DESIGN OF POLYMER-CERAMIC COMPOSITES FOR MEMBRANE CHROMATOGRAPHY
4.3 Results and discussion
4.3.1 Phase inversion micromoulding feasibility
mg/mL. To measure the static binding of the polymeric membrane references, a membrane with a known volume was immersed in a 2 mg/mL BSA solution and gently mixed for 48 hours. The absorbance of the solution was then measured using an Agilent 8453 UV/vis and the reported absorbance at 280 nm was used to determine the concentration of BSA in the solution. The mass of BSA bound was then determined using a mass balance.
A similar process was used to measure the static binding capacity (SBC) of the composite membranes. The samples were immersed in a 2 mg/mL BSA solution and gently rocked for 48 hours. The absorbance was then measured and the binding capacity calculated before the samples were rocked for another 72 hours. The absorbance was then measured again and the binding capacity calculated. The addition of the second absorbance measurement was to account for the increased thickness and reduced mass transfer in the composites. The initial experiments for comparison to the polymeric membranes used BSA dissolved in H2O, all subsequent measurements used BSA dissolved in 50 mM TRIS.
Dynamic binding measurements using 2 mg/mL BSA in 50 mM TRIS buffer were conducted using composites with formulations of 54-0.25 and 38-0.4. To run the measurement, the sample was first loaded into a Swinney filter holder (Pall Corp.) and was equilibrated using 50 mM TRIS buffer. The BSA solution was then introduced via a syringe pump to the device at a rate of 150 µL/min (or 2 membrane volumes/min). The filtrate was analyzed with time-resolved measurements on the Agilent 8453 UV/vis. The 10%
breakthrough curve method, as described in Ch. 3, was used to determine the dynamic binding capacity.
Figure 4.5: SEM micrographs showing the following: neat ceramic a) cross-section & b) surface, composite without surface functionality c) cross-section & d) surface, composite with ECH functionality e) cross-section & f) surface, and composite with PEI gel layer g) cross-section & h) surface.
that transverse the entire membrane. The corresponding surface (perpendicular to the freeze- casting direction), Fig 4.5b, demonstrates the morphology of the oriented pores as well as the average pore diameter of 20 µm. The composite presented in Figure 4.5c&d was infiltrated without modifying the surface of the ceramic. In panel c, there is a segment of the polymer matrix in the middle of the micrograph that has a morphology that closely matches the contours of the nearby ceramic pore wall. It is also noteworthy that the ceramic surfaces that are visible are all bare. In the surface view from panel d the pores are mostly filled with the polymer matrix, but there are many cases where there is a debonded interface between the polymer matrix and one side of the pore.
The composite presented in Figure 4.5e&f had the surface modified using reaction (1) from Figure 4.3b prior to the infiltration and phase inversion micromoulding. The polymer matrix once again fills the pores in panel e and the ceramic walls that are visible are lightly decorated in microparticles from the polymer matrix. Panel f shows that the ceramic pores are completely filled and there are no visible gaps between the polymer matrix and the pore wall. The composite presented in panels g& h of Figure 4.5 had the surface modified using reaction (2) from Figure 4.3b, producing a functional PEI gel layer prior to infiltration and phase inversion micromoulding. The polymer matrix fills the pores in panel g and the ceramic walls that are visible are decorated with a higher density of microparticles/polymer matrix than panel e. Panel h shows that the ceramic pores are once again completely filled and there are no visible gaps between the polymer matrix and the pore wall.
The observations of the behavior of the polymer matrix in panels c and d, provide several key insights on phase inversion micromoulding and how to stably integrate the mixed-matrix membrane with the ceramic scaffold. The match between the morphology of the polymer matrix and the contours of the scaffold wall in panel c indicates that phase inversion micromoulding is capable of replicating features on the order of 10 µm. However, the phase inversion process does not seem to prevent debonding of the polymer matrix from the pore wall as seen in both the bare pore walls in panel c and the gaps in between the polymer matrix and ceramic scaffold in panel d. While there is a significant probability that
the gaps between the polymer matrix and ceramic scaffold in panel d are due to the drying process, the presence of any gaps or debonding in the wet state could lead to channeling and result in poor membrane performance. Therefore, it was decided to covalently bind the polymer matrix to the ceramic scaffold to suppress debonding.
The surfaces in panels f&h show pores that are filled with no indications of gaps or debonding from the ceramic scaffold. Similarly, the cross-sectional images in panels e&g both exhibit ceramic walls that are decorated with PEI particles and small sections of the polymer matrix. However, the density of decorating material on panel e is less than half of what is observed in panel g. It was concluded that the decorating material in panel e stems from functionalizing the surface because of the complete absence of decorating particles when the ceramic surface has not been modified (Fig. 4.5c). The discrepancy in the density of adhered PEI particles and polymer matrix is attributed to the difference in the number of reactive sites available from the surface functionalization.
Consider first a single pore that is assumed to be a perfect cylinder with diameter of 20 µm and height of 1.6 mm. The corresponding surface area and volume are 1 ∗ 105 µm2 and 5∗ 105 µm3, respectively. Assuming a monolayer density of 4 ATMS molecules/nm2 on silicon dioxide18 and that only 50% of the SiOC ceramic scaffold is silicon dioxide19, there are approximately 3 ∗ 10−13moles of ATMS per pore. Using reaction (1) from Figure 4.3b to further functionalize the surface and assuming any side reactions of ECH may be ignored at room temperature, there are 6 ∗ 10−13moles of halide per pore in the ceramic scaffold available to react with the amines in the polymer solution. This concentration should be considered an upper bound due to secondary reactions, such as the halide on an already bound ECH molecule reacting with a nearby amine, reducing the actual number of halides.
Using reaction (2) from Figure 4.3b as the second functionalization step produces a conformal PEI gel layer with an average thickness of 500 nm. Following a similar analysis as above, the interface of the gel layer is assumed to form a perfect cylinder with a diameter of 19 microns and height of 1.6 mm. The corresponding surface area and volume are 0.96 ∗ 105 µm2 and 4.5 ∗ 105 µm3, respectively. Assuming that the concentration of halides may
be approximated as a monolayer of ECH that covers the entire gel layer, the monolayer density was estimated to be 8 ECH molecules/nm2 from the topological polar surface area of 0.125 nm2/ECH molecule20. The resulting concentration of halides is approximately 13 ∗ 10−13 moles of halide per pore. Although the calculated halide concentrations for the two reaction sequences are of the same order of magnitude, the value from reaction (1) is an upper bound that ignores a multitude of side and secondary reactions. In contrast, the value calculated for reaction (2) should be considered a lower bound due to TEP swelling the PEI molecules at the gel interface. The swelling of the interfacial region leads to more reactive sites being accessible further improving the bonding between the gel layer and the PEI microgels in the polymer solution. Due to the superior adhesion between the polymer matrix and ceramic scaffold when using the PEI gel layer, all composites used for BSA binding experiments were fabricated with a PEI gel layer unless otherwise indicated.