Correction: Carbon dioxide uptake from natural gas by binary ionic liquid–water mixtures

Correction for ‘Carbon dioxide uptake from natural gas by binary ionic liquid–water mixtures’ by Kris Anderson et al., Green Chem., 2015, DOI: 10.1039/c5gc00720h.

However, it should be noted that the majority of the research on carbon dioxide solubility in ionic liquids to date has, curiously, neglected a major issue related to the way in which uptake is reported. For physical absorbers, carbon dioxide uptake is typically described in terms of Henry's constant, Equation (1).
Here, p is the partial pressure of carbon dioxide, and c is concentration, typically expressed as mole fraction. When Henry's constant is expressed as a mole fraction, however, the molecular weight of the ionic liquid is not accounted for. As most of the ionic liquids extensively studied consist of relatively large organic cations paired with high molecular weight anions (often containing fluorine, oxygen, and/or sulfur atoms), the molecular weight of one ionic liquid "ion pair" is often much, much larger than the molecular weight of carbon dioxide. For example, one readily available ionic liquid is trihexyl(tetradecyl)phosphonium bis[(trifluoromethyl)sulfonyl]amide, [P 6 6 6 14 ][NTf 2 ], with a molecular weight of 764.01 a.m.u., while the molecular weight of carbon dioxide is 44.01 a.m.u. Therefore, high uptake in terms of molar fraction corresponds to a much lower uptake in terms of weight fraction (approximately 17 times lower). From an industrial perspective, efficient uptake in terms of solvent weight or volume is far more important than uptake in terms of mole fraction; the important empirical ratio is the weight of carbon dioxide absorbed per litre of solvent. Most ionic liquid studies have reported uptake in terms of Henry's constant using mole fractions. This has inadvertently led to artificially inflated views on the carbon dioxide capacity of many ionic liquids.
To counteract the above problems, in this work it was decided to determine carbon dioxide solubility in terms of weight and volume of solvent. This allows a more industrially relevant measure for determining the relative value of ionic liquids as gas uptake solvents. In addition, ionic liquid design was redirected towards physically smaller cations paired with low molecular weight anions. In this way, carbon dioxide uptake may be maximised in terms of solvent weight or volume.

Chemisorption with Ionic Liquids
In 2002, Davis Jr. and co-workers proposed the concept of a "task-specific" ionic liquid, now preferably known as functionalised ionic liquids. 39 Initially, they incorporated an amine functionality onto the terminal end of the side chain of an imidazolium cation. Coupled with the tetrafluoroborate anion, this generated the functionalised ionic liquid shown in Scheme 1. This ionic liquid was shown to result in the chemisorption of carbon dioxide at a 1:2 CO 2 :IL ratio in accordance with Scheme 1. 40 Scheme 1. Reaction between original functionalised ionic liquid and carbon dioxide. 40 Since these landmark publications on carbon dioxide absorption, there has been a flood of derivative publications, many of which have been reviewed elsewhere. [41][42][43] Although these papers largely play a variation on a known theme, many contained extremely reliable and valuable data on the physisorption of carbon dioxide in a wide range of ionic liquids. 42,44,45 It rapidly became clear from the reported Henry's constants for CO 2 in a wide range of ionic liquids, from about 0.1 to 0.25 (mol CO 2 ) (kg IL) -1 at 5 bar, 42 that a system based on physisorption alone would never be commercially viable, but that the systems based on chemisorption were constrained by a low molar uptake. Work from Notre Dame recognises this, resulting in new functionalised ionic liquid systems based on amino acids, which doubled the carbon dioxide uptake by operating on 1:1 stoicheiometry (Scheme 2, where [P 6 6 6 14 ] + is the trihexyl(tetradecyl)phosphonium cation). In this, and many other papers, the authors stressed the importance of using dry ionic liquids. 46,47 Scheme 2. Reaction between a functionalised amino acid anion and carbon dioxide. 48 Here, we report the development of a new acid gas stripping solvent for removal of carbon dioxide from natural gas sources, based on ionic liquids with carboxylate anions, as reported in recent patents. 49,50 Unlike previous literature studies, the water content of these ionic liquids is carefully controlled at near equimolar amounts. The formulated ionic liquid/water mixtures show greatly enhanced carbon dioxide solubility relative to both anhydrous ionic liquids and aqueous ionic liquid solutions.

Equipment and Uptake Experiments Ionic Liquid Selection
An important design factor that may be incorporated is dependent on ionic liquid basicity, as the mechanisms of carbon dioxide chemisorption involve nucleophilic attack. Typically, to generate strongly basic ionic liquids, a basic functional group has been appended to the side chain of the quaternising alkyl chain: see, for example, Scheme 1. 40 However, a facile consideration of electron density suggests that placing the basic group on the cation will reduce its basicity, while placing the basic group on the anion should enhance basicity. Bases are, more or less by definition, species with regions of high electron density. It makes little sense to append an electron-rich functional group to an inherently electron-poor species. To quantify this subjective argument, computational studies were performed on a cation and anion with basic functionalisation, using Spartan '10 with density functional theory (DFT) at the B3LYP level, and a 6-31G* basis set. 51 Figure 4 illustrates electrostatic potential maps for the [2-[2-(diisopropylamino)ethoxy]ethyl]-(ethyl)dimethylammonium cation (top) and the ethanoate anion (bottom). Electron density increases from the extremes of positive blue to negative red. On the cation, the lone pair on the nitrogen atom on the far right (green) is significantly less basic than the oxygen atoms (red) on the ethanoate anion. Hence, our ionic liquid design strategy was directed towards placing the nucleophilic site on the anion, rather than the cation. In addition, candidate cations and anions were also selected with an eye towards reducing molecular size and mass. All ionic liquids in this study have been synthesised from aqueous or methanolic solutions containing the cation of interest paired with the methyl carbonate, hydroxide, or hydrogencarbonate anions. In all cases, the conjugate acid of the anion of interest, typically a carboxylic acid or an amino acid, is added slowly to the aqueous or methanolic solution in sub-equivalent amounts (normally up to 80% of the required amount). The solution was stirred with mild heating, resulting in reaction to form water (hydroxide and hydrogencarbonate), methanol (methyl carbonate), or carbon dioxide (hydrogencarbonate and methyl carbonate). Where carbon dioxide is produced, greater heating and longer reaction times were required, as the carbon dioxide buffers the solution and must be released. The reaction is allowed to proceed until carbon dioxide is no longer generated (in the case of hydroxide solutions, the reaction shows no visible indication of completion but is quite rapid). The pH of the reaction mixture was then monitored as additional acid was added to the equivalence point of that particular species. Hydrogencarbonate and some hydroxide solutions were purchased from chemical supply companies, while most hydroxide and methyl carbonates were generated in-house using previously published methodologies. 52,53 Some of the tributylmethylphosphonium methyl carbonate in methanol was received as a gift from Cytec Canada.

Anhydrous Carboxylate Ionic Liquids
The absorption of carbon dioxide by ionic liquids has recently been reviewed. 43 The authors concluded that "the overall tendency is that adding a certain amount of water into the pure ionic liquids will slightly decrease the absorption capacity of CO 2 , but the lower viscosity of the mixture is favourable as it can promote gas diffusion in the solvent and reduce operational constrains on the use of ionic liquids fluids". 43 There have been many studies published on carboxylates, and especially on the [C n mim][CH 3 CO 2 ]-CO 2 (n = 2 or 4) system, 45,46,[54][55][56][57][58][59][60][61][62][63] but the focus has largely been upon a possible interaction between the carbon dioxide and the anion; two groups have noted a positive effect of the presence of water, but ascribed it to a lower viscosity of the system. 62,64 Of the multitude of papers on carbon dioxide solubility in ionic liquids, those that mention water content, with very few exceptions, state that the ionic liquids must be as dry as possible. 46,47 When considering physical absorbers, this is a sensible precaution. Although carbon dioxide is significantly more soluble in water than most other gases, 65 solubility is still quite low (Henry's constant on the order of 1640 bar) relative to "good" physical solvents such as methanol (on the order of 100 bar), used in the Rectisol process. 66 Thus, addition of water is expected to effectively dilute the uptake solvent and reduce solubility. To establish the viability of carboxylate anions for carbon dioxide uptake, a series of ionic liquids containing carboxylate anions of varying alkyl chain length and branching, paired with the trihexyl(tetradecyl)phosphonium cation, were prepared. This initial series was generated because the chloride form of the cation is readily available, and easily converted to the hydroxide using the reported method. 53 The solubility of carbon dioxide in this ionic liquid series, at 25 °C and 5 bar, is presented in Table 1. This reveals excellent uptake in terms of Henry's constant (19-32 bar), with reasonably good uptake in both weight and volume terms -an uptake of 0.30-0.58 (mol CO 2 ) (kg IL) -1 . Buoyed up by these results, it was decided to determine if altering the cation size would significantly change the carbon dioxide uptake in terms of weight and volume. These studies included symmetrical cations: Table 2 shows an uptake of 0.36-0.58 (mol CO 2 ) (kg IL) -1 for the methanoate anion, and 0.39-0.50 (mol CO 2 ) (kg IL) -1 for the ethanoate anion. The effect of cation chain length is detailed in Table 3 for the alkyltrioctylphosphonium cation series, showing an uptake of 0.28-0.48 (mol CO 2 ) (kg IL) -1 for the methanoate anion, and 0.39-0.43 (mol CO 2 ) (kg IL) -1 for the ethanoate anion. Finally, Table 4 shows an uptake of 0.37-0.45 (mol CO 2 ) (kg IL) -1 for the methanoate anion, 0.35-0.39 (mol CO 2 ) (kg IL) -1 for the ethanoate anion, for the alkyltributylphosphonium cation series. Curiously, changes in cation size, even quite large changes, showed little effect on the CO 2 uptake. In addition, the changes in uptake across series that did occur were inconsistent and difficult to explain in physical terms. As these ionic liquids all show at least a degree of hygroscopicity, it was thought that small variations in adventitious water content (due to subtle handling differences prior to uptake experiments) may have been responsible. This was one rationale for the water content studies outlined later.

Ionic Liquids based on Amino Acids
Alkaline metal amino acid salts are the primary component of an industrial CO 2 absorption system known as the DECAB process. 67 Thus, in a parallel study to the carboxylates above, ionic liquids with anions based on amino acids were explored as a potentially useful set of chemical absorbers. 68,69  Indeed, the tetrabutylphosphonium salts of amino acids prepared here show excellent carbon dioxide uptake (with the exceptions of the two amino acids with tertiary amine groups), achieving approximately 1:1 molar absorption (Table 5). Unfortunately, as pure, anhydrous compounds, these ionic liquids solidify or form thick pastes upon CO 2 uptake, rendering them unsuitable for real world applications.

Effect of Water on Carbon Dioxide Uptake
The solubility of carbon dioxide in water as a function of temperature (see Figure 5) and pressure is well known. 70 It can be seen that the solubilities in water are three orders of magnitude lower than those in simple ionic liquids (see Tables 1-5). It is perhaps not surprising, therefore, that the literature has emphasised the need for the ionic liquids to be dry. Figure 5. The temperature-dependent solubility of carbon dioxide in water at atmospheric pressure. 70 Our initial observations of carbon dioxide solubility in "dry" carboxylate ionic liquids showed a scatter over and above that expected for the experimental design. This suggested that it would be prudent to determine the effect of water on carbon dioxide uptake, as these ionic liquids are particularly hygroscopic. In addition, both raw natural gas and flue gas are saturated with water, and hence (during any industrial process) it would be expected that ionic liquids would absorb water if used as CO 2 absorbers. To determine the effect of low levels of water on these carboxylate ionic liquids, initially noted on exposure to ambient air, two ionic liquids were chosen for uptake trials. The liquid trihexyl(tetradecyl)phosphonium methanoate, [P 6 Figure 6 displays the carbon dioxide solubility of each ionic liquid:water mixture studied, resulting in curves that increase with increasing water content until a maximum is reached, followed by a reduction in uptake in the case of [P 4 4 4 4 ][HCO 2 ]. In each case, carbon dioxide solubility is at a maximum at a 1:1 molar ratio of ionic liquid:water, suggesting that the water is bound to the anion with a structure similar to that illustrated in Figure 7. Interestingly, and significantly, Kazarian and coworkers found that 1-alkyl-3-methylimidazolium ionic liquids with eight varied and different anions, X -(including [CF 3 CO 2 ] -, but no other carboxylates), formed Scheme 3. Proposed reactions between a tetraalkylphosphonium carboxylate ionic liquid, water, and carbon dioxide. "H 2 CO 3 " is a convenient shorthand, and makes no assumption about the correct speciation of this moiety, which may be simply hydrated CO 2 ; whichever extreme is nearest to the correct formulation, it does not alter the overall equilibria described.
To determine if this was indeed the reaction taking place, 13 C{ 1 H} nuclear magnetic resonance studies of water-formulated [P 4 4 4 4 ][CH 3 CO 2 ], before and after exposure to carbon dioxide, were undertaken, with comparison to aqueous sodium hydrogencarbonate solutions, as depicted in Figure 9. Figure 9 focusses on the 13 C chemical shift range between 150 and 215 p.p.m., and so does not reveal the observed peaks due to dissolved CO 2 (δ C = 125 p.p.m.), the phosphonium cation (δ C = 10-35 p.p.m.), 74 or the CH 3 resonance of the ethanoate group; it does encompass the resonances of hydrogencarbonate and the carboxylate group of ethanoate. Figure 9(a) defines the 13 C chemical shift of the carboxylate function of ethanoate in water, at δ C = 176.5 p.p.m., and Figure 9(d) shows the same mixture saturated with carbon dioxide, with a peak around 160 ppm. Figure 9(c) defines the 13 C chemical shift of the hydrogencarbonate anion in water, at δ C = 161.5 p.p.m, very close to the new peak in Figure 9(d). The 13 C NMR spectrum, Figure 9(b), of an aqueous mixture of [P 4 4 4 4 ][CH 3 CO 2 ] and sodium hydrogencarbonate reveals that the chemical shifts of both the carboxylate and the hydrogencarbonate anions move to lower field, suggesting that they are involved in a dynamic equilibrium such as that illustrated in Scheme 4. .  The 13 C{ 1 H} NMR results (Figure 9) suggest that Scheme 4 accurately represents the reaction with carbon dioxide taking place in this series of ionic liquids. This also explains the reduction of carbon dioxide uptake as the water content is increased past the 50 mol% ratio. When it is added in greater than equimolar amounts, water simply acts as a diluent, decreasing the uptake.

Crystal structure of [P 4 4 4 4 ][HCO 3 ]
Single crystals of [P 4 4 4 4 ][HCO 3 ] were harvested from a bulk wet sample of [P 4 4 4 4 ][HCO 2 ] under atmospheric air. First, the solid methanoate salt became liquid by absorption of atmospheric moisture, and after several weeks, abundant crystals had formed. Crystals of the methanoate salt were not found: only the hydrogencarbonate salt crystallised. It should be noted that the sample from which single crystals of tetrabutylphosphonium hydrogencarbonate grew had not been used for CO 2 absorption experiments (i.e. they were never in contact with pure commercial CO 2 ), and hence, the origin of the hydrogencarbonate anions forming the crystalline material must have been atmospheric CO 2 . Crystal growth was also attempted under a dinitrogen atmosphere from either pure or hydrated [P 4 4 4 4 ][HCO 2 ], but no signs of crystallisation were observed under such conditions. The fact that only tetrabutylphosphonium hydrogencarbonate crystals grew from wet tetrabutylphosphonium methanoate under atmospheric air confirms the high affinity for, and capacity of, phosphonium carboxylate salts for CO 2 in the presence of water.
Crystal data for [P 4 4 4 4 ][HCO 3 ] were collected using a Bruker kappaCCD diffractometer with graphite monochromated Mo-Κ α radiation (λ = 0.71073 Å) at ca. 150 K in a dinitrogen stream. Lorentz and polarisation corrections were applied. The structure was solved by direct methods and the non-hydrogen atoms were refined with anisotropic thermal parameters, with the exceptions of the disordered carbon atoms in one of the butyl chains. Hydrogen-atom positions were added at idealised positions and a riding model with fixed thermal parameters (U ij = 1.2U eq for the atom to which they are bonded (1.5 for methyl)), was used for subsequent refinements. The function minimised was Σ[w(|F 0 | 2 -|F c | 2 )] with reflection weights w −1 = [σ 2 |F 0 | 2 + (g 1 P) 2 + (g 2 P)] where P = [max|F 0 | 2 + 2|F c | 2 ]/3. The SHELXTL package 76 CO 2 ] led to the growth of a crystalline material (large plates) in the bulk viscous liquid medium after several days under atmospheric air. A sample of that material was successfully analysed by single crystal X-ray diffraction (data not presented here). Surprisingly, the actual composition of the crystals was [P 4 4 4 4 ][H(CH 3 CO 2 ) 2 ], where an ethanoate anion forms an adduct with an ethanoic acid molecule. This is in accord with Scheme 3, where the liberated ethanoic acid is reacting with the excess of ethanoate to form the stable protonated dimeric anion. This structure has been previously described, although no details on crystal growth were given. 78 It should be also noted that the crystal structure of tetramethylphosphonium hydrogencarbonate  82 have been crystallographically studied. Moreover, the crystal structure of the cyclic 1,1-dimethylpyrrolidinium hydrogencarbonate ([C 1 C 1 pyrr][HCO 3 ]) has also been reported, 83 and very recently the anhydrous salt 1-ethyl-3,4,5trimethylimidazolium hydrogencarbonate was determined. 84 The asymmetric unit of [P 4 4 4 4 ][HCO 3 ] consists of one hydrogencarbonate dimer and two tetrabutylphosphonium cations. In the refined structure, disorder is found for one of the butyl chains in one of the cations over two positions, with a 50:50 occupancy factor ratio. When comparing both disordered butyl chains, it is observed that one of them is in all-trans conformation, whereas the second one is in a distorted gauche-trans conformation as revealed by its torsion angles (90.1 and 176.1 ° for P-C α -C β -C γ and C α -C β -C γ -C δ , respectively). On the other hand, the average P-C bond distance in the [P 4 4 4 4 ][HCO 3 ] structure is 1.80 ± 0.01 Å, in good agreement with known quaternary phosphonium salts, 1.80 ± 0.02 Å. [NOTE: Searches on the CSD for any phosphonium cations where the phosphorus atoms is bonded to four carbon atoms; 4453 hits found.] Furthermore, the arrangement around the central phosphorus atom in [P 4 4 4 4 ][HCO 3 ] is tetrahedral (average C-P-C angle = 109.6 ± 3.6 °), with a slight deviation due to the distorted geometry of the disordered butyl chains. The hydrogencarbonate dimers in the crystal structure of [P 4 4 4 4 ][HCO 3 ] are surrounded by five [P 4 4 4 4 ] + cations. A number of short interionic contacts have been found between hydrogen atoms in the cations and anion oxygen atoms, being the mean C-H···O distance 2.49±0.17 Å. Such contacts, although considerably longer than those for the strong hydrogen bonds responsible for holding the [HCO 3 ] − dimers together (vide supra), are still shorter than the sum of the Van der Waals radii (2.70 Å). 85,86 Therefore, in addition to the electrostatic attraction between tetrabutylphosphonium cations and hydrogencarbonate anions, the sum of a number of weak cation-anion hydrogen bonds should also be significant in determining the solid state structure of [P 4 83 are listed in Table 6. In all cases, a number of weak hydrogen bonds around the anion dimers was found. a Distances averaged for all contacts shorter than the sum of Van der Waals radii. b Van der Waals radii taken from ref. 85,86 . c Distances averaged for all contacts shorter than the sum of Van der Waals radii plus 1.2 Å (arbitrary limit, see text). d Closest approach of an oxygen atom to a phosphorus or nitrogen atom (E = P, N). e Data retrieved from the CSD, referred to by their six-letter code. f #UGUKAI, ref. 79 . g #LUQFUZ, ref. 83 .
In  83 are also assembled as dimers.
Further analysis was focussed on the interionic E···O (E = P or N) distances (see Table 6). It was observed that the oxygen atoms not taking part in the formation of the hydrogencarbonate dimers are, in general, closer to the central phosphorus (or nitrogen) atoms, whereas several other contacts exist involving all oxygen atoms. Due to the ubiquity of such contacts, statistical analyses become ambiguous. For that reason, only the contacts for the first sphere of coordination were considered (arbitrary limit set below the sum of Van der Waals radii plus 1.2 Å). The resulting average E···O distances were marginally shorter for the smaller and less hindered tetramethylphosphonium cation than for the tetrabutyl analogue, although within a wide distribution (see Table 6). Minimum E···O distances are more informative about the degree of cation-anion interaction. The minimum P···O distance in [P 4 4 4 4 ][HCO 3 ] is ca. 3.42 Å, slightly longer than that found in [P 1 1 1 1 ][HCO 3 ] (3.36 Å), owing to the lower steric encumbrance of the cation in the latter. Both are slightly longer than the sum of Van der Waals radii. This, in combination with the tetrahedral arrangement around the central phosphorus, is indicative of very weak P-O interactions. Notably, the minimum N···O distance in [C 1 C 1 pyrr][HCO 3 ] is longer (3.54 Å) than P···O for the phosphonium salts, and considerably longer than the sum of Van der Waals radii, indicating a clearly decreased anion-cation interaction in the ammonium (pyrrolidinium) salt. This is in agreement with our previous study on halide structures, 74 where it was shown that the smaller size of nitrogen is responsible for a more crowded environment around the central atom in ammonium cations as compared to phosphonium. Therefore, the more open structure around the phosphorus allows a deeper penetration of the anion through the protective "shrubbery" of the alkyl groups. This trend is confirmed here for hydrogencarbonates, thus demonstrating the validity of the reasoning for a wider range of anions.

Regeneration experiments
A comparative study of the recyclability towards CO 2 capture of some selected ionic liquid:water (for [P 4 4 4 4 ][HCO 2 ] or [P 2 2 2 8 ][CH 3 CO 2 ]) systems and the commercially used aqueous MEA absorber has been carried out. These systems were chosen for their high capacity to dissolve CO 2 . As can be seen in Figure 11, the [P 4 4 4 4 ][HCO 2 ] system recovers essentially 100% of its CO 2 capacity at as low as 70 °C, over two consecutive recycles. The [P 2 2 2 8 ][CH 3 CO 2 ] system recovers ca. 90% of its capacity by recycling at 70 °C. Finally, the industrially used chemical absorbent (MEA/H 2 O) only recovers ca. 82% at temperatures as high as 120 °C, using the same experimental set-up. These results prove that the regeneration conditions for the tetraalkylphosphonium carboxylate ionic liquid systems for CO 2 absorption might be much less energetically demanding than those for aqueous amine systems. In addition, the [P 4 4 4 n ] + cations (and, indeed, their hexyl and octyl analogues) have been shown, by us, elsewhere [87][88][89] to be thermally and chemically stable under much harsher than those used here. These carboxylate anions, of course, are well recognised to be stable in aqueous conditions below 100 ºC. As a general trend, they may also indicate that methanoate systems might be easier to regenerate as compared to ethanoate ones, which would be consistent with the lower basicity of the former anion. These conditions need to be optimised in order to determine the practical recycling conditions for other compositions of ionic liquid absorbents. In addition, if water is to be removed in the same stripping unit, the corresponding temperatures should be further determined. Figure 11. CO 2 solubility data on fresh and recycled liquid absorbents, both ionic liquid:water (1:1) and MEA:water solutions. All recycles were done by sparging a dry gas (N 2 or air) through the stirred liquids for 15 min. Water content was determined by Karl-Fischer titration after the stripping step, and corrected in accordance before subsequent CO 2 absorption measurements. Mole ratio of the recycled absorbents might have an associated error of ±0.3.

Selection of the Best System
On establishment of the reaction mechanism, several carboxylate ionic liquids were generated to improve this system. As the ionic liquids based on the tetrabutylphosphonium cation were either solid at room temperature when dry, or produced solids on exposure to carbon dioxide, it was decided to use the lower symmetry tributylmethylphosphonium cation for further studies. The methanoate, ethanoate, propanoate, 2-methylpropanoate, butanoate, hexanoate, octanoate, lactate, or glycolate anions, paired with either the tributylmethylphosphonium or tributylmethylammonium cation, formed the initial ionic liquids synthesised and formulated with 50 mol% water. Initial carbon dioxide uptake was determined at five bar CO 2 pressure and 25 °C, followed by a second study at 40 °C. The most common commercial chemical absorber (MEA in water) and physical absorber (Genosorb ® ) were trialled in the same manner, as both a test of the methodology and to generate a realistic baseline. Tables 7 and 8 present the results of these initial single point experiments. Under these conditions, the commercial chemical absorber shows the best performance (30 wt% MEA in water, 3.58 mol l -1 ), followed closely by tributylmethylphosphonium ethanoate, propanoate, and 2-methylpropanoate. The pK a values of the conjugate acids of the ethanoate, propanoate, and 2-methylpropanoate anions (4.76, 4.87, and 4.86) are quite close to each other, while significantly higher than the conjugate acid of the methanoate anion (3.75). Thus, it would be expected that these anions would show similar basicity to each other, while acting as stronger bases than the methanoate. High base strength is expected to enhance the reactions shown in Scheme 4, as can be observed in Table 7. Table 7. CO 2 solubility for equimolar tributylmethylphosphonium carboxylate-water systems at 5 bar.  In the case of the tributylmethylammonium cation, the ethanoate immediately forms a solid. The results show that the propanoate, butanoate, and 2-methylpropanoate salts exhibit the highest carbon dioxide solubility, whereas the methanoate and octanoate show significantly lower uptake. The ammonium based ionic liquids, shown in Table 8, had a tendency to form solids on carbon dioxide exposure, and were generated from a less thermally stable cation than the phosphonium series; they were therefore not considered further.
The ionic liquids derived from amino acids (see Table 5) were also mixed with varying levels of water, in a bid to establish carbon dioxide uptake behaviour as a function of water content (Table 9). Here, it can be seen that addition of water results in a similar improvement in carbon dioxide uptake, with up to 1.5 moles of carbon dioxide absorbed per mole of ionic liquid, although the alaninate ionic liquid showed a reduction in CO 2 uptake. Addition of water did not prevent solidification, however, resulting in these systems being abandoned. Brennecke and coworkers have also reported similar results with these systems. 48 The ionic liquids tributylmethylphosphonium ethanoate and tributylmethylphosphonium 2-methylpropanoate show some solid formation on exposure to carbon dioxide, leaving tributylmethylphosphonium propanoate as the preferred choice for further studies. As this research is industrially oriented, it was decided to further study both the pressure and temperature dependence of carbon dioxide solubility in the formulated ionic liquid. In this particular case, the water content was increased to 60 mol%, partly to ensure enough water was available for reaction, and partly to allow for water uptake from the inlet gas stream. Inlet temperatures of both raw natural gas and flue gas are expected to be 40 °C and above, while regeneration temperatures may be as high as 150 °C. In natural gas processing, inlet pressures may be in the range of 50 to 70 bar, while carbon dioxide content may be 2 vol% or higher (typical sales gas contains less than 2 vol% CO 2 ). Thus, carbon dioxide partial pressures range from ca. 1 bar to 20 bar or higher (depending on the quality of the natural gas field). To establish a reasonable operational range of the formulated ionic liquid, it was decided to test carbon dioxide pressures between about 0.5 bar and 20 bar and temperatures between 40 °C and 150 °C. The results of this study are included in Figure 12, where some striking features may be seen. At 40 °C, the carbon dioxide uptake is consistent with chemical absorbers up to pressures between 8 and 10 bar, where uptake is equivalent to ca. 1 mol CO 2 per mol IL (discounting the moles of water). At this point, the mixture continues to behave in the manner of a good physical absorber, showing an uptake that is equivalent to 2 mol CO 2 per mol IL at 20 bar partial pressure carbon dioxide. As the temperature is increased, the chemical absorption becomes less prominent until the formulated ionic liquid behaves principally as a weak physical absorber at 150 °C. This indicates that regeneration by temperature-pressure swing is feasible for these systems. 56,90 Figure 12. Carbon dioxide uptake vs. pressure for tributylmethylphosphonium propanoate (0.6 mole fraction water) at six temperatures.

Conclusions
We have described a series of ionic liquids based on carboxylate anions formulated with water with a high chemical absorption (up to a 1:1 mole ratio of carbon dioxide:IL), followed by high physical absorption. The basicity of the ionic liquid anions activates the reaction between water and carbon dioxide to form the hydrogencarbonate anion and the conjugate acid of the anion (Scheme 4), which appears to exist as a complex within the ionic liquid solution. The resulting mixture of ionic liquid cation, water, hydrogencarbonate anion, and carboxylic acid behaves as a strong physical absorber for carbon dioxide, resulting in CO 2 uptake that is comparable to commercial chemical absorbers at relatively low carbon dioxide partial pressure, while exceeding them at higher pressures. Regeneration methodology is indicated by the dependence of carbon dioxide uptake on temperature and pressure. These studies show the importance of careful control of water content for these systems. A detailed engineering analysis for this project was performed by Worley-Parsons, and the main points are summarised in Table 10. It was found that, relative to a traditional absorber tower utilising activated methyl diethanolamine (aMDEA is MDEA (see Figure 1), containing up to 10 wt% of corrosive and volatile piperazine), 91,92 tributylmethylphosphonium propanoate (non-corrosive and involatile) with 60 mol% water required 5% less CAPEX and 22% less OPEX. Although costs vary unpredictably with time, this system is undoubtedly the best known based on ionic liquids, and is competitive with the best of the amine systems. There is much flexibility for further optimisation of this system (improvements in viscosity will lead to even better economics), which will be particularly attractive for implementation in areas of the world where the use of amines is sensitive or impractical.