A detailed study of cholinium chloride and levulinic acid deep eutectic solvent system for CO2 capture via experimental and molecular simulation approaches

Choline chloride + levulinic acid deep eutectic solvent is studied as a suitable material for CO2 capturing purposes. The most relevant physicochemical properties of this solvent are reported together with the CO2 solubility as a function of temperature. The corrosivity of this solvent is studied showing better performance than amine-based solvents. A theoretical study using both density functional theory and molecular dynamics approaches is carried out to analyze the properties of this fluid from the nanoscopic viewpoint, and their relationship with the macroscopic behavior of the system and its ability for CO2 capturing. The behavior of the liquid–gas interface is also studied and its role on the CO2 absorption mechanism is analyzed. The reported combined experimental and theoretical approach leads to a complete picture of the behavior of this new sorbent with regard to CO2, which together with its low cost, and the suitable environmental and toxicological properties of this solvent, lead to a promising candidate for CO2 capturing technological applications.


Introduction
Carbon dioxide capture is one of the most relevant technological, 1 environmental, 2 social 3 and economic 4 needs nowadays. The effect of anthropogenic CO 2 atmospheric emissions on global warming has been widely proven, 5,6 which in all the considered scenarios will lead to increasing temperatures at a global level, 7 with large economic, social and environmental consequences. 8,9 The most important source of anthropogenic CO 2 emissions rises from the combustion of fossil fuels for transportation, power generation and industrial purposes. [10][11][12] In particular, electricity production from fossil fuelled power plants is responsible for a large percentage of total anthropogenic CO 2 emissions, 13,14 and thus, reducing these emissions has a pivotal role for controlling climate change. In spite of the remarkable advances in renewable electricity production approaches, the state-of-the-art of these methods is not an alternative to fossil fuels, and thus, all the projections show that power plants based on fossil fuels will develop a pivotal role for energy production in the coming decades. 15,16 Therefore, a realistic approach for reducing CO 2 emissions in a reasonable time frame is to develop suitable technologies that allow for the capturing of CO 2 from flue gases in fossil fuelled power plants. [17][18][19][20][21][22] The state-of-the-art of CO 2 capture technologies does not allow carbon capture without an unacceptable increase in the costs of electricity generation, [23][24][25] which hinders their application for the required large scale. The most common approach for carbon capture is post-combustion sorption based on alkanolamine liquid sorbents, 26,27 which although showing large affinity for CO 2 molecules has serious technological drawbacks, such as facilitating corrosion, solvent evaporation or degradation, and large capturing costs. 23,[28][29][30][31][32][33][34][35] Therefore, alternative CO 2 sorbents have been studied in the last few years such as metal-organic frameworks, 36 solid sorbents, 37 carbon-based materials, 38 or membranes. 39 One alternative that has attracted great attention are ionic liquids, IL, [40][41][42] because of the possibility of tuning their properties, and thus their affinity for CO 2 molecules, through the selection of suitable ions. Nevertheless, ILs have also showed some problems with regard to their application for carbon capture purposes, such as their high viscosity or their cost, which have hindered their application for capturing purposes at the industrial scale. [43][44][45][46][47][48] Therefore, although these drawbacks should not be considered as a motive for discarding ILs as possible CO 2 capturing candidates, [49][50][51][52] several alternatives have been proposed to maintain the design flexibility of ILs while avoiding their well-known problems. Deep eutectic solvents (DES) are among the most promising options to overcome these IL problems while retaining their positive properties and the ability of tuning their behavior. 53,54 DESs are usually a binary mixture of a salt (or IL) with an hydrogen bond donor (HBD), which at a certain molar ratio leads to a system with a melting point close to ambient temperature and lower than those for the pure compounds of which the DES is comprised. 53,54 Although several salts have been proposed for developing DESs, most of the available literature is based on cholinium chloride ([CH][Cl]) ( Fig. 1). 54 [CH][Cl] is a nontoxic and biodegradable compound, and it can be obtained at very low cost. 55,56 [CH][Cl] may lead to a DES when mixed with different types of HBDs such as urea, polyols (glycerol or ethylene glycol), sugars or carboxylic acids. To maintain the suitable characteristics of [CH][Cl] when developing DESs, this salt has to be combined with suitable HBDs, and thus, an approach is vary the HBDs used. 57,58 In recent works Maugeri et al. 59 and Florindo et al. 60 proposed the use of HBDs combined with [CH] [Cl]. Levulinic acid ([LEV]) is a compound which is fully biodegradable and non-toxic, that may be obtained from biomass at low costs, 61 59,60 The available studies on CO 2 capture using [CH] [Cl] based DESs are mostly limited to systems containing HBDs such as urea, glycerol, ethyleneglycol, or carboxylic acids such as malonic or lactic, [64][65][66][67][68][69][70][71][72] but no studies for DES involving [LEV] are reported. Therefore, a study on the suitability of CHCl_LEV_1_2 DES for CO 2 capturing purposes is reported in this work. Considering that a full characterization of this material is required in order to analyze its weaknesses and strengths for carbon capture purposes, a physicochemical characterization of CHCl_LEV_1_2 was carried out, in which the most remarkable properties were selected, both because they are required for process design purposes and/or because they provide information about the structure and behavior of the fluid. CO 2 absorption was studied at isothermal conditions at pressures up to 30 bar. Likewise, in order to understand the mechanism of CO 2 absorption at a nanoscopic level, a theoretical study using both molecular dynamics (MD) and quantum chemistry -in the density functional theory (DFT) framework -approaches was reported. The fluid's structure was studied for pure CHCl_LEV_1_2 and for the same fluid after CO 2 absorption as a function of temperature, pressure and amount of absorbed CO 2 via MD simulations. Short-range interactions in the fluid were analyzed in detail from DFT calculations. Likewise, considering the relevance of liquid-gas interfacial behavior for CO 2 capturing purposes, MD simulations were also carried out to analyze the behavior of CO 2 molecules at the DES surface. The reported study shows for the first time a combined full experimental and computational characterization of DESs as sorbents for CO 2 capturing purposes.

Materials
Levulinic acid (CAS: 123762, Aldrich) and choline chloride (CAS: 67481, Iolitec) were used as received. A mixture of 1 : 2 molar ratio of [CH][Cl] : [LEV] acid (CH_Cl_LEV_1_2 DES) was prepared by mixing 46.54 g of chlorine chloride and 77.41 g of [LEV]. The resulting mixture was stirred for 2 hours using a magnetic stirrer. The formation of a eutectic solution was examined and confirmed using both NMR (Varian Unity Inova 400 MHz) and FTIR spectrometry (Spectrum 400, PerkinElmer, USA). CH_Cl_LEV_1_2 DES samples were prepared by mixing the ionic liquid and acid at room temperature in a glove box compartment. Water content (1.69 wt%) was measured using Karl Fischer coulometric titration (Metrohm 831 KF coulometer) to 0.3% accuracy in water mass content.

Experimental details
NMR measurements were performed using a Varian Unity Inova 400 MHz instrument at 298.15 K (AE0.1 K) for the 1 H and 13 C nuclei. CHCl_LEV_1_2 samples were prepared in 5 mm NMR tubes, with 0.005 g of CHCl_LEV_1_2 mixed with 1 mL of D 2 O.
The formation of DES was examined and confirmed using a FTIR spectrometer (Spectrum 400, PerkinElmer, USA). Other than the FTIR studies for the characterization of the DES system, in situ FTIR measurements were carried out using a Bruker s Vertex 80 Fourier transform infrared spectrometer, which was coupled with a temperature controlled high pressure liquid cell (HPL-TC) that can accommodate pressures of up to 30 bar, supplied by Harrickt Scientific. HPL-TC was equipped with high-resolution Si windows, teflon spacers of 2280 micron path-length and Viton s O-rings. The path-length was intentionally kept long to provide more volume, and therefore increase the contact opportunity with CO 2 . In order to adjust CO 2 pressure, a gas feeding manifold system was designed and the schematics are shown in Fig. S1 (ESI †).
Density and viscosity were both measured using Anton Paar DMA 4500M and Lovis 200 M/ME units respectively. The density meter uses the oscillating U-tube sensor principle and it has a volume requirement of 1 mL of sample. The density meter has a reported accuracy of 0.00005 g m À3 in density and AE0.05 K in temperature. The Anton Paar Lovis 200 rolling ball viscometer measures the rolling time of a ball through transparent and opaque liquids according to Hoeppler's falling ball principle with AE3% viscosity uncertainty and AE0.02 K accuracy in temperature measurements.
The pH and conductivity measurements were performed at 293.15 K using a pH/conductivity meter (3200M Multiparameter Analyzer, Agilent Technologies, USA). The conductivity and the pH of the final product was 752.85 AE 5.2 ms cm À1 and 2.7 AE 0.2, respectively, at a temperature of 293.15 K. Indeed, conductivity is an important indicating property, determining if an IL can play the role of both the solvent and the electrolyte in electrochemical reactions. Ionic liquids exhibit a broad range of conductivities from 0.1 to 20 mS cm À1 as conductivity is affected by many factors such as density, viscosity, anionic charge, ion size, delocalization, aggregation and ionic motion. 73 Recently, some DES liquid systems based on choline chloride, 74 with properties similar to those of ionic liquids, have been prepared by mechanically mixing two different components, where choline can be used as an alternative cation in combination with a suitable anion to generate ionic liquids.
Refractive indices were measured relative to the sodium D-line (AE1 Â 10 À5 ) using an automated Leica AR600 refractometer, with the sample temperature being controlled using a Julabo F25 external circulator and measured with a built-in thermometer (AE0.01 K). A standard supplied by the manufacturer was used for refractometer calibration.
Corrosion experiments were conducted using circular specimens of carbon steel 1018 (0.186 wt% C, 0.214 wt% Si, 0.418 wt% Mn, 0.029 wt% P, 0.019 wt% S, balance Fe) with 1.93 cm 2 surface area of specimen exposed to the medium. Carbon steel 1018 is selected for corrosion experiments since it is a common material that is used in process equipment in chemical industries. The specimens were prepared according to ASTM G1-03 standard by wet grinding and polishing using 320 600 and 1200 grit SiC papers. The specimen was then degreased by high purity acetone, rinsed with deionized water and dried with hot dry air. 75 Electrochemical experiments were conducted in a 100 cm 3 jacketed micro cell supplied by Autolab (Fig. 2). The microcell was equipped with a potentiostat (Autolab PGSTAT101) and data acquisition system (NOVEA 1.7). The corrosion cell consisted of an Ag/AgCl reference electrode filled with 3 M KCl solution, two 316 stainless steel counter electrodes and the working electrode. A heated water circulator was connected to the outer cell jacket throughout experiments for temperature control. A water-cooled condenser was utilized to minimize vaporization losses of the test solution. Gas supply streams and a CO 2 gas flow meter was connected to the cell. Throughout all experiments, a gas stream was maintained in the gas phase of the cell. Open circuit potential OCP measurement of the specimen against the reference electrode was recorded until it was stable (defined as AE0.01 mV between successive readings). The polarization curve was generated directly by the data acquisition system in the range of AE250 mV vs. OCP and a scan rate of 0.001 V s À1 . Tafel extrapolation method was used to determine the corrosion current (I corr ) which was converted to corrosion rate by the following equation: where CR is corrosion rate in (mm year À1 ), I corr is corrosion current in (mA cm À2 ), EW is the equivalent weight of the carbon steel specimen (g per equivalent) and r is the density of the specimen (g cm À3 ). The numerical value that appears in eqn (1) is a constant that defines the units for the corrosion rate and it is calculated from the calibration of the apparatus. Thermal stability analysis of materials was performed with a Perkin Elmer Pyris 6 TGA instrument, where samples were heated in a N 2 environment from 303 K to 873 K at the rate of 5 min À1 .
For carbon dioxide adsorption measurements, a highpressure magnetic suspension sorption apparatus (MSA) made by Rubotherm Präzisionsmesstechnik GmbH was used. A schematic of the apparatus is given in Fig. S2 (ESI †). The MSA apparatus is rated up to 350 bar at 373 K. The MSA has two different operation positions. First, the measurement cell is filled with CO 2 gas, and the MSA records the weight change of the sample that is placed in the sample container as the high-pressure gas is absorbed by the sample. The second measurement position is used to measure the in situ density of the high-pressure gas, which is required to calculate the amount of the gas adsorbed onto the sample in the highpressure cell. In this work, pressures of up to 30 bar were used as the maximum pressure, and at the end of each isotherm, a hysteresis check was conducted for each isotherm by collecting desorption data as the system was depressurized. The nature of physical adsorption was later cross-checked by comparing before and after FTIR measurements of the DES sorbent. CO 2 adsorption-desorption isotherms were selected at 298.15 K and 323.15 K. Detailed operating principles and data correlation of the magnetic suspension force transmission is also discussed previously elsewhere. 76,77 For buoyancy calculations used in sorption measurements, the in situ density of the pressurized gas in the high-pressure cell was measured. Archimedes' principle was used for density measurements by utilizing a calibrated silicon sinker placed just above the sample basket in the pressure cell. The silicon sinker used in this apparatus had a volume of 4.4474 cm 3 measured at 20 1C with a 0.0015 cm 3 uncertainty and a density of 4508 kg m À3 measured at 293.15 K with a 4 kg m À3 uncertainty. On the other hand, the uncertainty for pressure measurements was AE0.05% in the full scale of the 350 bar rated pressure transducer and AE0.1 K for the temperature measurements.

Simulations
Optimizations were performed from those systems composed of one isolated molecule (i.e. ions, levulinic acid or CO 2 ) up to systems composed of DES and CO 2 (CHCl_LEV_1_2Á Á ÁCO 2 ). Optimized minima were checked trough their vibrational frequencies. For those simulations wherein two or more molecules are present, different starting points were employed in order to study different relative dispositions, focusing our attention on the disposition of minimal energy. All these calculations were carried out using B3LYP 78,79 coupled with dispersion corrections according to Grimme's scheme 80 (B3LYP-D2), with the 6-31+G** basis set. B3LYP was selected since it has showed a remarkable performance over a wide range of systems, 81 while dispersion corrections are adequate since we are considering systems with dispersive interactions such as hydrogen bonds. 80 Besides, calculated energies after dispersion corrections are comparable with more reliable values, such as those obtained at the MP2 level. 82 From those systems composed of two or more molecules, computed energies were corrected (to avoid basis set superposition error) according to counterpoise procedure. 83 Interaction energies (DE) for different processes related to CO 2 capture and formation of DES were computed. Thus, DE for CHCl_LEV_1_2Á Á ÁCO 2 (i.e. CO 2 capture using DES) was calculated as: DE DES-CO 2 = E DES-CO 2 À (E DES + E CO 2 ), E DES-CO 2 , E DES and E CO 2 being the (counterpoise corrected) energies for CHCl_LEV_1_2Á Á ÁCO 2 , CHCl_LEV_1_2 and CO 2 , respectively. Intermolecular interactions were analyzed by means of Atoms in Molecules (AIM) theory. 84 Topological analyses according to AIM theory were carried out using the MultiWFN package. 85 According to Bader's theory, 84 there are four kinds of critical points, but given the characteristics of the studied systems and to improve and clarify data analysis we have mainly focused over bond critical points (BCP), which raises the criteria for considering the presence of intermolecular interactions. Finally, atomic charges were computed to fit the electrostatic potential according to the ChelpG scheme. 86 All calculations were carried out with Gaussian 09 (Revision D.01) package. 87 MD simulations were carried both for pure CHCl_LEV_1_2 and for CHCl_LEV_1_2 + CO 2 systems. In the case of pure DES, 250 [CH][Cl] ion pairs plus 500 [LEV] molecules were considered for all the simulations in the 298 to 348 K temperature range at 0.1 MPa of pressure. Mixed CHCl_LEV_1_2 + CO 2 were prepared according to the experimental solubility data obtained in this work, and thus, four different mixed systems were prepared for simulations at 298 K, all of them containing the same number of ion pairs and [LEV] molecules as for pure DES simulations and with a different number of CO 2 molecules to mimic experimental adsorption isotherms up to 1 MPa (samples CHCl_LEV_1_2_CO 2 _I, CHCl_LEV_1_2_CO 2 _II, CHCl_LEV_1_2_ CO 2 _III, and CHCl_LEV_1_2_CO 2 _IV; Table S2, ESI †). All these simulations were carried out in the NPT ensemble, with initial low density boxes (B0.2 g cm À3 ) built using the Packmol program. 88 Simulations were carried out starting from these initial boxes being equilibrated during 5 ns with additional 10 ns runs collected for production purposes.
In the case of MD simulation for the analysis of interfacial behavior, a box of CHCl_LEV_1_2 with the same characteristics of those used for the simulation of pure DES, and previously equilibrated, was put in contact along the z-direction with (i) a vacuum layer (with the vacuum layer dimension in the z-coordinate three-times larger than the DES layer) for describing the DES-vacuum interface, and (ii) with a CO 2 gas phase (with gas layer also being three-times larger in the z-dimension than the DES liquid layer) for considering the DES-CO 2 gas interface. The density of the CO 2 gas layer corresponded to that at 298.15 K and 10 bar obtained from CO 2 reference equations of state. 89 These interfacial simulations were carried out in the NVT ensemble at 298.15 K.
All the MD simulations were carried out using the MDynaMix v.5.2 molecular modeling package. 90 Pressure and temperature were controlled using the Nosé-Hoover thermostat. Coulombic interactions were handled with the Ewald summation method, 91 with cut-off radius of 15 Å. A Tuckerman-Berne double time step algorithm, 92 with long and short time steps of 1 and 0.1 fs, was considered for solving the equations of motion. Lorentz-Berthelot mixing rules were used for Lennard-Jones terms.
The forcefield parameterization used along MD simulations is reported in Table S1 (ESI †). This parameterization was developed using DFT calculations for clusters formed by 1[CH][Cl] + 2[LEV] molecules which were optimized as reported in the previous section, and thus, ChelpG 86 charges were calculated for this cluster. The optimized structures for this model cluster show that LEV molecules may interact in two different positions (see Results and discussion section), with [LEV] molecules having different charges depending on the interacting site. Therefore, the parameterization used considers two types of levulinic acid molecules (LEV_I and LEV_II) with different charge parameters (Table S1, ESI †). The proposed parameterization leads to charges of +0.8254 for [CH] + , À0.6849 for Cl À , À0.0663 for LEV_I, and À0.0743 for LEV_II. This parameterization was developed to obtain a more realistic physical picture of the charge distribution in the studied DES instead of applying the simplified option of AE1 total charge for the cation/anion and null charges for any LEV molecule.

Density, refraction index and viscosity
These properties were collected (Table S3, ESI †) and are presented in Fig. 3. Similar DES systems have been studied and their density and viscosity profiles have been reported elsewhere for CHCl and levulinic acid, 60 glyceric acid, 60 phenol, 93 urea, 94 and glycerol 95 containing DES systems. Density and viscosity values are important for these exotic mixtures for the development of suitable equations of state, which has a crucial role in the calculation of further thermodynamic properties for developing industrial processes including gas separation operations that run with novel solvent substitutes for amines. Most of the studied DES system densities are in the range of 1-1.35 g cm À3 at 25 1C. DES systems that contain metallic salts show a tendency to have higher densities at room temperature in the range of 1.3-1.6 g cm À3 .
[CH][Cl] and urea DES systems (with 1 : 2 ratio) have been reported to have 1.25 g cm À3 density at 293.15 K. 96 For the experimented CHCl_LEV_1_2 DES system, density values fall in the range of 1.14-1.10 g cm À3 within the temperature range of 293. 15 96 On the other hand, viscosity has also been extensively measured for the available DESs, due to its importance for industrial purposes. Viscosity data are required for DES applications not only in gas solubility and separation processes but also fields such as lubrication or any other potential high-pressure operation have been considered as well. 97 Moreover, viscosity data is essential to realize the possible mass transfer coefficient limitations as well as fluid pumping issues for advanced process system and equipment design purposes.  60 also reported a viscosity value of the same system for temperatures between 298.15 K and 350 K (e.g. 226.8 mPa s at 298.15 K). The difference between the viscosity data reported in this work and those by Florindo et al. 60 rises from (i) the different methodology for sample preparation (grinding of the DES components by Florindo et al., 60 cf. mixing in this work), and (ii) the different water content. The most remarkable feature should be the water content, which develops a pivotal role for viscosity measurements. The sample used in this work has a water content of 1.6 wt% which is higher than the 0.23 wt% of the sample used by Florindo et al., 60 which would justify the lower viscosity data reported in this work (Fig. 3b). This is confirmed by the viscosity data reported by Florindo et al. 60 for CHCl_LEV_1_2 DES saturated with water (9.88 wt%), which shows less viscous behavior than the sample used in this work.

NMR and FTIR
1 H and 13 C NMR features are summarized in Table S4 (ESI †), the data are in good agreement with those by Florindo et al. 60 and it discards the presence of relevant impurities at least at the NMR detection levels. FTIR characterization of the DES structure provides information about interactions and complexation between constituents. FTIR spectra of pure [LEV] and choline chloride are given in Fig. S1 (ESI †). Regarding pure [LEV], vibrational bands at 1720-1695 cm À1 , 1435-1400 cm À1 , 1380-1345 cm À1 and 1225-1100 cm À1 refer to the presence of aliphatic ketone, while the vibrational bands at 1720-1701 cm À1 refer to a carbonyl compound. The FTIR spectrum of the [LEV] confirms its classification as a keto-acid. For pure choline chloride, vibrational bands at 3540-3200 cm À1 and 1205-885 cm À1 refer to the presence of a hydroxyl or amino group, while those bands at 2990-2855 cm À1 and 1485-1415 cm À1 refer to the presence of an alkyl group. On the other hand, the formation of DES was confirmed by the FTIR spectrum of CHCl_LEV_1_2 DES 1 : 2 molar ratio as shown in Fig. 4, where, the vibrational bands at 2990-2855 cm À1 and 1485-1415 cm À1 indicate the presence of an alkyl group, the bands at 1720-1695 cm À1 , 1435-1400 cm À1 , 1380-1345 cm À1 and 1225-1100 cm À1 represent the aliphatic ketone group, 1720-1701 cm À1 bands refer to a carbonyl compound, and 1745-1710 and 1300-1000 cm À1 bands refer to an ester or ketone compound.

In situ FTIR
For in situ FT-IR measurements, a temperature controlled highpressure liquid cell (HPL-TC) manufactured by Harrick Scientific was used. HPL-TC was equipped with specific internal parts: high-resolution Si windows, Teflont spacers of 2280micron path-length and Viton O-rings. The path-length was intentionally kept thick to provide more volume, and therefore increase the contact opportunity between CO 2 and the solvent molecules as well as to provide a more stable sealed system. In order to manipulate CO 2 injection at the desired pressures under isothermal conditions, a particular system was designed and attached to the spectrometer. In situ FTIR experiments were performed at 50 1C over a range of low CO 2 pressures. To ensure thermal stability, the thermo-regulator was kept operating for at least one hour prior to the experiment. A background spectrum of the HPL-TC system under vacuum was collected before and after each DES in situ FTIR measurement and this background was subtracted from the DES + CO 2 spectrum data. To this end, the sample spectrum was collected upon exposure to CO 2 . This sample spectrum is believed to be influenced by the contribution of both CO 2 in the absorbed phase (dissolved in DES) and gaseous CO 2 in the bulk especially as the pathlength is thick. Hence, in order to eliminate the gaseous CO 2 contribution, pure CO 2 spectra were collected separately at the same conditions of the sample spectrum, and then each was subtracted from the sample spectrum that corresponds to same operating conditions. In situ FTIR experimental findings of CHCL_LEV_1_2 DES with the presence of pressurized CO 2 at different pressures are shown in Fig. S3 (ESI †). Near the CO 2 fundamental bending mode (600-700 cm À1 ), interesting gradual bands appear around 619 cm À1 analogous to the redshifted bending mode band observed in the computational results which was overlapping with other CO 2 -free DES and CO 2 -loaded DES bands. The subtraction of gaseous CO 2 and DES spectra resulted in a loss of CO 2 related bands that are overlapped and appeared as negative split bands. Nevertheless, the effect of the increasing pressure was observed and recognized by the increase in intensity that corresponds to the increase in the concentration of absorbed CO 2 in the liquid DES phase.

TGA
Thermal gravimetric analysis (TGA) of the DES system was performed to investigate the temperature limitation of the absorbent. Fig. 5 shows the TGA profile of the experimented DES system. Single step degradation behavior was observed and the analysis shows that the DES system is stable up to 453-473 1C, which makes the experimented DES system suitable for high temperature post combustion CO 2 capture process conditions as well.

Corrosion measurements
Needless to say that corrosivity is one of the most important concerns from a process operation point of view. It is necessary both to forecast the equipment depreciation with time, and also to make plans not only during material selection for equipment design, but also during operation by selection of appropriate corrosion inhibitors to prevent potential corrosion within the pipe and absorption column vessel. For this purpose, corrosion experiments have been conducted for the DES system and monoethanolamine (MEA) system, as it is the most widely used CO 2 capture agent currently used in chemical plants. Fig. 6 shows the current density vs. potential plot for the two studied DES and MEA systems. The details on how to interpret the polarization curve are given in Section 2.2 of this work. Two identical carbon steel samples were used for the corrosion experiments and they were prepared for the measurements as per ASTM G1-03 standards as mentioned in above section. The surface area of the specimens was calculated to be 1.93 cm 2 . The analysis on the polarization curves for the CO 2 loaded DES system showed that the corrosion potential is À0.43 V and the corresponding corrosion rate behavior is 0.027 mm year À1 . On the other hand, the corrosion behavior of the identical carbon  View Article Online steel specimen in the CO 2 loaded MEA solution was also conducted. The corrosion potential was recorded to be À0.75 V and the corresponding corrosion rate was calculated to be 0.54 mm year À1 . These results show that from a corrosion point of view, the DES system exhibits much more corrosion resistant behavior when compared with the MEA system, and this considerable order of magnitude difference is a great advantage in reducing the both operating and fixed cost of the absorber column and its ancillary equipment in a CO 2 capture plant.

CO 2 solubility
CO 2 solubility studies were conducted by using state of the art magnetic suspension based MSA and the details of the apparatus are mentioned in the experimental section. Pressure transducers (Paroscientific, US) were used in a range from vacuum up to 35 MPa with an accuracy of 0.01% in full scale. The temperature was kept constant with an accuracy of AE0.5 K for each measurement (Minco PRT, US). In situ density values for CO 2 were measured during sorption measurements as it is necessary to calculate the absorbed amount of CO 2 , and density values were cross-checked with REFPROP 9.0 98 for consistency purposes. Absorption measurements were carried out using 2 to 3 mL DES samples. First the system was placed under vacuum for 8 hours at 293 K, 298 K, 308 K, 318 K and 323 K. CO 2 was then pressurized through a Teledyne Isco 260D fully automated compressor and fed into the high-pressure measurement cell in which the CO 2 absorption process begun. Once the solubility equilibrium was reached, measurements were taken for a period of 10 minutes; each data point was collected at 30 second time intervals. At the end of each measured pressure point, a gas-dosing system triggered the compressor to the next pressure measurement point, which increased the measurement cell pressure in a step-wise manner. In this work, pressure up to 3 MPa (30 bars) was used for the maximum pressure and 293 K, 298 K, 308 K, 318 K and 323 K isotherms were investigated for CO 2 absorption in the DES system. A solubility hysteresis check was conducted at each isotherm by collecting desorption data as the system was depressurized. Details of the solubility experiment were based on the amount of absorbed carbon dioxide in the sample as calculated by using the below equation: W + W buoy,sample + W buoy,sink = m abs + m sample + m sink (2) where W = signal read by the instrument; W buoy,sample = V sample Â d gas = buoyancy correction due to sample; V sample = volume of the sample; d gas = density of the gas; W buoy,sink = V sinker Â d gas = buoyancy correction due to sinker; V sinker = volume of the sinker; m abs = absorption amount; m sample = mass of the sample; m sink = mass of the sinker (sinker is a float that is used for in situ gas density measurements). Fig. 7 shows the findings of the solubility measurements at 5 different isotherms up to a pressure of 30 bar. It was observed that the maximum amount of CO 2 solubility was decreased with an increase in temperature. The DES system absorbed 2.316, 2.220, 2.100, 2.027 and 1.934 mmol of CO 2 per gram of DES sample at 293 K, 298 K, 308 K, 318 K and 323 K, respectively, at the maximum operating pressure of 30 bar. The direct weight measurement technique allows in situ observation of the behavior of the measured sample. In other words, a potential swelling effect can be observed during the measurements. In the case of swelling of the sample, the increased sample volume would cause an increase in the buoyancy on the measurement cell, which in turn would give the impression of a decrease in the amount of captured CO 2 . However, it can be seen from Fig. 7 that the CO 2 absorption trend has been reported as a curve  with a positive curvature. During the solubility measurements, sorption hysteresis was also checked by conducting absorption/ desorption cycle analysis. None of measurements at each isotherm did not show a sign of chemisorption, and weight measurements were obtained during the desorption along with the same absorption path. Moreover, in order to check whether there is a degradation of the sorption activity with the experimented DES system, each isotherm measurement was repeated three times and there was no significant absorption activity loss observed.
On the other hand, the kinetics of the absorption in the DES system has been investigated. Fig. 8 shows the amount of absorbed CO 2 in the DES system with time at both low and high pressures. At a pressure of 1 bar, an average time of 7 minutes was required to reach the equilibrium for a fully saturated DES solution. Whereas at 30 bar, 12 minutes passed before equilibrium conditions were reached.

Nanoscopic behavior from molecular dynamics
The forcefield parameterization used in this work was validated through the comparison of predicted physical properties with experimental values. Deviation between experimental and simulated density, r, data in the 298 to 348 K range are reported in Fig. 9. Density data from MD are slightly lower than experimental data, but are closer to those obtained by Florindo et al. 60 Nevertheless, deviations are lower than 0.9% in the whole temperature range, which show the suitable performance of the used parameterization for describing the macroscopic properties of the studied system. The temperature evolution of density, both for experimental and molecular dynamics predicted data, allows the obtaining of the isobaric thermal expansion coefficient, a p , from its thermodynamic definition, leading to 0.583 Â 10 À3 , 0.576 Â 10 À3 and 0.593 Â 10 À3 K À1 at 298 K, for the experimental data obtained in this work, from Florindo et al., 60 and from molecular dynamics simulations, respectively, which shows the excellent agreement between the simulated data and experimental values.
Self-diffusion coefficients were also calculated, using Einstein's equation from mean square displacements (msd), for all the involved molecules leading to values of 0. [Cl] + urea in 1 : 2 molar ratio), their values being lower than those obtained in this work for CHCL_LEV_1_2 (roughly half) in agreement with the larger viscosity of reline in comparison with CHCL_LEV_1_2. Likewise, self-diffusion for [CH] + in reline is lower than for Cl À and the corresponding HBD (urea), which is also in agreement with values for CHCL_LEV_1_2, but HBD in reline has larger molecular mobilities than that of [Cl] À , whereas they have similar mobility in HCL_LEV_1_2. Perkins et al. 99 justified the self-diffusion data of [Cl] À and HBD in reline, indicating that in spite of the strong hydrogen bonding between both compounds their movements are not tied. On the contrary in the case of CHCL_LEV_1_2, the mobility of [Cl] À and HBD (LEV) seem to be strongly correlated.
A key point in the determination of self-diffusion coefficients from molecular dynamics simulations is if the condition of a fully diffusive regime is reached in the time frame used, which is commonly measured through the so-called b parameter, defined as the slope of log-log plots of msd vs. simulation time.
A fully diffusive regime is reached when b = 1 whereas subdiffusive regimes are characterized by b o 1. Perkins et al. 99 showed that in the case of reline at 298 K, a fully diffusive regime is reached at 30 ns, whereas it is reached at 10 ns for CHCL_LEV_1_2 in this work. This may be justified considering that reline viscosity 94 is twice that of CHCl_LEV_1_2, which leads to increased molecular mobilities (as shown by the larger self-diffusion coefficients) and thus diffusive regimes are  reached at shorter simulation times. Dynamic viscosity was calculated using Green-Kubo method leading to 265 mPa s at 298 K, which is in fair agreement with 226.8 mPa s as obtained by Florindo et al. 60 Vaporization enthalpy may be obtained from molecular dynamics simulations as the internal energy difference between liquid and gas phases (plus and R Â T term), which can be assimilated to the difference between potential energies. 100,101 The gas phase was modeled considering a low density cage  Table 1, showing that the main contribution to the potential energy differences between liquid and gas phases is a non-bonded term arising from intermolecular interactions, whereas bonded and intra-molecular non-bonded terms are almost negligible. Likewise, the vaporization enthalpy of CHCL_LEV_1_2 is lower than those values reported in the literature for ionic liquids, 100 which is in agreement with the strong effect of strong coulombic effects arising in ionic liquids, whereas the strength of LEV-ions intermolecular interactions in the studied DES are weaker. This effect should be confirmed experimentally in future works.
The structure of CHCL_LEV_1_2 is strongly dominated by the strength and nature of intermolecular interactions, E inter , and thus, intermolecular interaction energies are quantified and reported in Fig. 10 The main structural features for CHCL_LEV_1_2 may be obtained from the radial distribution functions, g(r), reported in Fig. 11. The results in Fig. 11a show ion-ion interaction in CHCL_LEV_1_2. The anion-cation strong interaction describing the ion pair is well defined by the strong and narrow peak at short distance in Fig. 11a, whereas very minor features appear for this interaction at larger distances. Anion-anion interaction is characterized by a broad peak followed by a shoulder in the 5 to 10 Å range, which is in contrast with the cation-cation interactions, which are characterized by a strong intermolecular hydrogen bonding inferred from the narrow first peak in Fig. 11a followed by three weaker and wider peaks, which show that [CH] + develop well-defined hydrogen bonding through the cation hydroxyl group. For the analysis of ion-[LEV] interactions, it should be remarked that LEV has three sites that may act as donors and/or acceptors for hydrogen bonding: (i) the oxygen atom in -CQO of the -COO group, Ol1, (ii) the -OH in the -COO group, Ol2 and Hl, and (iii) the oxygen atom in the -CO terminal group, Ol3, Fig. 11. Radial distribution functions for these three groups are reported in Fig. 11b with regard to ion-[LEV] interactions. The results in Fig. 11b  being larger for the interaction through Ol1 site. The interaction through Ol2 has shown poor interactions according to the wide peak in Fig. 11b. Likewise, results in Fig. 11b Fig. 11c, showing that the main interaction is developed between the Hl and Ol1 sites, with weaker contributions arising from hydrogen bonding through the Ol3 position and almost negligible interactions through the Ol2 site. Therefore, the [LEV] hydroxyl group acts as a strong hydrogen bond donor, both with [CH] + and other [LEV] molecules, and the intermolecular hydrogen bonding is developed mainly through Ol1, and in minor extension, Ol3 acceptor sites.
A most detailed picture of the molecular arrangement in CHCL_LEV_1_2 can be obtained from spatial distribution functions, SDF, as reported in Fig. 12 Fig. 12a showing the high density cap above the oxygen atom in the hydroxyl group (acceptor for H-bonds) and below the hydrogen atom in hydroxyl group (donor for H-bonds), together with a big cap surrounding the remaining [CH] + molecular regions showing the prevailing role of the hydroxyl group for the development of intermolecular hydrogen bonds. Cl À anions develop a high-density region in the vicinity of the Hc group (Fig. 12b) in agreement with the strong peak in Fig. 11a. The arrangement of LEV Hl atoms around [CH] + (Fig. 12c) follows a similar pattern to that of Hc (Fig. 12a) (Fig. 12d) but being placed at larger distances than in the case of Cl À around [CH] + . Finally, LEV-LEV interactions lead to high-density regions above the Ol1 group (Fig. 12e). The structural features inferred from Fig. 11 and 12 show the development of H-bonding between all the involved molecules through the [CH] + hydroxyl site and LEV -COOH group, and thus, the number of H-bonds is reported in Fig. 13 The development of strong H-bonding between all the involved molecules in CHCL_LEV_1_2 requires an additional analysis of the lifetimes of these intermolecular interactions. This analysis was carried out through the calculation of residence time, t res , of relevant sites around others in order to characterize H-bonding. Residence time was calculated from the exponential decay of the conditional probability for a site to stay within a sphere of defined radius around another molecular site, as explained in previous work. 103 Sun et al. 102 reported lifetimes of hydrogen bonds in reline DES, showing lifetimes lower than 13 ps for all the reported H-bonds in the eutectic composition, with values of 12.574 ps for the [CH] + -Cl À H-bond, 2.397 ps for the urea (HBD)-[Cl] À H-bond, and 2.952 ps for the urea-urea interaction. These values are lower than those reported for CHCL_LEV_1_2 in Fig. 14a, but it may be argued that the distance criteria used to define residence times in Fig. 14a (6.0 Å) is roughly double that used by Sun et al. 102 to define their H-bond lifetimes (roughly 3 Å). The criteria used in Fig. 14 consider the existence of a second solvation shell around each hydrogen bond donor site, and thus, although H-bonds are destroyed for distances larger than 3.0 Å, those atoms reamining in the 3-6 Å range arise from the reorganization of the local H-bonding structure and they may develop a new H-bond in a fast way, and thus using 6.0 Å may give a realistic picture of the H-bonding around a certain site. The residence times reported in Fig. 14a are very similar for all the considered interactions being in the 35 to 55 ps range. Moreover, the analysis of the temperature evolution of residence times follows a similar non-Arrhenius pattern for all the studied interactions, thus confirming a strong correlation between the dynamics of all the studied H-bonds.

. The [CH] + -[CH] + interaction is well defined in
The results in Fig. 15 show the changes in intermolecular interaction energy upon CO 2 absorption for CO 2 pressures up to 10 bar. The structure of CHCL_LEV_1_2 does not change upon CO 2 absorption from an energetic viewpoint; ion-ion, ion-LEV and LEV-LEV E inter suffer very minor changes with increasing CO 2 mole fraction (Fig. 15a) (e.g. for [CH] + -[Cl] À , E inter decreases just a 2.5% in absolute value). Therefore the liquid structure of CHCL_LEV_1_2 is able to accommodate CO 2 molecules without remarkable changes in its mechanism of intermolecular interaction. Regarding the strength of the interactions between CO 2 and molecules present in CHCL_LEV_1_2, the results in Fig. 15b show that the strength of interactions   Fig. 16 for radial distribution functions. The arrangement of CO 2 molecules around [CH] + (Fig. 16a) show that CO 2 molecules are placed both around the cation hydroxyl group and also around the methyl groups, following a similar pattern for the first solvation sphere, although for the hydroxyl group two additional maxima in g(r) points to a slightly stronger interaction through that site. This is confirmed by spatial distribution functions as shown in Fig. 17a, which although they show high density caps around the [CH] + hydroxyl group, they also show a very symmetrical distribution of CO 2 molecules around the methyl group, which would justify the sharp and narrow peak around the [CH] + nitrogen atom in g(r) as shown in Fig. 17a. The CO 2 molecules are also strongly structured around [Cl] À (Fig. 17b), the sharp peak in g(r) following a similar pattern to the interaction between CO 2 molecules and the Hc site in [CH] + ; therefore as the absorption of CO 2 molecules does not disrupt anion-cation interactions, CO 2 may interact efficiently with both ions at the same time. The CO 2 -[LEV] interaction is characterized by the strong features in g(r) around the Hl and Ol3 sites, with weaker interactions through the Ol3 site, leading to a CO 2 distribution around the terminal [LEV] methyl group. Therefore, CO 2 molecules are placed around the LEV hydroxyl group and also around the terminal groups of the [LEV] molecule ( Fig. 17b) with minor concentration around the Ol1 group.   Upon increasing the absorbed amounts of CO 2 , these molecules tend to form clusters. At a CO 2 pressure of 1 bar, the number of absorbed molecules is very low and thus CO 2 clusters were not observed (Fig. 18), but with increasing CO 2 mole fraction (increasing pressure) a well defined peak at 4.3 Å in g(r) is obtained followed by a second and wider peak. The position of these peaks does not change with increasing pressure, only their intensity increases with increasing amounts of absorbed CO 2 . The corresponding running integrals for these g(r) show that the first solvation sphere, integrating up to the first minimum, around a central CO 2 molecule contains at least two additional CO 2 molecules, whereas another two additional CO 2 molecules may be found in the second solvation sphere for the highest studied pressures.
The dynamics of CO 2 absorbed molecules may be analyzed considering the residence times of CO 2 around both ions and [LEV] molecules, as shown in Fig. 19. The results show that the mobility of CO 2 molecules is almost twice around [Cl] À and [LEV] than that around [CH] + , and in all the cases these mobilities increase with increasing CO 2 pressure, with the exception of the change from 1 to 4 bar. This result is surprising considering the strength of the interactions between CO 2 molecules and CHCL_LEV_1_2 molecules as reported in Fig. 15b which led to larger E inter for CO 2 -LEV than for the other interaction pairs -but it may be explained considering steric factors around the [CH] + which hinder the molecular mobility of CO 2 molecules interacting with this ion in comparison with the CO 2 absorption sites around LEV.
A remarkable feature that should be analyzed in order to understand the mechanism of absorption of CO 2 molecules in CHCL_LEV_1_2, is the interfacial behavior of CHCL_LEV_1_2-CO 2 systems. The CO 2 capture from flue gases requires that CO 2 molecules are adsorbed at the sorbent-gas interface and then they diffuse from the interfacial region toward the bulk DES. Therefore, to analyze this process step a simulation on the properties of the CO 2 -DES interface was carried out using molecular dynamics in this work. Pure CO 2 gas was put in contact with CHCL_LEV_1_2 and the behavior was followed as a function of time. For comparison purposes the CHCL_LEV_ 1_2-vacuum interface was also simulated. The density profiles shown in Fig. 20a-c allow the characterization of the molecular arrangements of ions and LEV molecules in the interfacial region with vacuum. The density profiles for ions and LEV center-of-mass in Fig. 20a show that [LEV] molecules occupy outer regions close to the vacuum layer with ions being placed in inner regions, therefore, the CHCL_LEV_1_2-vacuum interface is composed mainly of [LEV] molecules (Fig. 21a). Fig. 20b shows that [CH] + cations tend to orientate themselves parallel to the interface, whereas results in Fig. 20c show how [LEV] molecules are placed perpendicular to the vacuum interface with the -COOH group pointing toward the bulk fluid and the terminal methyl group exposed to the vacuum layer. These molecular arrangements do not change significantly when CHCL_LEV_1_2 is put in contact with a CO 2 gas layer. Very subtle structural changes occur to accommodate adsorbed CO 2 molecules at the interface (Fig. 20d-f), but ion and LEV arrangements are similar to those at the vacuum interface (Fig. 21b).
The main structural feature of the CHCL_LEV_1_2-CO 2 interface is the formation of a strongly adsorbed layer of gas molecules (Fig. 21b) which is formed in the first stages of the simulations (0 to 1 ns); then upon the formation of this initial layer its size increases upon increasing simulation time (Fig. 22). The main characteristic of this adsorbed layer is that CO 2 molecules are adsorbed very quickly but once they are in this layer above CHCL_LEV_1_2, in contact mainly with [LEV] molecules, their mobility is significantly reduced, and thus the Fig. 18 CD-CD site-site radial distribution functions, g(r), in CHCL_LEV_1_2 + CO 2 systems calculated from molecular dynamics simulations as a function of CO 2 pressure at 298 K. CD stands for carbon atoms in CO 2 . Fig. 19 Residence time, t res , of CO 2 carbon atoms around the center of mass of [CH] + , [Cl] À , and LEV molecules in CHCL_LEV_1_2 + CO 2 systems calculated from molecular dynamics simulations as a function of CO 2 pressure, p, at 298 K. CD stands for carbon atoms in CO 2 . t res was calculated from the exponential decay of conditional probability P with R = 6.0 Å.
time that CO 2 molecules remain in this layer on top of CHCL_LEV_1_2 is long (Fig. 22). In the first 6 ns of simulation, all the CO 2 molecules were placed in the adsorbed layer whereas no CO 2 molecules were able to diffuse toward the bulk CHCL_LEV_1_2. The layer of CO 2 adsorbed molecules develop strong interactions with [LEV] molecules and weaker ones with [CH] + , especially when the adsorbed layer is wide enough (simulation times > 4 ns, Fig. 23), which is in agreement with the interface structure inferred from Fig. 20a. This persistence of the CO 2 molecules in the initially adsorbed layer at the CHCL_LEV_1_2-gas interface before moving toward the bulk liquid phase should be considered for any industrial application of this DES as a carbon capture agent. The kinetics of the CO 2 absorption process (largely controlled by the high affinity of CO 2 molecules for the [LEV]-rich interfacial region) has to be analyzed together with thermodynamic factors controlling the capturing process.

Short-range properties through DFT
Previously, CHCL_LEV_1_2 structural features and CO 2 capture mechanism have been analyzed in the context of MD simulations. In this section, the main features relating to CHCL_LEV_1_2 and CHCL_LEV_1_2Á Á ÁCO 2 systems are assessed trough DFT simulations. The bulk effects that are studied though DFT simulations  allow detailed insights of the studied systems at the molecular level which is useful for forecasting macroscopic behavior. Fig. 24 (Fig. 24). As expected, the main interaction between both ions is due to a coulombic attraction between both ions, which agrees with the high binding energy for the ionic pair interaction, |DE IP | = 511.81 kJ mol À1 . For the ionic pair, the intermolecular charge transfer (CT IP ) is equal to 0.161e À . In addition, the chloride anion stabilizes four intermolecular hydrogen bonds with the cation (d 1 -d 4 ), wherein the strongest H-bond is via the OH (cation) group.
In accordance with MD simulations, the optimized structure for CHCL_LEV_1_2 (Fig. 15) is mainly characterized by intermolecular hydrogen bonds between the chloride atom and both levulinic acid molecules. Both levulinic acid molecules H-bond (through H atom of COOH group) with the chloride atom (d 5 and d 6 for LEV1 and LEV2, respectively). In addition, both molecules also establish several H-bonds with the choline cation. All these bonds (d 7 , d 8 and d 9 , d 10 for LEV1 and LEV2, respectively) are formed through the COOH group (levulinic acid) and methylic hydrogen atoms (choline), except d 7 , which is formed between the COOH group and the -CH 2 group adjacent to the OH motif. As can be seen in Table 2, Cl À -LEV interactions yield the largest electronic density values, i.e., the key interactions in the DES structure are those carried out between the levulinic molecules and the chloride atoms. The interaction energy between the ionic pair and the [LEV] molecules was |DE DES | = 141.93 kJ mol À1 (Fig. 25), favorable lower energy state DES structure. Fig. 25 also shows the computed charges over cation, anion and levulinic acid motifs. The largest change on the charges is noted for the chloride atom, which decreases its charge 0.155e À (relative to the [CH][Cl] ionic pair). For CHCL_LEV_1_2 some negative charge is transferred from the anion up to the levulinic acid molecules, 0.066e À and 0.074e À for LEV1 and LEV2, respectively. In addition, there are also some hydrogen bonds between both ions, very similar to those reported in Fig. 24. In this sense, the chloride atom forms three hydrogen bonds (d 1 -d 3 ) with the cation, which are   ) and chloride (q Cl À ) motifs, as well as the binding energy (DE IP ) for the ionic pair interaction are also shown.
weakened compared with the ionic compound. Thus, d 1 suffers an elongation of 0.157 Å, while its electronic density value is diminished by 0.0108 a.u. From the CHCL_LEV_1_2 optimized structure, the binding energy for the ionic pair has been also calculated through a single point calculation, which is also used for estimating the interaction strength. For CHCL_LEV_1_2, |DE IP | = 503.87 kJ mol À1 , i.e. DES formation results in a waning of the ionic pair interaction of around 7.93 kJ mol À1 . Fig. 26 shows the optimized structure for CHCL_LEV_1_2Á Á ÁCO 2 . Three different arrangements (a, b, and c) were found for the interaction between the DES and the CO 2 molecule. As noted from MD simulations, the structure of CHCL_LEV_1_2 does not change upon CO 2 absorption. In this sense, the same intermolecular H-bonds reported for the DES structure (d 1 -d 10 ) were also found in presence of a CO 2 molecule. As can be seen in Table 2   Optimized structures for CHCL_LEV_1_2 at B3LYP-D2/6-31+G** level, along with the main structural parameters related with intermolecular interactions. Computed charges over choline (q CH + ), chloride (q Cl À ) and levulinic (q Lev ) motifs, as well as the binding energy (DE DES ) corresponding to DES formation are also shown. molecule (whose charge is 0.027e À ). However, the negative charge over LEV2 in arrangement c does not vary, the charge transfer up to CO 2 molecule being around 0.014e À . As far as the Cl charge is concerned, the largest charges are noted for arrangement c. With regard to arrangement b, the CO 2 molecules are mainly interacting with the COOH group of LEV2, through d 13 which is stronger (based on lower/larger intermolecular distances/ electronic density values, respectively) than those interactions found in arrangements a and c. Although the charge over LEV2 is still very small, LEV2 decreases its charge 0.037e À , while for the charges over LEV1, choline and chloride remain unaffected. These interactions provide the largest binding energies for CO 2 capture, DE c DESÀCO 2 ¼ 24:90 kJ mol À1 . According to the DFT results, CHCL_LEV_1_2 displays three different positions for CO 2 absorption, whose high values would indicate that this is an energetically favored process.

Conclusions
This extensive study combines both experimental and theoretical investigation of the DES system made by mixing CHCl and LEV with a 1 : 2 mixing ratio. Experimental studies are focused on obtaining detailed physical characteristics of the studied system; and CO 2 solubility at high pressures was the most important part of the experimental study. A promising absorption of 2.316 mmol of CO 2 per gram of DES sample at 293 K has been obtained at a pressure of 50 bar. Moreover, a detailed corrosivity study that deals with the CO 2 saturated DES system demonstrated a great corrosion resistance, with a corrosion rate of 0.027 mm year À1 , whereas the same experiments showed a corrosion rate of 0.54 mm year À1 for the CO 2 saturated MEA system. Molecular dynamics and DFT results allowed the analysis of the structural properties of CHCL_LEV_1_2, both in the pure state and after CO 2 absorption. The CHCL_LEV_1_2 liquid phase is characterized by the development of intermolecular hydrogen bonding between [LEV] molecules and both Cl À and [CH] + ions, with [LEV] molecules mainly accepting H-bonds through the Ol1 atom in the -COOH group and also acting as a H-bond donor through the hydrogen atom of the hydroxyl group. Likewise, [CH] + is also self associated through H-bonding, and thus the characteristics of the functional groups available in the molecules comprising CHCL_LEV_1_2 allow the development of a very efficient network of H-bonds, which is the main reason for the properties of this fluid. This structuring does not change upon CO 2 absorption for pressures lower than 10 bar, with the CO 2 molecules developing stronger interactions with [LEV] molecules than with the ions. The CO 2 capture process by CHCL_LEV_1_2 is also characterized by the strong affinity of CO 2 molecules for the interfacial DES-gas region, which is very rich in [LEV] molecules, thus leading to large lifetimes of the adsorbed molecules in the interfacial region with slow migration toward the bulk fluid region.
The combined experimental and computational results reported in this work allow a detailed characterization of CHCL_LEV_1_2 DES, from both macroscopic and nanoscopic viewpoints, and demonstrate its suitability for possible use as carbon capture agent. Optimized structures for CHCL_LEV_1_2Á Á ÁCO 2 system (three different arrangements were found) at B3LYP-D2/6-31+G** level, along with the main structural parameters related with intermolecular interactions. Computed charges over choline (q CH + ), chloride (q Cl À ), levulinic (q Lev ) and CO 2 (q CO2 ) motifs, as well as the binding energy (D E DES-CO 2 ) corresponding to CO 2 capture by DES are also shown. For clarity, intermolecular bond labels corresponding to DES have been omitted, since we followed the same nomenclature used in Fig. 25.
Computing and Advanced Technologies Foundation of Extremadura (CénitS, LUSITANIA Supercomputer, Spain), and Consortium of Scientific and Academic Services of Cataluña (CSUC, Spain) for providing supercomputing facilities. The statements made herein are solely the responsibility of the authors.