The sol-gel method is versatile, offering great pore size tailoring in the sub-nanometer range in the preparation of amorphous silica membranes. As such, these membranes are ideal to separate gases based on their sizes, a mechanism known as molecular sieving. Further improvements with silica membranes have been attained by functionalization of the silica matrix with carbon templating and lately with metal oxide doping. Cobalt oxide and nickel oxide have been used as the metal oxides of choice by the scientific community. However, there is a knowledge gap associated with the effect of other types of metal oxides such as iron oxide in silica membranes. Of particular attention, the effect of multiple metal oxides co-embedded in silica membranes for gas separation has yet to be reported. Therefore, this thesis has focused on the morphological effect of single and binary metal oxide doping in silica matrices for hydrogen separation at high temperatures (up to 500 oC). In this thesis, iron and cobalt were selected for incorporation into the silica matrix, both as a individual metal oxides (iron oxide silica) and as a binary metal oxide (iron cobalt oxide silica) and the effect on the structure, chemistry, stability and performance of resulting membranes was systematically investigated.
The incorporation iron oxide has a significant effect on the resultant silica matrix. Interestingly, the increase of the iron content led to higher weight losses of resultant xerogels during calcination though this effect resulted in no significant variation of silica functional groups. The heat treatment was more significant, as surface areas and pore volumes of xerogels varied proportionally with the calcination temperature leading to the densification of the material matrix at higher temperatures. Nevertheless, the water content greatly affected the formation of functional groups and microstructure of iron oxide silica matrix. A low water/silica ratio resulted in almost dense materials and by increasing the water/silica ratio, the formation of siloxane bridges (Si-O-Si) increased. As a consequence, the surface area and total pore volumes of xerogels increased, whilst the average pore size remained almost constant. Hence, high water content was beneficial for the microstructural formation, and drove the reaction towards the condensation reaction of silica alkoxides. By the same token, the iron oxide particles did not affect the microstructural formation of xerogels, which was dominated by the amorphous silica structures.
Xerogels prepared by the binary doping of iron and cobalt oxides resulted in amorphous silica with γ-Fe2O3 and Co3O4 particles. It was found that iron and cobalt oxide particles dispersed quite well in the silica matrix, though minor patches of cobalt oxide were observed. The Fe/Co ratio altered the chemical signature of the resultant silica functional groups though the variations in pore size and pore size distribution were not significant as a function of the Fe/Co ratio. Hence, this key finding clearly indicates that microporosity is still controlled by the silica instead of iron and cobalt oxides.
This confirms the postulation that the sol-gel synthesis method to the point of gelation is the key parameter in the formation of microporous structures.
The membranes were tested for single gas permeation of He, H2, N2 and CO2. The iron cobalt oxide silica (FeCoOxSi) membranes complied with activated transport, following a temperature dependent gas permeation behaviour, where the smaller kinetic diameter gases (H2 and He) permeance increased with temperature whilst the larger kinetic diameter gases (CO2 and N2) showed the opposite effect. These results are attributed to the membrane’s precise pore size control and narrow pore size distribution with an average pore size below the kinetic diameter of CO2 of 3.3 Å. Best single gas selectivities of binary metal oxide silica membranes (Fe/Co of 10/90) reached 170 (He/N2) and 141 (He/CO2).
The best FeCoOxSi membranes (Fe/Co of 10/90) also delivered excellent performance for gas mixture separation. The permeate H2 purity reached ~98% for a retentate stream of H2/Ar (70/30) concentrations. The permeation and selectivity of gas mixtures were greatly affected by the composition of the gas mixture, which in turn affected the driving force of gases. Although H2 preferentially permeated though the membranes, H2 permeances and selectivities were lower as compared to single gas values. The H2 purity was evidently a function of the driving force. The diffusion of H2 in gas mixtures complied with a temperature dependent transport, similar to that which was observed for single gas. In addition, the H2 permeance was constant and was not affected by the concentration of Ar in the feed stream. The dynamic effect of Ar diffusion in the bulk phase may have blocked pores accessible to H2 for membranes showing high selectivities. In the case of low selectivity membranes, this effect was less pronounced as slightly larger pores in the silica membrane allow for the diffusion of both H2 and Ar molecules together instead of single file diffusion.
Finally, iron oxides formed small particles (2-4 nm) whilst cobalt oxides particles were larger (10-30 nm). However, it was interesting that silica films with high iron oxide content generally cracked resulting in membranes with no separation properties. However, the membranes with Fe/Co of 10/90 content delivered excellent results. It was noteworthy to observe that upon reduction of the FeCoOxSi membrane, there was a reversible switch in gas permeation from a reduced to an oxidizing state which was more noticeable for the permeations of H2 and CO2. These results may be attributed to a binary molecular gap distribution cause by the redox effect. The first molecular gap with dimensions up to ~3.0 Å is associated with increasing H2 permeation and the reduction the smaller iron oxide. The second molecular gap with dimensions up to ~3.5 Å is related to increasing CO2 permeation and resulted from the reduction of the larger cobalt oxide. The small molecular gap was reversible, whilst the redox effect for the larger molecular gap tends to disappear due to densification of the silica cobalt oxide interface.