Effects of solute polarization on the hydrodynamics within membrane-fluid-solute systems, with special application to hollow fiber bioreactors.

Abstract

The primary objective of this study was to assess the influence of osmotically active macromolecules on the hydrodynamics of hollow fiber membrane systems which, in turn, influences solute distributions within these bioreactors. This included verifying the phenomenon of concentration polarization of bovine serum albumin (BSA) in a rectilinear flow cell designed to incorporate the salient features of the hollow fiber bioreactor. Following characterization of the membrane used in the study, experimentally determined concentration profiles of BSA were compared with those predicted from theoretical analysis. The membrane characterization revealed local variability in the permeability of the membrane: L$\rm \sb{p}/\mu$ was found to vary between 2.80 $\times$ 10$\sp{-11}$ and 8.19 $\times$ 10$\sp{-11}$ m$\sp2$s/kg. The membrane permeability did not exhibit hysteresis effects due to pressure nor were there substantial changes in the permeability as a result of membrane aging. The permeability of the membrane decreased by approximately 30% following conditioning with a solution of BSA and was successful at retaining the protein. Transient BSA polarization was observed in the extra-luminal space of the rectilinear flow cell after 12, 24 and 36 hours of operation. The theoretical concentration profiles of BSA over predicted the experimental concentration profiles when assuming a membrane permeability equal to the average of the membrane coupon measurements. Further experiments consisted of examining possible dilution of the BSA samples obtained from the flow cell as well as investigating the theoretical assumption of a one-dimensional model. Sedimentation experiments showed that settling of the protein was not responsible for diluting the BSA samples. The local variability in the membrane permeability is thought to contribute to the discrepancy between the theoretical and experimental concentration profiles, as well as transverse concentration gradients not considered by the model. (Abstract shortened by UMI.)