Design, fabrication and characterisation of H2-selective Pd-alloy membranes for application in pre-combustion CO2 capture cycles
Climit-finansiering100 % from the Research Council
Partnere• Stiftelsen SINTEF (SINTEF) • Norges teknisk-naturvitenskapelige universitet (NTNU) • United Technologies Corporation, United Technologies Research Centre (UTRC) • Delft University of Technology (TU Delft)
Prosjektperiode2012 – 2017
Goal of project:
The pre-memCO2 project aims to research and develop improved H2-selective membrane materials for integration in pre-combustion decarbonisation process schemes, thereby targeting the priority areas in CLIMIT. The project is a collaboration between two SINTEF departments, NTNU, the Technical University in Delft (the Netherlands), and United Technology Research Centre (USA) combining membrane fabrication and performance testing with ab-initio modelling and advanced characterisation. The objective is to obtain a more fundamental understanding of permeation membrane materials, which will enable improved membrane design and fabrication, and finally more efficient and robust membranes in power generation processes.
The objectives of the pre-memCO2 project will be achieved through a multidisciplinary approach combining theoretical modelling based on first principle calculations, extensive materials characterization using high resolution microscopy and surface analytical techniques, ternary Pd-alloy manufacturing, and H2 permeation assessment efforts to develop improved membrane design and process integration.
The technology that is being investigated is palladium (Pd) membrane technology for H2 separation from syngas, a key enabling concept for pre-combustion CO2 capture. Pd-alloy membranes have been studied in membrane reactors for water-gas shift (WGS-MR) and steam reforming (SR-MR) reactions to simultaneously achieve a high CO or methane conversion and production of pure H2. A key feature of this process intensification, achieving pre-combustion decarbonisation (PCDC), is that such a membrane reactor would produce both a high pressure CO2 stream and high-purity H2 for power generation. This can greatly facilitate the economics of power generation with carbon sequestration. The technology is therefore an alternative to post-combustion decarbonisation, where more diluted CO2 is removed, by various technologies, from the flue gas after combustion. In the PCDC membrane process the CO2 stream is kept at high pressure reducing the final compression work for storage.
At the temperatures needed for direct integration in the reforming of natural gas, or other fuels like bio-ethanol or hydrocarbons in general (typically 550 – 600 C), current dense metal membrane technology is insufficient. This is because the state-of-the-art thin Pd-based membranes suffer from pinhole formation at such high temperatures, rapidly decreasing membrane life-time. As a solution, the Pd membrane thickness is increased in order to reduce the effect of pinhole formation, leading unavoidably to a decrease in H2 permeance and increase in price. Improved Pd-alloys and composite membrane structures are hence needed for the next generation of H2 separation membranes and membrane reactors. To solve this problem fundamental understanding of degradation mechanisms is required.
Results to date:
The pre-memCO2 project has combined membrane fabrication and performance testing with ab-initio modelling and advanced characterisation. Most of the work has focused on ternary Pd-Cu-Ag alloy system, relevant for the development of highly permeable sulphur-tolerant membranes. In summary, the main activities performed together with their main outcomes are listed below.
Membrane fabrication and testing
For the Pd-Cu-Ag material, a maximum value for the H2 permeability was observed at a Ag content of 1 at.% using membrane permeation (MP) experiments. Then, the H2 permeability decreased when the Ag content increased up to 7 at.%. In contrast, the H2 solubility was found to monotonously increase with Ag content and thus unit cell dimension. This thus indicated that the H2 diffusivity shows a distinct maximum. This phenomenon was further investigated by dedicated hydrogenography (HG) experiments on the same alloy system at the TU in Delft, the Netherlands. Hydrogenography is a compositional gradient thin film technique which exploits the change in optical properties as the hydrogen content in metals changes. This technique allows for sampling a large range of compositions with a small number of experiments. Crucially, the HG technique resulted in the same permeability () behaviour as function of Ag content as obtained for the MP experiments. This validated both techniques, in addition to providing a wide range of compositions; the silver concentration in Pd0.8Cu0.2-xAgx was actually varied by four orders of magnitude, from x=10-5 to 0.17. This surprisingly complicated behaviour can be explained by phenomena active at several different relevant size scales; from the upper atomic layer of the membrane (Å, x~10-5), through the grain boundary regions (nm, x~10-2) to the entire membrane (µm and beyond, x~10-1).
Microstructural changes due to long-term hydrogen permeation at moderate temperature and trans-membrane pressure difference could generally occur in Pd-based membranes. An understanding of the impact on membrane stability is crucial to design Pd-alloys for high temperature applications. Here we have continuously exposed Pd-Ag and Pd-Ag-Cu membranes to 4–5 bar H2 at 400–600 °C for 2–8 weeks, and investigated microstructural changes by high-resolution scanning transmission electron microscopy at the NORTEM facility. Compared to untested membranes, the material evidenced recrystallization and extensive growth of columnar grains presenting long axes close to <111> directions and a high density of 3 twins parallel to the membrane surface after membrane operation. No apparent segregation effects are found experimentally, but STEM analysis shows the presence of cavities with higher density along grain boundaries or in their vicinity that could be related to the growth of the <111> columnar grains. These cavities do initially not contribute to unselective leakage flux, but could be the origin of pinhole formation. For improved stability it would therefore be relevant to reduce the grain boundary diffusion. The low level of strain in the metal lattice around the spherical cavities demonstrates an internal pressure able to balance the constrictive surface tension, indicating that we are in fact in the presence of gas bubbles.
Atomic scale modelling by DFT
In order to rationalize the permeability measurements, atomic-scale models of grain boundaries were created. The fraction of atoms being present at the grain boundaries in multicrystalline membranes is in the order of 1%, thus a natural hypothesis would be that the mechanism is related to the minority element segregating towards the boundaries. A large number of coherent (Sigma) grain boundary models have been created to test this hypothesis, and the temperature dependent equilibrium concentration c_eq of the minority element has been assessed. Most of the models exhibited strong segregation tendencies with c_eq≈1 at relevant temperatures. More complicated models were then built, varying the local concentration of all alloy elements. A clear thermodynamic driving force was seen, encouraging segregation of minority elements towards one particular layer at the grain boundary while at the same time away from one or more of the neighbour layers. A small region of pure Pd may thus be created close to the grain boundaries. This region would have slightly extended lattice constants due to strain in the lattice arising from the nearby grain boundary. Then hydrogen diffusion paths parallel to and across the grain boundaries were investigated, quantifying the effect of the grain boundaries. It turned out that the effective diffusion barrier was significantly reduced in the pure Pd regions; these regions thus represent percolating hydrogen diffusion highways that can explain the anomalously high diffusivity at a silver content of x=1%. An accompanying hypothesis for the reduction in permeability when going from x=1 to 7% is that the diffusion highways are blocked by the additional silver, generating a more homogeneous composition also near the grain boundaries.
The pre-memCO2 project has so far resulted in 6 journal articles in peer-reviewed journals, 16 conference contributions, 2 book chapters, and 4 popular science presentations.