Thin supported silica membranes
Prof. dr. ing. D.H.A. Blank
Prof. dr. ir. A. Nijmeijer
Dr. H.J.M. Bouwmeester
This thesis discusses several transport-related aspects relevant for the application of thin supported silica membranes for gas separation and nanofiltration.
The influence of support geometry on overall membrane performance is investigated. Planar (i.e., flat plate), tubular, and multichannel support geometries are investigated in numerical simulations of the membrane performance in gas separation. The emphasis is laid on the last two membrane geometries which are considered more suitable due to their greater surface-area-to-volume ratio and mechanical robustness. The dusty-gas-model (DGM) is used, and the contribution of different transport mechanisms occurring in a porous system (Knudsen diffusion, bulk diffusion and viscous flow) is accounted for. The comparison of geometries is performed in terms of the calculated flux and selectivity, in case of separation of pure H2 and of a 50-50% binary H2/CH4 mixture. The multichannel configuration is found to be far less efficient than the multitubular one due to the inefficiency of the inner channels, even in the case of highly permeable silica - where the transport is governed by the support – and even though the surface-area-to-volume ratio is higher in the multichannel configuration. For a proper prediction of transport behaviour it is crucial to account for the three transport mechanisms included in the DGM, especially in case of the binary mixtures. Calculations for the binary mixture show that the inner channels contribute to a considerable decline of the selectivity. Even values below unity may be obtained in the case of highly permeable silica, i.e. indicating a higher transport rate of CH4 compared to H2. For a leaking inner channel, a maximum in selectivity is observed for a certain value of permeability of silica for H2. Further improvement of the silica membrane layer would thus result in a decreased performance of the multichannel membrane.
The transport of binary mixtures containing an inert mobile component such as He and a condensable component such as H2O through a microporous silica membrane has been investigated using spectroscopic ellipsometry. Attention is focussed to the correlation between the flux of the more mobile component and the sorption properties of the condensable component. A linear decline of the normalized He permeance as a function of water occupancy is observed. This behaviour is agreement with a theoretical description of the transport where the microporous medium is considered as an ideal lattice of sites. However, departures from linear behaviour are observed experimentally at low temperatures, where water has low molecular mobility, which can not be accounted for by the theoretical description.
It is demonstrated that an improved membrane performance in gas separation can be achieved by incorporating an additional intermediate surfactant templated silica layer between the supporting mesoporous g-Al2O3 and the amorphous microporous silica top layer. The dual-layered silica membrane shows improved values of H2 flux and H2/CH4 permselectivity compared with that of a standard silica membrane.
Throughout this thesis spectroscopic ellipsometry is used as a non-destructive characterization technique to determine the thickness and porosity of the asdeposited membrane layers, but also to monitor in-situ the sorption of water by the membrane.
Improved ion retention can be achieved by using a bilayered nanofiltration membrane. The novel bilayered membrane concept is based on a difference in iso-electric point between the two separating layers in the membrane allowing improved ion retention over a large pH range. The composite membrane consisting of a g-alumina layer and a surfactant templated silica layer, similar to that investigated for the improvement of membrane performance in gas separation, is analyzed in terms of the retention of monovalent ions (Na+, Cl-) over the pH range 4-10. A good agreement is found between experimental data and model predictions, with a discrepancy at high pH (pH 10) in case of sodium retention.