New Binary Gas Integrity Test Improves Membrane Quality Assurance

April 1, 2011

BioPharm International

Volume 24, Issue 4

The authors developed a test for defects in filter membranes.

The authors have developed a test for defects in filter membranes based on the principle of differing gas permeabilities through the liquid layer of a wetted membrane. Both the binary gas integrity test and the standard gas–liquid diffusion method were applied to a newly developed virus clearance filter. Results demonstrated that the binary gas test provided a significantly higher level of virus retention assurance compared to the air–water diffusion test. The authors conclude that the binary gas test provides superior defect detection sensitivity in virus filters.

Over the years, there have been a number of nondestructive integrity tests for filtration membranes conducted as part of manufacturing quality assurance. Tests are typically implemented by the filter manufacturer as an additional quality assurance test prior to shipping the product to users.

Millipore has developed a test designed to be more sensitive to detection of defects. The new high sensitivity binary gas integrity test is based on the principle of differing gas permeabilities through the liquid layer of a wetted membrane that results in a concentration enhancement of the faster permeating gas.

The test makes use of the fact that the permeate composition in an intact (integral) membrane, with no defects, can be predicted based on the transport properties of the gases permeating through the liquid layer and the known operating conditions. So, if the test spots a deviation from the expected concentration, it is a clear indication of a defect or the presence of open pores.

The binary gas test has low sensitivity to membrane porosity, liquid layer thickness, and membrane area. This means that integral devices will exhibit a relatively narrow range of test values, making it easier to spot defects.

To "test the test", researchers applied both the binary gas integrity test and the standard gas–liquid diffusion method to a newly developed virus clearance filter. Results demonstrated that the binary gas test provided a significantly higher level of virus retention assurance compared to the air–water diffusion test.

TESTING FOR DEFECTS THAT COMPROMISE PERFORMANCE

Integrity testing of microporous or ultraporous filters is routinely used to detect the presence of oversized pores or defects that can compromise the filter's retention capability. Test choices include the particle challenge test, the liquid–liquid porometry test, the bubble point test, the gas–liquid diffusion test, and diffusion tests measuring tracer components. Although each test has benefits, there are also compromises that must be made.

The particle challenge test is destructive and therefore not applicable as a pre–use test. The liquid–liquid porometry and bubble point tests are useful for ensuring that the user has selected a membrane with the proper nominal pore, but are not sensitive enough to identify small numbers of small defects, particularly for filters larger than 47 mm. With the gas–liquid bubble point test, a single or few small defects may add only a small amount of gas flow that cannot be distinguished from the background diffusive flow rate through the integral part of the membrane in a filter device.

The most commonly applied nondestructive integrity test for membrane filters, especially virus filters, is the gas–liquid diffusion test. A wetted membrane provides a liquid layer across which diffusive air flow occurs (see Figure 1a).

Figure 1: Gas diffusion through a wetted membrane: (a) integreal and (b) non-integral membrane.

As pressure is increased, diffusive flow increases linearly until either the liquid layer begins to thin or until the bubble point is reached, whereupon robust bulk air flow commences. As shown in Figure 1, a measured gas flow rate more than that predicted for an integral membrane signals the presence of a defect.

The sensitivity of this test is limited by the minimum detectable excess flow. There can be significant device–to-device variability in gas diffusion flow rates of integral membrane filter devices due to differences in membrane area, membrane thickness, membrane porosity, and pore tortuosity (twists and turns). Other factors such as thinning of the liquid due to evaporation, liquid retention of membrane support layers (porous non-wovens, for example), and membrane movement or compression can also affect measured gas diffusion rate. This variability in gas flow rate acts as "background noise" that can diminish the sensitivity of the gas–liquid diffusion test.

NEW INTEGRITY TEST DEVELOPED

The binary gas test was developed to increase the level of defect detection sensitivity, and has been found to be particularly useful for evaluating virus filtration membranes. It is based on the principle of differing gas permeabilites of a gas mixture's two components through the liquid layer of a wetted membrane.

Unlike the single gas in the gas–liquid diffusion test, the binary gas test relies primarily on the measurement of downstream gas composition rather than downstream flow rate. In an integral membrane, the permeate gas is depleted of the slower permeating gas. However, if a defect is present, the leak through the membrane will contaminate the permeate stream, resulting in an elevated concentration of the slower permeating gas.

A key advantage of the binary gas test is that the permeate concentration expected through an integral membrane is well defined and its sensitivity is not compromised by most of the background noise factors cited for the gas-liquid diffusion test. Figures 2a and 2b illustrate how the test works.

Figure 2: Binary gas diffusion through a wetted membrane: (a) integral and (b) non-integral membrane.

USING GASES TO MEASURE DEFECTS

Diffusion can be specified for each component in a gas mixture permeating across a membrane. Assuming that the gas is completely mixed on both sides of the membrane, the composition of the permeate gas can be calculated from the ratio of diffusive flow rates of the two components, and the inlet side composition.

The composition of the permeate gas is independent of membrane thickness, tortuosity, porosity, and area. It is also independent of the pressure difference across the membrane but instead is dependant on the pressure ratio. The permeate composition does, of course, depend on the feed side composition. To maintain a constant feed side composition, a constant sweep flow must be applied.

A measured binary gas composition can be used to estimate a defect size in a device. The calculated defect size can, in turn, be used to estimate the liquid flow rate through the defect and the total volume of liquid passing through the defect for a given time period. Therefore, the permeate gas concentration can be used to predict the loss in virus log reduction value (LRV) due to the defect.

The sensitivity of the binary gas test is related to the selectivity (ratio of permeabilites) of the gases through the liquid layer. To maximize the sensitivity of the binary gas test, the gas pair should have a high selectivity. Figure 3 shows normalized permeabilites of some common gases in water. For this study, researchers selected a carbon dioxide/ hexafluoroethane (CO2/C2F6) pair, with a selectivity of about 1000.

Figure 3: Normalized permeability of common gases in water at 25 °C.

The selection of the concentration of the gases in the mixture was influenced by a number of factors, including the ease of composition measurement, gas flow rate through the membrane, and economic considerations. The concentration selected enables convenient flow and composition measurement for even relatively small membrane areas (as low as 3 cm2 membrane area).

COMPARING HOW THE TESTS MEASURE CONTROLLED DEFECTS AND VIRUS RETENTION

The research included conducting integrity tests on virus filtration membranes using both the air–water diffusion test and the binary gas test. The virus filtration membrane used was Millipore Viresolve Pro (asymmetric PES) membrane in flat sheet or disc formats.

Researchers introduced defects by laser drilling 2–10 µm diameter holes in the center of the membrane discs (see Figure 4).

Figure 4: Laser hole drilled through a Viresolve Pro membrane disc.

For the air diffusion test, pressurized air was applied to the upstream side of the membrane and downstream air flow rate was measured using a mass flow meter.

For the binary gas test, the CO2/C2F6 test gas was introduced to the membrane at 345 kPa, and a constant sweep gas rate at a 4:1 ratio relative to the permeate flow rate was maintained through the vent port of the filter holder. Gas composition was measured using fourier transform infrared spectroscopy (FTIR) (inDuct FTIR, MKS Instruments). Measurements were recorded continuously until an essentially steady state permeate composition was achieved, typically within 15–20 minutes, which was the time required to fully flush out the residual air from the volume downstream of the membrane, the sample lines leading to the FTIR, and the FTIR sample chamber. A small volume custom cell was procured for the FTIR in order to minimize the total internal volume downstream of the membrane and thereby reduce the overall test time.

After the initial air diffusion and binary gas testing, researchers challenged the membrane devices with a solution consisting of a bacteriophage mixed with polyclonal human immunoglobulin in a buffer solution. The solution was filtered through the membrane until flux had declined by 75% compared to the clean buffer. They then collected feed and permeate samples, determined the infectious titer, and calculated the virus LRV.

Figure 5: Defect detection by the air diffusion test.

Figures 5 and 6 show the air diffusion and binary gas test results as functions of defect size. The solid lines are the model predictions for the air diffusion and binary gas tests. The shaded regions in each graph show typical test value ranges for integral membranes. These regions represent background noise against which a signal for a defect must be compared. Figure 5 shows that a 2 µm defect was not "visible" to the air diffusion test because the additional flow rate due to the defect was not large enough to increase the total flow rate beyond the range typically measured for integral membranes. In contrast, as shown in Figure 6, the elevated C2F6 concentration in the permeate is a clear signal for the same 2 µm defect. This result was an unambiguous demonstration of the binary gas test's superior defect detection sensitivity. For defect sizes larger than 2 µm, both tests provided a strong signal for a defect.

Figure 6: Defect detection by the binary gas test.

The research also showed that the loss in LRV compared to an integral membrane due to a single defect can be predicted from the binary gas value. This means that a maximum allowable binary gas value can be established for a desired level of LRV assurance based on a worst case assumption of a single defect.

In addition to using the test in conjunction with controlled defects, researchers looked at defect detection sensitivity in manufactured devices. Both tests were applied to a set of prototype Viresolve Pro devices. A portion of the two–layered membrane used to manufacture the devices was tested for virus retention and the LRV of the three discs was determined to be 5.9. The LRV of the devices, however, ranged from 5.0 to 5.9. While the air diffusion test did not differentiate among these devices, the binary gas test showed clearly elevated values for the two devices with lower LRV values (see Table I).

Table I: Comparison of integrity test sensitivity between the air-water and binary gas tests on prototype Viresolve Pro devices.

CONCLUSION

Compared to the conventional air-water diffusion test, the binary gas test provides superior defect detection sensitivity in virus filters. While the air diffusion test provided an LRV assurance of about 4.5–5.0 for the virus filters studied, the binary gas test can provide an LRV assurance of greater than 6.0.

The greater sensitivity of the binary gas test is due to a much more favorable signal–to–noise ratio than the air-water diffusion test. Unlike the gas–liquid diffusion test, the binary gas test has low sensitivity to membrane porosity, liquid layer thickness, and membrane area. Other factors that can confound the sensitivity of the air diffusion test such as thinning of the liquid due to membrane asymmetry or evaporation, liquid retention of membrane support layers (porous non–wovens, for example), and membrane movement or compression have a much lower impact on the sensitivity of the binary gas test.

Sal Giglia*is principal applications engineer, and Mani Krishnan is director of single use technologies, both at EMD Millipore, sal.giglia@merckgroup.com