. The total volume of retentate remains constant throughout the process and requires some method of level control that will meter the addition of DF buffer to keep the retentate volume constant.

Best Practice: Since you must typically remove contaminants or residuals from the product to very low levels, incorporate a safety factor of at least two extra diavolumes and test the process to ensure that actual residual levels are acceptable.

Characterization of Performance (Optimization Trial)

You must test your process performance on actual feedstock. One of the most important experiments for performance characterization is to generate flux versus TMP curves at several feed flow rates and protein concentrations. This will determine product retention at each point. If the process contains a diafiltration step, generate the flux versus TMP curves in both the starting and final buffers, since flux and retention can change significantly with buffer conditions.


A typical trend of flux versus transmembrane pressure for a TFF process

Best Practices: Determine TFF performance at approximately three different feed flow rates (see manufacturer-recommended rates for the module used). Test a minimum of two different protein concentrations (initial and final feedstock concentration). Investigate at least five transmembrane pressures for each feed flow and protein concentration. TMPs will vary but will typically be in the range of 5 to 50 psid. Always begin with the least fouling conditions and move towards more fouling conditions, since the start-up method and order of conditions tested can impact the results. (Least fouling conditions are those at low protein concentrations, low TMPs and high feed rates.) Start with the highest feed rate and lowest protein concentration and TMP. For each test point, calculate flux, TMP and retention. Then, generate graphs showing flux versus TMP at different feed flow rates and protein concentrations as well as retention versus TMP at different feed flow rates and protein concentrations. It is important to calculate yield as well as mass balance. The collection of this data enables the choice of successful and robust operating conditions. Mass Balance states the principle of conservation of mass: VinitialCinitial = (Vretentate x Cretentate)+ (Vpermeate x Cpermeate)

Determine Concentration of Diafiltration in the Process
After choosing a control mode, determine the concentration of the diafiltration step within the process. Flux typically drops as a protein is concentrated:

• Diafiltration at lower protein concentrations maximizes flux; however, the total volume of product to diafilter is higher, increasing the membrane area and buffer volume required.
• Diafiltration at higher protein concentrations reduces flux, increasing the need for membrane area but decreasing the required buffer volume.

Therefore, there is an optimum protein concentration at which to perform diafiltration where the tradeoff between flux and volume is balanced and the minimum membrane area or process time is achieved.

To determine the optimum point to diafilter for standard pressure-controlled UF/DF processes:

• Plot flux versus the log of protein concentration. (Plot this data with the protein in both the initial and final buffers, since flux can change with different buffers).
  
  
  
  A typical trend of flux versus protein concentration in different buffers
  
  
• Choose several protein concentrations along each curve that span the range from initial to final concentrations expected in the process and calculate the value of the DF optimization parameter at each point.

DF Optimization Parameter : = C * Jf where: C=product concentration in feedstock at data point [g L-1] Jf=filtrate flux at data point [L m-2 h-1]

Determine the Product Concentration
Plot the DF optimization parameter versus protein concentration for each buffer to find the product concentration that maximizes the value of the optimization parameter (this is the optimum concentration at which to diafilter to minimize membrane area). If the goal is to minimize the expense of DF buffer, select a concentration higher than the CbDF, keeping in mind that higher membrane area or process time will need to be balanced. Selecting a concentration point below the optimum CbDF will result in higher membrane area or process time, and higher buffer usage/larger recirculation volume. The ideal is to select a concentration between the DF optimization points, or the lower concentration.

Membrane Area Required to Perform Process Simulation

You can estimate the membrane area for the initial process simulation from the optimization data. The membrane area required for the final unit operation can be determined from a linear scale-up.
Learn about Pellicon® ultrafiltration cassettes for demanding ultrafiltration processes.

Since flux is filtrate flow rate divided by both area and time, membrane area is a function of the total process time:

Membrane Area [m2]: Filtrate volume [L]/Flux [L m-2h-1]* Process time [h]

A longer process time leads to lower membrane area requirements; a shorter process time will require higher membrane area. Your facility constraints will influence your choice of membrane area based on this trade-off.

Flux typically drops as protein concentration increases, so choose an average flux. For a robust scale-up, always incorporate a safety margin into the membrane area requirements to account for lot-to-lot variability in membrane permeability, feedstock characteristics and batch volumes. Typically, a safety margin of at least 20% extra membrane area is used.

To determine the membrane area required for the final unit operation, read A Hands-On Guide to Ultrafiltration/Diafiltration Optimization using Pellicon® Cassettes.

Average Flux* = Total process volume / Total process time *Determined from process simulation data

During the process simulation, collect the following data:

• Monitor flows and pressures and retention.
• Collect samples of all initial and final streams.
• Calculate process time to ensure that it is within the expected range.
• Test the quality of the final product with reliable assays.
• Calculate not only the yield, but also mass balance.

Best Practices: Run the process several times to understand how robust it is to feedstock variability and multiple cycles, and to guard against unexpected performance degradation.

Contact the Ultrafiltration Team to explore how we can help you avoid costly errors by verifying that your design falls within Merck’s guidelines (actual process performance to be verified with optimization and process simulation trials).