Carbon dioxide (CO₂) has been shown to be the major contributing gas to global warming and detrimental climate change (IPCC, 2014). Carbon capture utilization and storage (CCUS) is one of the key frameworks being developed to counter this threat. Without further development and subsequent deployment of CCUS and associated technologies, models indicate that warming will not be limited to the previous 2°C target, let alone 1.5°C target contained in the COP21 Paris Agreement (IPCC, 2014; United Nations, 2015).

Within CCUS, once the CO₂ is captured, it may be either sequestered into geological storage or utilized as a feedstock in other products through CO₂ utilization. Both approaches require significant energy input. Either way, both methods first require a CO₂ capture step, although in the utilization processes, this may be a reactive capture directly to a product. One of the key challenges facing CCUS remains the question of which approach to use in order to capture dilute CO₂ from exhaust gases when over 90% of stationary emitters produce emissions concentrations of 15% by volume of CO₂ or even less (Gale et al., 2005).

Removal and capture of CO₂ from these point sources is an area of intense research. Even when only post-combustion removal of CO₂ is considered, approaches range from adsorption and absorption, membrane and cryogenic separation, and microbial or algal biosystems (Figure 1) (Chou and Chen, 2004; Figueroa et al., 2008; Stolaroff et al., 2008; Brennecke and Gurkan, 2010; Karadas et al., 2010). Of the overarching technologies listed in Figure 1, adsorption and absorption systems are the most advanced and are themselves split into two main types; temperature swing absorption (TSA) and pressure swing adsorption (PSA). TSA processes typically rely on high temperatures in order to regenerate a solvent that readily captures CO₂ liberating a high-concentration CO₂ stream, whereas PSA processes rely on a drop in pressure, most usually from near-atmosphere to vacuum to release the more weakly bound CO₂ (MacDowell et al., 2010; Samanta and Zhao, 2011).

The aforementioned choice between the capture approaches will likely be decided upon relative economic viability of the processes. This hinges on their energy requirements, prospective plant footprint area, and the balance of capital and operating costs. In turn, these economic elements will depend on both the nature of the process and the sorbent taken together. For example, a low-cost sorbent that has exorbitant energy requirements for CO₂ capture and that will require a large plant footprint may struggle to compete with a far higher-cost sorbent that has lower energy requirements and smaller footprint. Location also needs to be considered. If there is surplus low-grade heat available, this can be used in the TSA process to afford desorption of the captured CO₂. However, there are many potential emission sources where surplus heat is not available. In these cases, the PSA process using renewable energy becomes more economically efficient as the heating duty on solvent regeneration by fat outweighs compression duty.

Ionic liquids (ILs) and particularly room temperature ionic liquids (RTILs) are particularly expensive sorbents. Some of these show very promising uptake performance for CO₂ capture such as high selectivity and inherent flexibility, allowing task-specific sorbents to be synthesized (Bates et al., 2002; Ramdin et al., 2012; Moen and Stene, 2014). These RTILs can also bridge both the physisorption and chemisorption classes of sorbent, potentially allowing both high uptake capacity and low regeneration costs in typically TSA processes.

One of the main drawbacks when using RTILs is the very slow rate of CO₂ diffusion through the bulk liquid. Compared with benchmark amines, the diffusion rates of gas through the RTILs is up to 19 orders of magnitude slower (Jassim et al., 2007; Moya et al., 2014). This largely prevents rapid CO₂ sorption and cycling and, therefore, increases the total IL inventory required to treat a given flow rate or volume of gas. Compounding this problem is the high cost of the RTILs. However, in some cases, this might be erroneous as only a few cases of cost modeling have been carried out for IL production. In addition, these are frequently estimates that are based on laboratory-scale syntheses or small-scale production, which can easily imply costs approaching £1 million per tonne (Baltus et al., 2005). A recent full techno-economic analysis of larger scale syntheses have given more likely IL prices ranging from £1,700 to £35,000 per tonne, based on manufacturers estimates but with a strong dependence on the IL structure (Klein-Marcuschamer, 2011). For comparison, the benchmark amine, MEA, has a market cost of £1,100 per tonne. Naturally, it should be noted that the wide range of results is due to scaling assumptions (Chen et al., 2014).

However, what is more certain is that even if the lower-cost figure for the production of ILs is the more accurate, the slow diffusion rate will require infeasibly large inventory supplies to be available at any point source emitter.

One potential solution to the twin problems of low-gas diffusion rates and high sorbent costs may be to use SoILs as sorbents. While these may appear to be a contradictory name, ILs are typically defined as salts with melting points below 120–140°C, allowing organic salts that have high melting points to be entirely solid at room temperature.

Previously published work (Styring et al., 2016) by this group has investigated SoILs and molecular organic salts as an alternative IL approach to CO₂ capture, where PSA is used on finely ground ILs with melting points above room temperature. These solids are often structurally simpler and cheaper to synthesize than RTILs and naturally avoid gas diffusion issues by allowing the gas immediate access to a large IL surface area. They can also potentially avoid sorbent pumping issues as fine powders rather than typically viscous fluids and are significantly easier to dry (Styring et al., 2016).

Through this work, it was found that the high selectivity to CO₂ over nitrogen, which is typical of RTILs, is maintained in the SoILs. However, in contrast with the RTIL capture methods that typically use TSA, PSA is required to counteract the fact that typically crystalline and impermeable SoILs will not allow full gas–solid contacting. As a result, since not all of the IL is available for gas–solid interaction, uptake capacity is limited compared to RTILs. By applying a high pressure swing process, where adsorption and separation occurs at high pressures (5–30 bar) and desorption occurs at atmospheric pressure, uptake capacities can be boosted. Once again, this method reduces the inventory of IL required to treat a given gas stream. An additional benefit of using a high pressure swing process is a potentially reduced capture plant footprint made up of less complex unit operations than those found in a TSA process, and operation at ambient temperature. Overall, there is a small temperature change of less than 1°C over the compression–decompression cycles. Furthermore, it should be noted that if the captured CO₂ is to be transported by pipeline, large-scale compression and pumping equipment will already be present (Pershad et al., 2010).

These positive effects come at the price of the energy costs of pressurization. Costs have been estimated for a model pressure swing capture system using a range of operating pressures and separation performances based on experimental data. Costs estimated by this model compare well with literature examples of high pressure swing systems for CO₂ separation. Table 1 compares model results with the most prominent separation technologies shown Figure 1. This table shows that when considering total energy costs, high pressure swing processes have a decided advantage over other separation technologies.

A frequent comment on this observation is that these overall energy comparisons can be misleading when different energy sources are used for each of the various capture methods (Wilson et al., 2016). For example, in a power plant context, temperature swing processes (MEA and amines primarily) could in principle be powered by low grade and low value heat. This would render the overall energy cost less relevant, especially if assuming this energy might otherwise be wasted. In comparison, the costs of gas compression or vacuum generation, while representing far less total energy, could have a larger associated energy penalty to the power plant if the assumption is made that higher-value electricity is required. This position naturally ignores the possibility that the same low-grade heat that putatively powers MEA capture could be used to power pressurization equipment, which fundamentally requires only mechanical motion and does not necessarily require electricity. Nor does it allow for situations where such waste low-grade heat is unavailable or insufficient for MEA capture to proceed without large economic costs. In such situations, overall energy costs of each approach, as shown in Table 1, is a fair basis of comparison. If renewable energy is used for the pressure swing then the situation is further improved.

The action of supporting ILs has previously been used in applications such as supported IL catalysis, a concept that combines the advantages of ILs with those of heterogeneous support materials; and gas separation, where ILs are supported on membranes to improve permeability (Mehnert, 2004; Ilconich et al., 2007). Supporting SoILs to specifically aid CO₂ capture is a relatively unexplored area of research and may allow significantly improved CO₂ uptake per unit mass of IL in the total sorbent. This will further reduce the quantity of IL needed to treat a given gas stream. This would primarily be achieved by coating a thin layer of the SoIL on a support particle with which it has good adhesion and full surface coverage. Assuming the support material is low cost, this in turn may substantially reduce sorbent inventory costs. Further benefits may include unique combinations of properties from both the IL and the support itself.

Supports such as activated carbon, alumina, and cellulose among others, could be considered as viable options. Of these, there is a history of using ILs with cellulose, typically in efforts to separate cellulose from biomaterials as cellulose is known to dissolve well in RTILs (Swatloski et al., 2002). It would, therefore, seem reasonable to assume that SoILs could have a good surface interaction and adhesion to the cellulose particles, allowing the desired thin layer to be deposited, giving total coverage of the support particle. The insolubility of cellulose in other solvents allows easy recovery of any supported SoIL. Furthermore, cellulose is also very cheap, is an abundant bio-renewable material and has only weak interaction with CO₂. The tough fibrous structure of the cellulose support also leads to high mechanical strength, preventing the sorbent particles from breaking up under pressure swing conditions.

As with previous work (Styring et al., 2016), the cations and anions selected for the SoILs focused on low complexity cations with abundant precursor materials. Those chosen for cellulose support were tetraalkylammonium, pyridinium, and imidazolium cations. For the anions, acetate and bromide were chosen. RTILs using acetate have a track-record of high uptake capacity. Bromides, which typically have very low uptake capacities, have a facile synthetic pathway and were chosen to see if supporting them on cellulose would help improve sorption capacity (Shiflett and Yokozeki, 2008). The use of a cellulose support lowers the cost of the materials relative to a pure SoIL while also reducing the inventory of the SoIL. This also improves uptake capacity so has a double benefit. The materials are more expensive than activated carbons but comparable or cheaper than advanced solids such as molecular sieves/zeolites. We have chosen low-cost SoIL materials with market values close to pure MEA for example. These are not highly complex ILs needing exotic anions and cations.

All reagents were purchased at highest available purity from Sigma-Aldrich and used without further purification, other than drying. Amine quaternization reactions were carried out using Schlenk-line techniques under an inert (N₂) atmosphere. Reactions using methyl iodide were carried out using aluminum foil protection on the exterior of the reaction vessel and low lighting. All solvents were HPLC grade. CO₂ and N₂ and were supplied by BOC-Linde. 1-Butyl-4-methylimidazolium bromide ([Bmim][Br]) and tetra-octylammonium bromide ([N₈₈₈₈][Br]) were directly purchased from Sigma-Aldrich and thoroughly dried under a flow of 20 mL/min nitrogen at 60°C for 24 h before uptake tests were carried out.

High pressure adsorption experiments were carried out using a bespoke packed-bed adsorption column constructed from Swagelok™ piping and fittings (Figures 2 and 3) using a Jasco PU-1580-CO₂ supercritical CO₂ pump, a Jasco BP-1580-81 back pressure regulator, an Omega PX409USB High Accuracy Pressure Transducer, a 42AAV48 Midwest Pressure Systems Gas Pressure Booster, and an AND GF-1000 High Capacity 3 decimal place balance. The reactor was isolated from the system using valves and the assembly weighed on the balance. Desorption was measured by slowly opening the valves while still on the balance. Supported sorbent packed densities were measured using a Micromeritics AccuPyc 1340 Pycnometer and surface area analysis was carried out on a Micromeritics 3Flex gas sorption analyzer. High pressure adsorption capacities were further verified using a Hiden Isochema IGA-0002 adsorption apparatus with a pressure range of 0–10 bar.

Further testing was carried out to determine whether the trends seen for [N₄₄₄₄][Ac] hold true for other SoILs. [N₁₈₈₈][Ac], [Bmim][Br], and [N₈₈₈₈][Br] were chosen since they are structurally simple, low in cost, have readily available starting materials, and are solids at room temperature. Previous tests with [N₈₈₈₈][Br] have shown it to be particularly poor as a CO₂ capture agent. It was selected in order to indicate whether the increase in CO₂ capacity was proportional to the starting performance of the pure ionic liquid. If so, this would indicate that the improvement seen after cellulose support was simply down to an increase in available surface area of the IL for CO₂ capture and not due to surface effects potentially suggested by SEM images.

[Bmim][Br] had a fairly poor starting capacity when using it in its pure form as it forms large crystals on synthesis and has a high affinity to water making it challenging to handle as fine powder. The resulting large crystals lead to a poor surface area to volume ratio, which was expected to be the reason for its poor uptake performance. However, once supported, [Bmim][Br] showed a similar increase in its CO₂ uptake capacity at 75 wt% loading as was found for [N₄₄₄₄][Ac]. Following the pattern of the [N₄₄₄₄][Ac] (Figure 8), 70 and 80 wt% loadings showed poorer total uptake. The 75 wt% cellulose support loading gave an increase of a factor of 2.3 over pure [Bmim][Br], which amounts to an increase of CO₂ adsorbed per unit mass of IL of 9.4, nearly an order of magnitude improvement in the effectiveness of the IL.

Similarly, the 75 wt% cellulose sorbent shows a tripling of surface area by BET measurement (Table 2). It should also be noted that it is possible in this case that there is a sharp optimum peak loading for [Bmim][Br] in the 70–75 or 75–80 wt% range that may afford even further improvements in uptake performance. After loading, [Bmim][Br], which is highly hygroscopic, also showed a reduction in its water affinity and remained as a free-flowing powder during handling. This could also be attributed to changes in the surface properties, however, may be due to cellulose acting as a desiccant.

[N₁₈₈₈][Ac] has the highest uptake capacity of the pure ILs that have been previously investigated by this group (Styring et al., 2016). In contrast, [N₈₈₈₈][Br], as previously mentioned, has one of the lowest uptake capacities measured. However, as shown in Figures 9 and 10, respectively, [N₁₈₈₈][Ac] showed only very modest improvement at 75 wt% loadings and loadings higher or lower than this amount actually gave a diminished performance. By contrast, [N₈₈₈₈][Br] showed an order of magnitude improvement in total CO₂ capacity, with the highest capacity seen at 70 wt% cellulose loading. Furthermore, [N₁₈₈₈][Ac] exhibited a much larger increase in surface area, of around 40%, when compared to [N₈₈₈₈][Br] after cellulose loading (Table 2). However, since its capacity improvement was only marginal, surface area was unlikely to be the driving factor in this case. This perhaps indicates that the surface effects suggested by the SEM images on [N₄₄₄₄][Ac] may be important and, therefore, indicates that surface properties of [N₁₈₈₈][Ac] were already well-suited for CO₂ adsorption. Therefore, the addition of the cellulose support had a negative or near-neutral effect at best. However, it still shows a total improvement, after the loading factor is considered, of a factor of 4.2 times better than the pure SoIL.

On the other hand, the improvement of [N₈₈₈₈][Br] indicates the opposite and that the unsupported salt surface morphology seems likely to be hostile to CO₂ adsorption, despite its demonstrated capability to perform once supported. This is backed up by the surface area measurements shown in Table 3 where [N₈₈₈₈][Br] only exhibited a small increase in surface area, yet the increase in performance was large. This further suggests that surface properties play a major role in these cases.

It can be assumed that while the cellulose increases the surface area of the total sorbent, albeit with varying effect, the improvement in CO₂ capacity after cellulose loading is neither proportional to the surface area increase, nor is it proportional to the pure IL uptake and is, in fact, more likely due to the increase in surface roughness imparted by the support. Overall, as cellulose loading increases, the total sorbent uses less and less ionic liquid, meaning the SoIL is used much more efficiently. The previously inaccessible centers of the SoIL particles are now instead spread over the cellulose allowing it to capture more CO₂ per unit of SoIL used. In addition to what appears to be significantly improved surface effects, increased effectiveness at different cellulose loadings can be seen in Table 3: depending on the nature of the IL in question. Table 3 shows the amount CO₂ captured per gram of SoIL within the total sorbent at different cellulose loadings.

It is clear that further advancements in CO₂ capture systems are needed in the near future to develop CCUS in order to help mitigate the potential problems caused by further carbon emissions. Existing state-of-the-art CO₂ capture and separation technologies suffer from high energy costs, primarily associated with the CO₂ desorption inherent in chemisorption-based TSA processes. Alternative PSA processes show many advantages in terms of both cycle speed, energy costs, and sorbent stability, due to the absence of thermal cycling. Therefore, PSA coupled with appropriate solid adsorbents would appear to be a promising alternative gas separation pathway. SoILs show good capacity and excellent adsorption rate, making them desirable for use in a PSA process. However, currently, even simple ILs have a high manufacturing cost, which reduces their attractiveness as large-scale carbon capture sorbents.

Significant reductions in amount of IL required to capture CO₂ have been demonstrated by coating inexpensive cellulose particles with thin layers of IL. In addition, surface characteristics imparted to the IL by the cellulose supports have further increased CO₂ uptake capacity. Together, these two effects may allow even expensive, specialized, and previously overlooked IL sorbents to be used more effectively. Fast sorption kinetics are observed, which results in large capture capacity per unit time when compared to conventional TSA systems, which exhibit significant temperature lag between adsorb and desorb cycles. Further research must be done to show their suitability as a post-combustion capture sorbents.

These studies were fundamental in nature, carried out with pure CO₂ and pure nitrogen. The results of these studies will be used to monitor performance of different gas mixtures ranging from blast furnace top gases at high CO₂ concentrations through to flue gas from power stations down to the parts per million concentrations of atmospheric CO₂. By determining the fundamental data, we are now in a position to test these on low concentration simulated and real waste gas streams.

Each author played a major role in the development of the work and writing the manuscript. PS is the research supervisor and group leader. DGR is a postgraduate research student who undertook the work as part of his Ph.D. programme. GRMD is a postdoctoral research associate who was involved in all aspects of experimentation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.