1-butyl-3-methylimidazolium acetate ([bmim] [ace]) and 1-butyl-3-methylimidazolium iodide ([bmim] [I]) were purchased from Sigma Aldrich and used as received. Poly (allyl amine) hydrochloride (PAlAm∙HCl) (MW: 15,000, Sigma Aldrich) was neutralized by NaOH and then filtrated through a Ultrafilter UP-10 (Advantec) for several times with pure water to remove inorganic ions and low molecular weight polymers. Concentration of PAlAm thus purified was estimated by conductometric titration to be ca.9.9%.

Hydrophobic fumed silica particle (HDK-H18, primary particle size: 5 - 30 nm), the surface OH groups of which are methylated with poly(dimethyl siloxane) by 75%, was purchased from Wacker Asahikasei Silicone Co. and used as received.

D-IL preparation was attempted by using [bmim] [ace] and the silica nanoparticles. Using the bulk liquid without added water, only a creamy or paste-like substance was obtained (Table 1). This failure in preparation of a powdery D-IL may be ascribed to relatively low surface tension (γ) of the pertinent IL; although γ of the IL is not available in literatures, γ of 1-ethyl-3-methylimidazolium acetate was reported to be ca.46 mN/m [23] . Thus, γ of [bmim] [ace] may be lower than that of [emim] [ace] because

the former alkyl group is larger and more hydrophobic than the latter. So far, only one successful preparation of powdery D-IL. Only with large amounts of water, it was successful; although the product prepared with the IL containing 50 wt% water was “souf- flé-like” to some extent, one with 60 wt% water was wholly powdery. Effects of the water addition and the resultant changes in the appearance of D-IL on the CO₂ absorption performance were examined as shown below.

CO₂ absorption was monitored by measuring CO₂ partial pressure as a function of time for two kinds of bulk ILs; [bmim] [ace] and [bmim] [I]. The latter was used because it is the only IL that provides powdery D-IL without water addition [8] . If it were also a CO₂-absorbing IL, the D-IL would be promising from a view point of the expected high absorption speed. The experimental results are shown in Figure 2 and the experimental

conditions and the respective CO₂ absorption abilities are given on Table 2. In the case of [bmim] [I], the decrement in the partial pressure was only slight and the absorption soon reached at the saturation.

On the other hand, CO₂ absorption by bulk [bmim] [ace] was significant even at low pressures as expected from literature data [24] [25] while it took so long time (ca.72 h) to reach the saturation. The respective performance on the basis of mol of CO₂ absorbed by mol of IL were 0.014 at the partial pressure p ~ 0.94 bar for [bmim] [I] and 0.32 at p ~ 0.48 bar for [bmim] [ace]. The latter value was comparable with the literature data by Yokozeki et al. [24] [25] ; 0.336 at p = 0.502 bar and t = 298.1 K. The smaller absorption (Figure 2), it may be due to the larger amount of IL contained in the D-IL compared with the bulk system (cf. Table 3). In fact, except for the initial stage of the absorption, those two profiles showed a similar tendency of pressure decrement. efficiency obtained in the present study may be safely ascribed to the lower pressure.

As a next step, CO₂ absorption performance by D-IL of [bmim] [ace] and those prepared by adding water to the IL was investigated in a similar way as the bulk IL cases. First of all, the creamy substance prepared without water addition was tested. The result was given in Figure 3 and on Table 3 together with the other ones. The creamy D-IL showed a similar performance with the bulk IL; although the initial speed of the CO₂ absorption, i.e., the initial decrement in the partial pressure was more significant than that of the bulk.

For example, at t = 50,000 s (~14 h), ca.83% of the initial amount of CO₂ was absorbed to the respective samples. The equilibrium states of CO₂ absorption seem to be

fter 2.5 - 2.6 × 10⁵ s (69 - 72 h) for both systems. Thus, CO₂ absorption speed of the creamy D-IL proved to be no superior to that of the bulk IL sample. These results suggest that the IL in the creamy D-IL formed a continuous phase without increasing the effective surface area. As for the absorption ability, the molar ratio of CO₂ absorbed to the IL was estimated to be 0.25. This value, which was obtained at p = 0.18 bar, is comparable to literature values of the bulk IL at p = 0.1 bar; 0.19 - 0.23 [24] [25] , as a matter of course.

Next we tested powdery samples to confirm the expected enhancement in the absorption speed. The results are shown in Figure 3. As described in the previous section, a powdery D-IL of [bmim] [ace] was successfully prepared only when substantial amount of water was added as much as 60%. As seen from the figure and Table 3 the CO₂ absorption speed was so rapid that the equilibrium was established just within several minutes. The slight increase in the pressure after 5000 s may be due to evaporation of water. On the contrary to the advantage in the absorption speed, the 60% water containing D-IL showed a much lower CO₂ solubility than the bulk IL, just one eighth [26] . This deterioration in the absorption ability may be caused from the existence of water. According to a proposed mechanism for the chemical absorption of CO₂ to the relevant IL [26] [27] , C2 proton of the imidazolium ring is abstracted by an acetic acid counter anion to form a site reacting with CO₂. This mechanism has in fact been supported by some researches [28] [29] , one of which also suggested that water may inhibit the C2 proton abstraction reaction [28] . Thus, in order to obtain a D-IL material combining a high CO₂ solubility and an excellent absorption speed, it should be necessary to reduce the amount of water in the inner phase of the relevant D-IL samples as much as possible with a powdery state retained. Since preparations of D-IL with [bmim] [ace] containing 50% or less of water provide only creamy substances (Table 1), we attempted to remove water from the D-IL containing 60% of water with an IR humidity meter at 40˚C and 80˚C. With this treatment, water content of the D-IL sample was reduced to 30% and 7.6% at the respective temperatures. The absorption performance of these modified samples is given on Table 3 and Figure 3. With decreasing the water content, the CO₂ solubility into the samples increased as can be judged from the almost flat level of CO₂ pressure. The mol-based absorption efficiency of the 7.6% water sample was 0.27, which is still lower than 0.32. Since the equilibrium pressure at which the former value was obtained (61.2 kPa) was higher than the latter (47.8 kPa) by ca.30%, actual difference of the solubility should be more significant. This means that only a small amount of water may affect the CO₂ absorption to [bmim] [ace]. However, the absorption efficiency value, 0.27, is never low in light of such a relatively low equilibrium pressure (0.61 bar). Further, the absorption speed was also excellent; 90% of the saturation level was absorbed within ca.20 min. Although the 7.6% water sample was not powdery but Souffle-like, the speedy absorption suggests that the aqueous IL did not form a continuous phase but was finely divided by the silica particles with the surface wetted to some extent by the liquid. Thus, the “post-dehydration” procedure seems to be effective to prepare D-ILs as a high performance material for CO₂ absorption.

Nonvolatile CO₂ absorption substances are not limited to ILs; polymers containing amino groups may also be utilized for such purpose. For a polyamine to be incorporated into a DM, the polymer must be solubilized in any liquid, ideally pure ILs. Thus, we attempted to dissolve PAlAm into [bmim] [ace] or [bmim] [I], while it proved to be in vain. Many neutral polymers are known to be soluble in ILs, while polyelectrolytes seem to be mostly insoluble in them [30] . Then, an aqueous solution of PAlAm was used to prepare a DM containing [bmim] [ace] and the polyamine. By setting the water content of the DM at 60 wt%, a Souffle-like DM was obtained.

Figure 4 and Table 4 show results of CO₂ absorption by the DM prepared above, together with those by the bulk aqueous PAlAm solution and the bulk IL + aq. PAlAm mixed solution (water content = 60 wt%). First of all, one should note the significant absorption ability of the bulk aq. solution of PAlAm as large as 0.39, a mol-based value, which is comparable with 0.32 of bulk [bmim] [ace]; if those values are converted to the mol bar-based efficiency, the former becomes 0.61 and the latter 0.67. This relatively high CO₂ absorption ability of the pertinent polymer is still lower than the theoretical or ideal value, 0.5; in principle, two free amino groups of PAlAm may react with one CO₂ molecule to form an ion-pair as the case of low molecular weight amines [31] .

In the case of polyamines like PAlAm however the above ion-pair formation serves as a physical crosslinking of the polymer chains, leading to a gel formation [32] . In fact, in the above aqueous solution of PAlAm a phase separation was observed after CO₂

*Values estimated on the basis of PAlAm mol.

absorption measurements. The post-measurement sample consisted of a turbid lower layer and a clear upper layer, which was separated by a white thin layer. By shaking the sample tube, the white layer was easily destroyed and a totally turbid liquid was obtained. If the white layer was a gel-like phase, it would retard the CO₂ absorption to some extent. In fact, after leaving the sample tube for two days, the whole lower layer became a soft gel.

A similar phase separation was also confirmed for the bulk [bmim] [ace] + aq. PAlAm system. In this case, however, the lower layer was already gel-like. Although the difference in the gel formation behavior between these two systems is not known at present, the slower CO₂ absorption of the bulk IL + PAlAm system may be ascribed to the more significant gel formation. As for the absorption efficiency, it became slightly worse than that of the bulk aq. PAlAm; if [bmim] [ace] in the mixed solution sample absorbed CO₂ to the same extent as the 60% water-containing sample (Table 3) in the mol bar-based absorption efficiency (0.044/0.967 = 0.046), the mol-based efficiency of the IL in the mixed solution sample may be estimated as 0.046 × 0.624 = 0.029. Then, a contribution from the IL to the total amount of CO₂ absorbed is estimated with this efficiency to be 2.6 × 10⁻⁴ mol. This means in turn that the contribution from PAlAm in the total absorption amount is ~2.5 × 10⁻³ mol of CO₂ and the efficiency becomes ca.0.33 mol/mol. This slightly lower efficiency, compared with 0.39 of the bulk PAlAm solution, may be ascribed to the more significant gel formation. Namely, the gel network might obstruct CO₂ to access to free amino groups. Thus, for the pertinent pressure profile given in Figure 4 to reach the true equilibrium, much more time would be necessary.

In the case of DM prepared with the IL + aq. PAlAm mixed solution, the CO₂ absorption speed was distinctly higher than those bulk systems. For example, it took ca.90 min to absorb 90% of the saturation, while it was ca.1600 min for the bulk mixed solution. The former elapsed time is, however, still much longer than that of the bulk IL system containing only 7.6% of water but no PAlAm, i.e., ca.20 min (Figure 3). This slowing down in the absorption speed, which may be due to the presence of PAlAm, may also be attributed to the gel formation as compared in the bulk systems. As for the absorption efficiency, however, the gel formation in the DM does not seem obstruct the CO₂ absorption; the mol-based absorption efficiency, 0.58, though it was estimated as a value per amino group of PAlAm, is higher than 0.5, the limiting value for the CO₂ absorption by PAlAm. This means that the IL also contributed to the significant CO₂ absorption to a substantial degree, although the respective contributions may not be separately estimated in the present study. Here we should note why [bmim] [ace] could absorb CO₂ to a significant degree irrespective of the coexistence of water. As discussed previously, the abstraction of the C2 proton of the IL cation by the counter anion (CH₃COO⁻) should be interfered by water. The observed high level of CO₂ absorption suggests that this interference mechanism by water is not well working. As a probable mechanism for that, we postulate that amino groups of PAlAm are involved in the C2 proton abstraction. In fact, basicity of a primary amine is much stronger than that of -COO⁻; basicity measured by pK_b values of ethyl amine is larger than that of acetate anion by a factor of ca.10⁶. Then, a next question is why the abstraction by the polymer in the bulk mixed solution system did not appear to contribute to an appreciable degree. This may be safely ascribed to the observed macroscopic phase separation. Because the polymer was mostly confined to the lower layer of the sample, the amino groups would be hardly available for the IL left in the upper layer. On the other hand, in the case of the DM system each DM particle contains only a tiny amount of liquid and has a relatively large surface area. For a typical DM having a diameter of 100 μm [8] , the surface/volume ratio is as large as ca.10⁶ times of the bulk. Thus, it may be reasonable to assume that the gelation in the DM system prevailed through the tiny sample without leaving a separate PAlAm layer. A non-monotonic change in the CO₂ pressure observed for the DM sample around at t = 10⁴ s may be corresponding to this rather rapid development of gelation through each droplet. A “transition” to a second “stationary” state found around at t = 1.3 × 10⁵ s (36 h) indicates that the CO₂ absorption process was not completed but just temporarily stopped by the gel formation. As stated above, PAlAm gel formed by ionic crosslinking according to Equation (1), where Rs should be replaced by polymer chains, seemed to be never robust. Thus, the gel layer, which must be much thinner than that formed for the bulk system, would be easily, at least partly, broken under a certain CO₂ pressure.