Properties and application of 1-5 μmPd/Ag23wt.% membranes forhydrogen separation
MetadataVis full innførsel
Thin (1-5 μm) Pd/Ag23wt.% films were produced by magnetron sputtering and subsequently removed from the sputtering substrate to create self-supported membranes. Two different reactor configurations were used; a flat configuration (2.4 cm2 membrane area) with gas flowing perpendicular to the surface, and a microchannel configuration with the gas flowing along the membrane in parallel channels (six 1 x 1 x 13 mm3 channels with 0.78 cm2 total membrane area or seventeen 0.2 x 0.2 x 13 mm3 channels with 0.44 cm2 total membrane area). The membranes were tested under different conditions, to investigate factors like thickness dependent behaviour, concentration polarisation effects, competitive adsorption by other gas species and the effect of different membrane treatments. For all membranes, permeance values were found through measurements with pure hydrogen feed and no sweep gas applied, after initial stabilization in a H2:N2 feed gas mixture. A relationship between hydrogen flux and membrane thickness was generally obtained. The permeability values were, however, not constant, and variations within sputtering batches as well as between batches were observed, thus indicating that bulk diffusion was not the sole rate-limiting step. Large concentration polarisation effects were observed in the flat membrane configuration, using H2:N2 feed gas mixtures, while they were found to have minor impact in the microchannel configuration. The self-supported membranes were found to withstand high differential pressures (>470kPa) when supported by the microchannel structure, conveniently enabling hydrogen separation without the use of a permeate side sweep gas. Several of the membranes were exposed to a thermal treatment in air. The treatment resulted in significantly improved permeation, that is, more than a doubling for the thinnest membranes. Investigations of surface topography by atomic force microscopy revealed a correlation between increased permeation and increased surface area, roughness and grain size. After treatment the permeability values were quite uniform (centred around ~2.1·10-8 mol·m·m-2·s-1·Pa-05 ), and bulk diffusion was the clearly most dominant permeation step. At the same time, the flux was found to be nearly independent on temperature after the treatment, indicating a non-activated rate-limiting step. A modelling study based on a formalism presented by Ward and Dao, taking into account all the different permeation steps in the solution-diffusion mechanism, shows that the changes due to the treatment can be modelled by a change in certain parameters like the hydrogen heat of adsorption, hydrogen heat of absorption, activation energy for diffusion and diffusion pre-exponential factor. The results were, however, far from unambiguous, due to limited information in both model and experimental data. CO was found to have significant inhibiting effect on hydrogen permeation. For instance, mixing 1 mol% CO into a feed gas consisting of 90 mol% hydrogen at 350°C resulted in a ~60% permeation decrease. CO2 was found to have minor effect at 350 C, while 10 mol% CO2 at 300 °C led to a steady decrease over time. It is suggested that competitive adsorption is the main mechanism for CO inhibition, while the inhibiting effect of CO2 is governed by slower processes like deposition and removal of strongly adsorbed species. The thermal treatment in air led to a significant reduction in CO inhibition. For instance, using 1 mol% CO at 350 °C, the reduction was now only ~15%. An approach combining a Langmuir-Fick model based on a work by Barbieri et al. and microkinetic modelling (through transition state and unity bond index-quadratic exponential potential theories) indicates that the change in inhibiting CO effect can at least partly be explained by a changes in CO and H2 heats of adsorption. No permeation decrease under CO2 exposure was observed after heat treatment in air.