Direct air capture

Capturing carbon dioxide from the air is very challenging.¹ This is primarily because of its low concentration (~400 ppm or ~0.04%). Anyone working on traditional carbon dioxide capture, where flue gases contain ~5 - 15% carbon dioxide, will note the two to three orders of magnitude lower concentration in air. Alarm bells will probably start ringing!

Nonetheless, direct air capture of carbon dioxide is one of a number of negative emission technologies that are necessary for meeting climate change mitigation targets. Indeed, it has been described as one of the "Seven chemical separations to change the world".² 

There are three key technical challenges for direct air capture. First, as the concentration of carbon dioxide is so low, vast quantities of air must be processed to capture a meaningful amount of carbon dioxide. There is no way around this requirement. In our work, we set out to tackle the other two technical challenges, namely the energy required for concentrating carbon dioxide (the thermodynamic challenge) and the slow kinetics of any separation process operating with such a dilute feed (the kinetic challenge).

Biological membrane inspiration

Inspired by active transport in biological membranes, where the energy released during the downhill transport of one species is used to pump another species uphill, we designed a synthetic membrane that can use the energy released during the downhill transport of water to pump carbon dioxide uphill.

The membrane was fabricated by laser-drilling ~2000 'pores' into a dense alumina tube, into which we infiltrated a eutectic molten-carbonate salt. A molten-carbonate salt was chosen as our past work had shown that the perm-selectivity for carbon dioxide was sector-leading.³ Moreover, in experiments initially targeted at understanding the role of humidity in traditional carbon capture (i.e., humid feed and sweep gases) we noticed that when water was fed to the membrane in the sweep gas, carbon dioxide flux was increased quite significantly. Surprisingly, when the concentration of carbon dioxide in the feed gas was decreased, the flux remained high with a humid sweep gas. This suggested that the water driving force (in the opposite direction to the carbon dioxide flux) might be exerting influence over the carbon dioxide flux.

Perhaps the most important moment in the development of the membrane was in thinking about whether the water driving force could drive carbon dioxide permeation in the absence of a carbon dioxide driving force. Thus, we supplied the membrane with air (~0.04% carbon dioxide) to both 'sides' (i.e. in the feed gas and the sweep gas). However, we humidified the sweep gas. Incredibly, the membrane permeated carbon dioxide to the humidified side, thus removing carbon dioxide from air.

Membrane performance

The real excitement of our work came when we realised that we could employ naturally occurring humidity differences. Operated in this way, i.e., without the application of e.g., pressure to drive carbon dioxide permeation, the membrane can capture 20% of the carbon dioxide in air, using ambient energy.

Comparing the membrane with other membranes we noted that significant caution was needed. This is because the membrane we report is unique; as it pumps carbon dioxide uphill against its own concentration difference, routine methods for calculating permeability break down i.e., the permeability would be negative! Nonetheless, to achieve the magnitude (but not the direction) of carbon dioxide flux that we reported, state-of-the-art polymeric membranes would require a driving force of 4,000,000 Pa. We operated our membrane at 40 Pa!  Put another way, those polymeric membranes would require a permeability five orders of magnitude higher than they currently offer to achieve the carbon dioxide fluxes we did with a ~0.04% carbon dioxide feed gas.

At this point, the membrane has addressed both the thermodynamic and kinetic challenges. A humidity difference is used as an ambient energy input, and due to the accelerating effect of water, the membrane provides carbon dioxide fluxes at levels associated with far more concentrated feed gases and permeabilities far beyond those of state-of-the-art membranes (with the caveats noted above due to the uniqueness of the membrane).

Permeation mechanism

As detailed in the manuscript, we employed molecular DFT calculations to probe the thermodynamics of the reactions of water and carbon dioxide with the molten salt phase. From the results above, we suspected that there should be a mutual carrier for carbon dioxide and water. This is because there was a 1:1 ratio of carbon dioxide and water transport (in opposite directions) across the membrane. This, alongside several other constraints based on the experimental results, guided the DFT calculations.

Ultimately, several viable mechanistic pathways were identified. These pathways were considered viable as long as they didn't involve significantly stabilised species that might stop the permeation process. Moreover, they must not require more energy than that supplied by the humidity difference. Put simply, the pathways that were identified involved water at the sweep side of the membrane releasing carbon dioxide, and carbon dioxide at the feed side of the membrane releasing water (all in a 1:1 ratio).

Summary

Those interested in this work are invited to read the paper via the link at the top of this blog post. Across the manuscript and the 74(!) pages of Supplementary Information, we also thoroughly characterise the membrane using X-ray micro-computed tomography, operate the membrane continuously for ~50 days, provide the largest single permeation dataset associated with molten-salt membranes,  and a very significant body of DFT work on the species that occur in molten-carbonate salts in the presence of carbon dioxide and water.

We hope that our work will inspire more efforts on this exciting new class of membrane, and in reimagining how direct air capture or other dilute separation processes might be 'driven'. We are grateful for the contributions of all the co-authors, and those of our funders, the European Research Council, Royal Academy of Engineering, and the Engineering & Physical Sciences Research Council.