Because they do their job, membrane filters are doomed. As they work, they foul because rejected material builds up on the top surface and microscopic particles accumulate internally. This fouling reduces a membrane’s flux, cutting the rate at which desired material can flow through. That degradation is a problem for biotechnology applications that involve filtering protein suspensions. Composed of an array of small pores, membranes are used to filter protein suspensions, separating the proteins from a solution containing proteins and cells. A microfiltration membrane has pores less than 1 μm in diameter. That’s much larger than proteins, which measure a few nanometers in size. Proteins should, therefore, be able to easily traverse the membrane, but they don’t. The membranes often foul, greatly reducing the protein flux. Eventually, the membrane is either replaced or cleaned. Membrane filters consist of an array of small pores that allow desired components of a solution through while keeping out everything else. The filters foul, as seen here with a multiphoton microscopy top view of the formation of a cake produced by a mixture of 6- and 10-μm microspheres. Images courtesy of Zhanfeng Cui, Oxford University. An ideal monitoring method would detect fouling and determine the culprit. Unfortunately, the best and most complete techniques require that the membrane be removed and analyzed. Existing in situ methods are limited in the membranes they can handle or the measurements they can make. Researchers from Oxford University in the UK turned to multiphoton microscopy. Chemical engineering professor Zhanfeng Cui said that the method offers the possibility for 3-D in situ observation of the formation of cake and deposition of foulants, such as proteins, onto the membrane. Observing fouling as it occurs and gaining 3-D information could allow scientists to see the structures being deposited and cleaned off. A membrane fouled by washed yeast cake formation was imaged using multiphoton microscopy, which provides a means to monitor and study such fouling online. As detailed in the July 4 issue of Langmuir, the researchers used multiphoton microscopy to image the fouling of polycarbonate membranes by fluorescently labeled bovine serum albumin (BSA) and ovalbumin. For this they employed a Zeiss multiphoton laser scanning system, which had a Coherent Ti:sapphire laser illumination source. The laser produced femtosecond pulses at 800 nm that traveled through the system’s optics and were focused down to a small volume on the membrane. Within that spot, two or more photons interacted with the sample, giving rise to effects similar to that of an incoming 400-nm or shorter wavelength photon. This example shows membrane filter fouling by proteins and their aggregates. Ovalbumin is labeled with a red dye, and bovine serum albumin, with a green dye. The view is from the right, as imaged with multiphoton microscopy. The fluorescence traveled back out and was captured by either the system’s detectors for 3-D imaging or by a spectral analysis instrument from Applied Spectral Imaging Ltd. of Tel Aviv, Israel, for an exact determination of the wavelength of the emission. The maximum spatial resolution of the system was 0.10 μm in X and Y and 0.52 in Z, although most of the images were captured at lower resolutions. The spectral detector had a resolution that ranged from The system acquired optical data in 2-D planes, changing focal depth to obtain the third dimension. The process took from one to 20 minutes, depending upon the resolution. Cui noted that, although this setup had been modified for another, unrelated set of experiments, a standard system would work for the fouling research. The researchers pumped fluorescent and native proteins from a reservoir through a cross-flow module housing a polycarbonate membrane with 0.22-μm pores and examined the membrane with multiphoton microscopy. They monitored the pressure up- and downstream of the membrane using pressure transducers. They completed the loop by sending fluid back into the reservoir. The two transducers allowed the researchers to keep the pressure driving the fluid across the membrane constant. They periodically weighed what made it through the membrane to determine the flux, returning the weighed permeate to the reservoir. After introducing the suspension and allowing it to flow briefly, they stopped the flow to take an image and then restarted it. They did this to avoid streaking or blurring while the image was being acquired. They looked at fouling using BSA labeled with fluorescein and ovalbumin labeled with Texas Red, and an equal mix of the two. The addition of the dye altered the fouling of BSA, causing it to occur much sooner. This, they noted, could be caused by the conjugation of the dye with the protein or perhaps could arise from differences in fouling performance between lots of BSA, which other groups have reported. There was no difference between labeled and native proteins in the case of ovalbumin or in the 50-50 mix of BSA and ovalbumin. The multiphoton microscopy images correlated well with flux decline curves. The BSA, for example, initially fouled in small patches that covered less than 10 percent of the membrane surface. Its flux did not change much initially. The ovalbumin, on the other hand, reached 96 percent coverage in five minutes, corresponding to a sharp decline in flux. Because the images agreed with filtration data and models that were derived from other studies, it showed that the method worked. Cui noted that this demonstration means that nonlinear laser optics could prove useful in research and other settings. “We may be able to develop online optical sensors to monitor fouling and cleaning in membrane systems in the future,” he said. The researchers plan to use the tool for further research. Cui said that a goal is to test membrane-solution compatibility and membrane suitability, as well as cleaning protocols. The technique also is being considered for more than research. “We have had inquiries to characterize solutions and their fouling tendency,” Cui said. Contact: Zhanfeng Cui, Oxford University, UK; e-mail: zhanfeng.cui@eng.ox.ac.uk.