CO2 capture by dry alkanolamines and an efficient microwave regeneration process

Removal of acidic gases such as H2S and CO2 is performed during the purification of raw natural gas, most commonly using amine gas treatment. However, this industrially entrenched method is limited by significant shortcomings including low operational capture efficiency, amine pipeline corrosion and a large energy penalty due to the sorbent regeneration process. To address these shortcomings, we have studied the use of perfluorinated silica-stabilized dry alkanolamines (DAf) for CO2 capture. Due to their micronized liquid domains, DAf display high operational CO2 capture efficiency. Further, to minimize energy requirements for sorbent regeneration, microwave-assisted regeneration of the spent DAf sorbent was also studied and shown to decrease the energy requirements by about ten times. In contrast to very recent work, our results show that the use of DAf exhibits extraordinary recyclability, with a negligible decrease in absorption capacity over at least ten absorption–regeneration cycles, indicating the potential of this material for gas treatment applications.


Introduction
Carbon dioxide capture is crucial not only for mitigating greenhouse gas emissions but also for natural gas processing. [1][2][3][4] The recent United States shale gas boom has precipitated an increase in the production of natural gas, making the need for effective CO 2 capture technologies more pressing than ever. 5 Examples of solid sorbents for CO 2 capture include metalorganic frameworks, 6-8 porous polymers 9-12 and zeolites. 13 Such sorbents have the advantage of a high surface area to accelerate CO 2 uptake. 14 Nevertheless, liquid sorbents, which oen require agitation to achieve high gas-liquid interfacial areas, are still predominant in industrial processes. In fact, amongst various CO 2 capture technologies, the most established is amine gas treatment, which relies on the use of alkanolamines to separate acidic gases such as H 2 S and CO 2 from raw natural gas. 15 Despite amine-based gas sweetening being an already wellestablished industrial process, it faces multiple shortcomings including low operational CO 2 capture efficiency, amine corrosivity and a signicant energy penalty due to the sorbent regeneration process. 1, 16 Although primary and secondary alkanolamines react with CO 2 in a 2 : 1 mole ratio to form their corresponding carbamates with a theoretical CO 2 absorption capacity of 50 mol% (Scheme 1), [17][18][19] industrial systems utilizing alkanolamines typically only operate at 15 to 30 mol% capture efficiency. 20 In addition, the high viscosity and corrosivity of alkanolamines necessitate their dilution with water to approximately 30 wt% of amines. 6 This leads to an associated energy penalty during sorbent regeneration, whereby CO 2 -loaded alkanolamine sorbents are heated to regenerate the amines, since the additional water content increases the overall heat capacity of the sorbent. To address this problem, researchers have turned to utilizing ionic liquid-alkanolamine blends 21 or to amine-functionalized porous solids for CO 2 capture. [22][23][24] Dawson et al. recently reported an alternative approach whereby they used hydrophobic silica-stabilized dry alkanolamines (DA h ) for CO 2 capture. 25 However, the DA h exhibited low recyclability and substantial sorbent loss during regeneration, as well as signicantly decreased CO 2 uptake aer the rst cycle. Nevertheless, this approach circumvents the need for dilution of the sorbent with water, thereby decreasing the energy penalty incurred during sorbent regeneration. Amine-related corrosion problems are also potentially reduced because the amine is isolated by the silica particles from pipelines.
We herein report the use of dry, undiluted alkanolamines for CO 2 capture whereby instead of hydrocarbon-functionalized, hydrophobic fumed silica, peruoroalkyl-functionalized oleophobic fumed silica was utilized for dry alkanolamine, DA, formation. These ndings were previously outlined in our patent led in May 2013, 26 prior to the work being reported by Dawson et al. We refer to this DA as DA f . In contrast to the ndings by Dawson et al., 25 we found that DA f showed excellent recyclability and stability. Studies were conducted on monoethanolamine (MEA) and diethanolamine (DEA) due to their widespread industrial usage. Furthermore, a novel method using microwave-assisted heating for DA f regeneration was also investigated as an energy-efficient alternative to the conventional (convective and conductive heating) regeneration method.
DAs are powdered materials and can be thought of as inverted foams, i.e. liquid droplets dispersed in air, whereby microscopic liquid droplets stabilized by nanoparticles form the dispersed phase and air is the continuous matrix ( Fig. 1). They appear as a free-owing powder and are analogous in form to dry water (DW) which has received signicant attention in recent years. [27][28][29][30][31] For dry liquid formation to occur, the stabilizing particles must be hydrophobic towards the encapsulated liquid, otherwise a foam or paste may ensue. 27 Generally, for a at surface to be hydrophobic towards a liquid (contact angle > 90 ), the solid surface tension g S must be g S ( g L /4, where g L is the surface tension of the liquid. 32 It has recently been demonstrated that the formation of dry oils with liquids of decreasing surface tension requires the use of nanoparticles with increasing degrees of uorination on their surfaces. 33 Based on the surface tensions of MEA and DEA (Table S1 †), it can be deduced that oleophobic peruoroalkyl-functionalized nanoparticles are preferable over oleophilic hydrocarbon-functionalized nanoparticles for stabilization of DAs (Fig. S1 †).

Methods
Synthesis of peruoroalkyl-functionalized oleophobic SiO 2 nanoparticles. 10 g of fumed SiO 2 was added to a well-mixed solution of 220 mL ethanol, 11 mL of 28% aqueous NH 3 and 2 mL of 1H,1H,2H,2H-peruorooctyltriethoxysilane, which was then homogenized (using an IKA T18 basic Ultra Turrax digital homogenizer) for 30 min until a homogeneous suspension was obtained. The resulting dispersion was stirred at 400 AE 50 rpm at room temperature overnight, aer which the dispersion was heated on a hot plate to evaporate the solvents and subsequently dried in an oven at 80 C for 1-2 days to collect the dry peruorinated silica powder. Thermogravimetric analysis (TGA) was performed using a TGA Q500 v6.7 Build 203 instrument, from room temperature to 800 C at a rate of 20 C min À1 with 40 mL min À1 of nitrogen gas as the balance gas and 60 mL min À1 of air as the sample gas. TGA of peruorinated silica versus unmodied fumed silica particles showed that the per-uorocarbon groups comprised approximately 35 wt% of the functionalized silica (Fig. S2 †). Dynamic Light Scattering (DLS) studies were performed with a Brookhaven ZetaPlus Zeta Potential Analyzer to determine the size distribution of unmodied fumed silica particles versus peruorinated fumed silica particles aer dispersing them in ethanol. DLS studies showed that the average aggregate size of peruorinated silica particles was smaller than that of unmodied fumed silica particles (282 nm vs. 367 nm, see Table S2 †). Preparation of dry alkanolamines. 5 g of peruorinated SiO 2 nanoparticles was placed in a plastic beaker while 5 mL of the selected alkanolamine (pre-heated for 1 h at 100 C to minimize the water content) was slowly added into the beaker and blended with the particles using a simple coffee whisk ( Fig. S3a †) at a rate of 500-1800 rpm for 1 min or until the formation of a free-owing powder.
Absorption of CO 2 into bulk or dry alkanolamines. The reaction was performed at room temperature and pressure. 5 mL of alkanolamine or approximately 10 g of DA f powder (containing 5 mL of alkanolamine) was added into a roundbottomed ask placed in a water bath to modulate the heating effects arising from the exothermic nature of the reaction. CO 2 gas was bubbled at a ow rate of 500 AE 50 mL min À1 through the ask with the outlet connected to a gas bubbler until the absorption prole plateaued. At short time intervals of 30 s or 1 min, the ask was taken out of the water bath, dried and weighed on a weighing balance to record the increase in mass. The maximum CO 2 uptake by alkanolamines (in the absence of water) was calculated by assuming a reaction mole ratio of 2 : 1 alkanolamine to CO 2 .
Regeneration of sorbents by conventional hotplate heating. The CO 2 -loaded alkanolamine or DA f powder was regenerated by bubbling argon gas at a ow rate of 500 AE 50 mL min À1 through the sorbent while the ask was heated in an oil bath at 120 C to release the absorbed CO 2 . The ask was equipped with a reux condenser maintained at 15 C to minimize the evaporative loss of the sorbents and the outlet was connected to a gas bubbler.
Regeneration of sorbents by dielectric heating. The CO 2loaded DA f powder was added into an open Teon cup. The cup was placed in a holder and subsequently placed in a programmable microwave synthesizer at 2.45 GHz. The maximum power of the synthesizer was set at 500 W. No purge gas was used. To maintain the stripping temperature of 120 C, a thermocouple was inserted into a separate Teon cup containing unloadedsorbent which provided feedback to the soware to control the microwave power. In addition, stirring of the sample within the cup was carried out as part of the microwave heating process. The microwave synthesizer was used to heat the CO 2 -loaded sorbent at 120 C for approximately 1 h until no further mass loss occurred, aer which time the cup was removed and weighed. The decrease in weight was attributed to the CO 2 loss (typical mass decrease for diethanolamine: 0.08-0.1 g g À1 DA f or 0.16-0.2 g g À1 bulk DEA).

Results and discussion
Brunauer-Emmett-Teller (BET) analysis performed on per-uorinated SiO 2 particles showed that CO 2 adsorption on these particles at room temperature was negligible (Fig. 2), because 1 g of peruorinated SiO 2 nanoparticles absorbed approximately 6 mg of CO 2 at atmospheric pressure while 1 g of DA f (containing 0.48 g of peruorinated silica) absorbed approximately 100 mg of CO 2 . Of this 100 mg, about 3% is attributed to adsorption by silica. Hence, the mass increase is largely due to CO 2 uptake by the alkanolamine.
Carbon dioxide absorption results for both DA f and bulk liquid are shown in Fig. 3. The theoretical maximum alkanolamine CO 2 uptake (in the absence of water) was calculated by assuming a reaction mole ratio of 2 : 1 alkanolamine to CO 2 . Hence, 1 g MEA theoretically absorbs about 0.36 g CO 2 at the maximum, and 1 g DEA theoretically absorbs 0.21 g CO 2 . As shown in Fig. 3, CO 2 absorption by dry and bulk MEA in the rst 10 min of CO 2 exposure reached 98 and 60 wt% of the theoretical maximum absorption, respectively. Similarly, dry and bulk DEA achieved about 90 and 35 wt% of the theoretical maximum within 20 min. Mass changes due to the replacement of air in the reaction vessel headspace with CO 2 were taken into account in the calculations.
For both amines, the DA f showed signicantly higher CO 2 absorption than their corresponding bulk liquid counterparts (Fig. 3). The lower absolute absorption of the bulk amines is due to mass transport effects arising from the large increase in the liquid viscosity caused by carbamate salt formation, which was veried by measuring the viscosities of the alkanolamine liquids before and aer CO 2 absorption (Table S3 †). The effect of viscosity of a liquid on the mass transfer kinetics is described by the Stokes-Einstein equation: 34 In this equation, D ij is the diffusion coefficient of solute i in pure solvent j, k is the Boltzmann constant, T is the temperature, n SE is the Stokes-Einstein number, h j is the solvent viscosity, R i is the radius of the solute molecule and R j is the radius of the solvent molecule. According to eqn (1), the diffusion coefficient of CO 2 gas in the amine sorbent decreases with increasing solvent viscosity. Further, given that the system studied is stationary and unagitated, the increased viscosity also signicantly affects the replenishing of the unreacted amine near the gas-liquid interface, which is dependent upon the movement of alkanolamine molecules towards the interface via convection currents or diffusion.
However, due to the small domain sizes within DA f (approximate diameter of a microscopic liquid marble is 80 AE 30 mm, as measured by cryo-SEM), the required distance of CO 2 diffusion within the droplets to contact the unreacted alkanolamine is signicantly less than for the bulk liquid. This helps to explain why the DA f achieve a much larger gas absorption relative to the theoretical capacity within the timeframe of the experiment than do the bulk amines. In addition, the reason for the slightly higher absorption capacity of dry MEA compared to dry DEA is due to the inherently lower viscosity of the former compared to the latter.
Based on the work of Wang et al., 31 who postulated that the higher methane gas hydrate formation rate in dry versus bulk water was due to the higher surface area to volume ratio in dry water, and taking the view that the rate of gas uptake is dependent upon the gas-liquid interfacial area, we expect that the CO 2 absorption rate by DA f would follow a similar trend (estimated surface area to volume ratio in DA f is 75 000 m À1 based on the volume of alkanolamine present). Nevertheless, the gas-liquid interfacial area of the microscopic liquid droplets is reduced by the presence of the silica shell around each droplet; CO 2 must diffuse across this porous shell to react with the encapsulated alkanolamines.
Our CO 2 uptake studies showed that for MEA, the absorption rate by the bulk liquid in the rst minute was faster than that by dry MEA (Fig. S4 †). This is presumably because the lower liquid viscosity of MEA compared with DEA means that CO 2 mass transfer effects do not outweigh the effect of the silica barrier initially. However, as carbamate formation proceeds, the viscosity of the liquid mixture increases (Table S3 †), and mass transfer effects become more important. The rate of CO 2 uptake therefore becomes faster for dry over bulk MEA. For DEA, dry DEA exhibits faster CO 2 absorption than bulk DEA throughout the entire absorption cycle (Fig. S4 †), because the mass transfer rate in the more viscous DEA liquid was much slower and hence the effect of higher surface area to volume ratio of dry DEA was more prominent.
Since the carbamate salts decompose in the temperature range of 100-150 C, 35,36 the CO 2 -loaded bulk alkanolamines were heated to 120 C to release the absorbed CO 2 gas and regenerate the alkanolamines. The decrease in mass was attributed to both CO 2 and H 2 O release, which was veried by thermogravimetric mass spectroscopy (TG-MS) analysis (see Fig. 4).
For each sample of bulk alkanolamine and DA f , three cycles of absorption and desorption of CO 2 were performed (Fig. 5). As shown in Fig. 5, DEA appears to be more recyclable than MEA, with a negligible change in CO 2 uptake capacity over three cycles. This is because secondary amines like DEA form weaker bonds with CO 2 than primary amines like MEA, allowing for easier regeneration of the free amines. 37 This is supported by the ndings of McCann et al., where the enthalpy of carbamate formation for MEA and DEA was found to be À29.7 kJ mol À1 and À23.7 kJ mol À1 , respectively. 38 Moreover, MEA has a lower boiling point than DEA (170 C vs. 271 C) and a much higher vapour pressure, leading to a greater degree of sorbent loss per cycle.
As heating is essential for the sorbent regeneration process, it is imperative to maximize the heat transfer efficiency within the amine sorbents. This is dependent on the three main mechanisms of heat transfer, namely conduction, convection and radiation. In the case of DA f , these three mechanisms happen to be hindered by the silica shell encasing the micronized alkanolamine droplets and the air pockets throughout the dry liquids. For heat transfer via conduction, fumed silica particles exhibit low thermal conductivity due to the nanoscale pores among them in their fractal arrangement, 39,40 causing the silica shell to block effective heat conduction from the heat source to the droplets. Importantly, many air pockets between the microscopic particle-coated liquid marbles in DA f also limit heat conductivity. With regard to convection, the liquid in DA f exists as dispersed microscopic droplets, thereby limiting convective heat transfer. As for radiative heat transfer, Taylan and Berberoglu have shown that the silica shell in dry water greatly attenuates infrared radiative heat transfer due to its high single-scattering albedo, i.e. a signicantly larger proportion of heat radiation is scattered compared to that absorbed by the silica shell, 41 thus preventing infrared radiative heat transfer from the heat source to the droplets.
Since the three major forms of heat transfer are signicantly inhibited for DA f , we turned to microwave heating as an alternative method to improve the efficiency of heat transfer during the regeneration process. This method is dependent on the ability of the irradiated material, e.g. a solvent, to absorb microwave energy and convert it to heat, which is quantied by a dielectric parameter called the loss factor tan d. This loss factor is expressed as: 42 where 3 00 is the dielectric loss, which is indicative of the efficiency with which electromagnetic radiation is converted into heat, and 3 0 is the dielectric constant describing the ability of molecules to be polarized by the electric eld. A solvent with a high tan d value of more than 0.5 is required for efficient microwave absorption. DEA, with a high tan d value of $0.6 at 2.45 GHz at 278 K can be rapidly heated to temperatures >100 C within a minute when irradiated under microwave conditions. 43 Owing to the higher recyclability of DEA over MEA, we chose to study DEA for microwave heating-based regeneration. Samples of approximately 10 g of dry DEA containing 5 mL of liquid DEA were loaded with CO 2 until the absorption plateaued, and the CO 2 -loaded samples were subsequently heated in a programmable microwave synthesizer at 120 C for approximately 1 h until no further mass loss occurred. To test the recyclability of the sorbent and conrm that the observed mass loss was correctly attributed to the CO 2 loss, 10 repeated cycles of CO 2 absorption and removal were carried out, as shown in Fig. 6. The level of CO 2 uptake by dry DEA was found to be stable over at least 10 cycles, showing an excellent recyclability of dry DEA. Similar regeneration results were obtained using a simple household microwave instrument, although the heating periods were limited to 10-20 s in order to avoid overheating.
Regeneration of DA f via microwave heating was found to be more efficient than conventional hotplate heating under the same conditions (whereby no purge gas was used). Conventional hotplate heating took more than 3 h and consumed about 0.38 kW h per cycle on average. Microwave heating on the other hand took about 1 h and consumed about 0.034 kW h per cycle on average, which is less than 10% of the energy consumption by conventional hotplate heating.
Finally, the recyclability of the peruorinated silica particles was investigated. Alkanolamines typically undergo degradation over multiple absorption and regeneration cycles aer which they must be replaced. 44 Nevertheless, the peruorinated silica particles could be separated from the amines by dispersion of DA f in ethanol and then centrifuged, and re-used in at least three subsequent batches of DA f . This leads to signicant costsavings since only the alkanolamines need to be replaced.

Conclusions
In conclusion, the dry alkanolamines described herein demonstrate signicantly enhanced performance with regard to CO 2 absorption capacity and sorbent recyclability. Furthermore, the novel dielectric heating technique for sorbent regeneration proposed here has proven to possess the benets of reduced energy consumption and minimal use of purge gas. The per-uorinated silica particles can also be recycled and re-used to prepare DA f with a negligible change in their wettability and overall quality for at least three cycles. Moreover, as suggested by the recent work of Dawson et al., 25 the amine component of the DA f is also isolated by the silica particles from pipelines and other corrosion sensitive materials, leading to further reduction in maintenance costs associated with the addition of corrosion inhibitors and repair of damage to pipelines. Last but not least, this novel method of using DA f is not restricted to MEA and DEA; in principle any dry liquid can be formed provided that nanoparticles of appropriate surface energy are selected. 33 The implications are that this idea can be extended to other liquid CO 2 sorbents, possibly leading to a similar improvement in the absorption rate and capacity. However, there still exist several challenges for the application of DA f industrially. Firstly, the powdery nature of DA f will likely require signicant changes in the current process technology and facilities specialized in handling liquid amine sorbents. Secondly, due to the nanoscale primary particle size of fumed silica, careful connement of the powder needs to be put in place to prevent contamination in the outlet gas stream. Thirdly, stripper columns need to be modi-ed to incorporate dielectric heating capabilities. Nevertheless, given the multiple advantages of utilizing DA f for CO 2 absorption and desorption, we believe that they should still be considered promising candidates for practical applications in CO 2 capture.