COSMO-RS Analysis of CO2 Solubility in N-Methyldiethanolamine, Sulfolane, and 1-Butyl-3-methyl-imidazolium Acetate Activated by 2-Methylpiperazine for Postcombustion Carbon Capture

Novel aqueous (aq) blends of N-methyldiethanolamine (MDEA), sulfolane (TMSO2), and 1-butyl-3-methyl-imidazolium acetate ([bmim][Ac]) with amine activator 2-methylpiperazine (2-MPZ) are analyzed through conductor-like screening model for real solvents (COSMO-RS) for possible application in the chemisorption of CO2. The molecules associated are analyzed for their ground-state energy, σ potential, and σ surface. Thermodynamic and physicochemical properties have been assessed and paralleled with the experimental data. Vapor pressure of the blended systems and pure component density and viscosity have been compared successfully with the experimental data. Important binary interaction parameters for the aqueous blends over a wide temperature, pressure, and concentration range have been estimated for NRTL, WILSON, and UNIQUAC 4 models. The COSMO-RS theory is further applied in calculating the expected CO2 solubility over a pressure range of 1.0–3.0 bar and temperature range of 303.15–323.15 K. Henry’s constant and free energy of solvation to realize the physical absorption through intermolecular interaction offered by the proposed solvents. Perceptive molecular learning from the behavior of chemical constituents involved indicated that the best suitable solvent is aq (MDEA + 2-MPZ).


INTRODUCTION
The quest to reduce CO 2 emissions via different routes has been a major concern over the past few decades. The process intensification of the existing CO 2 capture techniques and introduction of novel solvents for achieving the same through chemisorption or physisorption has been proposed by many researchers. 1,2 An extensive lab-scale development of vapor− liquid equilibria, 3 kinetic studies, 4 thermophysical properties, 5,6 calculation of binary interaction parameters, 7 improvement in the existing modeling techniques, 8 proposing new correlational analysis, 9 optimizing the reaction or process parameters, 10 defining the structural property relationships, 11 heat of absorption, 12 etc. are an integral part of the development of new solvents for CO 2 or other acid gas absorption applications. However, most of the experimental investigations at the pilot scale tend to be very expensive, and therefore, an efficient solvent screening through various analyses of the proposed solvents is the need of the hour to arrive at a conclusion of their possible applicability at the plant scale. Conclusively, researchers have shifted the research a step back at the quantum-molecular level to understand the basic phenomena of the novel solvents than to lab-or pilotscale studies for achieving the anticipated CO 2 separation. 13 −17 The conductor-like screening model for real solvents (COSMO-RS) is a method of quantum chemical calculations grouped with statistical thermodynamics. The same has been widely applied for accurate prediction of thermodynamic or essential behavioral properties such as Gibb's free energy, activity coefficients, partition coefficients, etc. Calculating the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) gaps to define the polarities or electronegativities associated with a specific molecule yields a useful insight for behavioral analysis of any molecule. 18 This significantly reduces the amount of time, energy, and cost associated with the detailed study required for a set of selected solvents. The variations of the predictive capabilities of COSMO-RS can be deduced from a brief literature survey. COSMO-RS applications in the membrane separation processes through COSMOmic simulations have been widely studied in the recent past, indicating the efficiency of estimation of partition coefficients with fewer deviations from experimental studies in comparison to molecular dynamics simulation evaluations. 19−23 The self-assembly of the surfactants Triton X-114 and Triton X-100 in water solutions at different concentrations and temperatures has been reported, 24 where the partition behavior of neutral solutes and micellar structures was studied. The research findings are of great value for extraction and purification through membrane processes using surfactants. The partition behavior of various amino acids at distinct ionization states was predicted successfully using COSMO-RS indicated 25 for the extraction of biomolecules using surfactants. A detailed octanol/water partition coefficient prediction through the conceptual application of molecular dynamics and the subsequent input to COSMOmic results in efficient assemblies of complex multiphase fluids are also reported. However, the prediction accuracy depends highly on the i/p molecular structures acquired from molecular dynamics simulations. 26 Similar studies can also be performed with respect to the CO 2 capture process of adsorption through membranes.
Further with respect to CO 2 absorption or adsorption, the important properties of understating are vapor pressure, excess enthalpy, excess Gibb's energy, infinite dilution activity coefficient (IDAC), activity coefficients, chemical potentials, physiochemical properties such as density, viscosity, refractive index, toxicity, biodegradability, thermal stability, corrosion behavior, and CO 2 solubility at a specified temperature and pressure, etc. Thorough knowledge of these properties plays a significant role in selecting a solvent. This is owing to the fact that each of these properties signifies the overall effectiveness of the CO 2 capture process. For instance, if the selected solvent exhibits a high viscosity at the absorption temperature and pressure, the overall pumping costs from the regenerator and absorber columns are expected to be very high. On the other hand, if the chemical potentials of the selected solvents are negative in the selected range of temperature, the solvents will not react with CO 2 chemically at all, leading to merely physisorption in the capture process.
Assessing the possibilities of COSMO-RS for the important parameters of prediction and analysis, the current study is proposed for CO 2 capture through absorption. The nonlinearity exhibited in the absorption of CO 2 in previously studied solvents 27,28 suggests that using a quantum calculation method may yield an accurate estimation of various involved thermodynamic properties. The selection of proposed blends is based on an extensive literature survey. The analysis of amine activator 2-methylpiperazine (2-MPZ) 4,29,30 for enhancement of CO 2 capture is carried out in tertiary amine N-methyldiethanolamine (MDEA), 31,32 physical solvent sulfolane (TMSO 2 ), 33,34 and imidazolium-based ionic liquid 1-butyl-3-methyl-imidazolium acetate ([bmim][Ac]). 35,36 MDEA is reported to exhibit a high CO 2 loading, although it has a low reaction rate. The latter is owing to the fact that MDEA acts only as a weak base to release free OH − that further interacts with CO 2 . Further, the reaction of MDEA with CO 2 is less exothermic when compared to primary amines. MDEA additionally offers various advantages such as high thermal and chemical degradation opposition and low solution vapor pressure in comparison to monoethanolamine (MEA) and diethanolamine (DEA). 37,38 TMSO 2 is also considered for its qualities of high physical absorption capacity, high thermal stability, low heat of absorption, and lower corrosion characteristics. 33,34 1-Butyl-3-methylimidazolium acetate, being a room-temperature ionic liquid, offers an insignificantly low vapor pressure and is recognized for its thermal stability and CO 2 capture capacity.
The solvents are chosen in such a manner as to provide molecular and thermodynamic insights into each category of solvents. The base solvents, i.e., MDEA, TMSO 2 , and [bmim][Ac], are proposed in the concentration range of 2.5− 3.5 mol·kg −1 , whereas the concentration range of the activator is varied from 0.5 to 1.5 mol·kg −1 . The selected temperature range is 303.15−323.15 K in view of application in absorption phenomena. The variables studied in the respective range of solvents are based on the recommendations in the literature. 29−39 Various important properties such as σ potential, σ profile, vapor pressure, pure component density and viscosity, infinite dilution activity coefficients, activity coefficients, Gibb's free energy, chemical potential, CO 2 solubility, Henry's law coefficient, etc. have been analyzed through the COSMO-RS theory using " is the number of distributed segments that has surface charge density σ • "A i (σ)" is the segment surface area that has charge density σ • "A i " is the area of the whole surface cavity rooted in the medium 3 chemical potential of a surface segment with screening charge density • "μ s (σ)" is the chemical potential of a surface segment • "σ" is the polarity of the surface under study 4 vapor pressure is the vapor pressure of the component under study is the vapor pressure of the reference component if any considered • "μ gas i " is the chemical potential of the component in the gas phase • "μ S i " is the chemical potential of the pure component in mixture S • "R" is the universal gas constant • "T" is the temperature at which the vapor pressure is to be estimated

Ring
" is the number of ring atoms is the pseudo-chemical potential of "i" at infinite dilution • "μ i 0 " is the chemical potential of "i" in its pure liquid state 9 COSMO model for VLE x i " is the mole fraction of the compounds in the liquid phase is the mole fraction of compounds in the gas phase 10 nonrandom twoliquid (NRTL) model • "λ ij ," "λ ji ," "V i ," "V j ," "a ji ," and "a ij " are binary interaction parameters of the system under study

COMPUTATIONAL METHODS AND THEORY
The molecules under study, i.e., N-methyldiethanolamine (MDEA), sulfolane (TMSO 2 ), 1-butyl-3-methylimidazole, cation ([bmim]), acetate anion ([Ac]), H 2 O, CO 2 , and 2-methylpiperazine (2-MPZ), were selected within COSMOtherm (COSMOlogic GmbH, Leverkusen, Germany). Single conformers with the least ground-state energy were selected for each of the molecular compounds obtained with BP-TZVPD-FINE-level COSMO calculations that incorporate a full geometry optimization by density functional theory (DFT) using the Becke and Perdew (BP) functional with the triple-ζ valence polarized (TZVP) basic set. The detailed specifications of the chemicals for the experimental work are presented in Table 1. The analysis of σ surface, σ potential, and vapor−liquid equilibrium and estimation of vapor pressure, pure component density and viscosity, infinite dilution activity coefficients, CO 2 solubility, Henry's law coefficient, and pK a values are carried out over a wide range of temperatures using the respective modules within COSMOtherm. For pK a calculations, the respective protonated structures of the molecules were developed using TURBOMOLE. All of the estimated values have been presented up to three significant digits after the decimal.

RESULTS AND DISCUSSION
The detailed mathematical relationships of various studied parameters with chemical potential are presented in Table 2. 40 The conforming significance of the simulated properties has been conversed alongside the analysis results.
3.1. σ Profile and σ Potential Analysis. The intermolecular interactions of the selected solvents among all of the constituents along with CO 2 contribute largely toward CO 2 solubility. This further hinges on the associated shape, size, polarity, and type of molecules. Chemical potential "μ" of any species in a solution is evaluated using screening charge density "σ" on the surface of molecules through the COSMO-RS theory within three major norms: (i) the liquid state is incompressible, (ii) all fragments of molecular surfaces can be in interaction is the combinatorial contribution signifying the entropic size and shape transformations of the compounds the parametric equations are as follows and "Θ i " are the normalized volume and surface area fraction of species "i" in the blended solvent x i " is the mole fraction the enthalpy interactions among various constituents in the UNIQUAC 4 model are quantified by the residual contribution in the calculation of the activity coefficient; the mentioned term is described as enthalpy, being closely related to temperature, the residual contribution of binary interaction major parameters "τ ij " is further taken as an inverse logarithm function of temperature and is given as the compound-specific UNIQUAC volume and surface area parameters are presented as " is the vapor pressure of the pure compound • "γ j " is the activity coefficient • "x j " is the mole fraction • "p j " is the partial pressure of compound "j"

14
Henry's law coefficient The σ profile and the corresponding σ potential for the molecules functional to majorly chemical potential are shown in Figure 2. These properties define the attraction of selected solvents with the desired solute, thereby determining the extent of possible separation.
The negative polarities of any molecule are indicated by positive screening charge density in a σ-scale and vice versa. 41 The least σ surface was obtained for H 2 O in the extensive range of −0.02 e·Å −2 , and +0.02 e·Å −2 specifies the positive and negative polarities of the associated atoms in the H 2 O molecule. Successively, it can also be seen from Figure 2a that key portions of σ charge densities of the [bmim] cation, TMSO 2 , MDEA, and 2-MPZ are negative and those for CO 2 and the [Ac] anion are positive in nature. Further, the peak intensities of MDEA and 2-MPZ are very competitive with each other, indicating the possible high CO 2 solubility offered by both. Also, if the peaks of TMSO 2 are analyzed, it can be perceived to have a positive and negative σ charge density with two sharp peaks. The higher peak is, however, present on the negative side. It can thus be concluded that as CO 2 and selected solvents present different charge densities, the selected solvents could provide good CO 2 absorption. This conclusion is in agreement with the fact that the negative surface pieces of  The intermolecular interaction of a solvent toward the molecular surface that it comes in contact with polar or nonpolar behavior can be qualitatively discussed in terms of the σ potential. 43 The positive σ potential of CO 2 over the studied charge density of −0.03 to +0.03 e·Å −2 indicates its capability as a H-bond acceptor (Figure 2b). The σ potential behavior of CO 2 is almost symmetrical and concave in nature over the entire charge density. On the contrary, the parameter is asymmetrical for [bmim], [Ac], 2-MPZ, MDEA, and TMSO 2 , leaning more to the negative side of charge density. Additionally, only the [bmim] cation is associated with a positive σ potential in the negative surface charge density region. The molecules signifying a −ve σ potential act as H-bond donors, whereas the +ve σ potential suggests that CO 2 is a H-bond acceptor. However, understanding the [bmim] cation alone does not provide any technical application since, in the present study, it is associated with the [Ac] anion. Combining the chemical potentials of both the [bmim] cation and the [Ac] anion leads to the overall negative charge density, proposing it to be a H-bond donor. The formation or loss of a H bond is usually at the S, N, or O atoms of any molecules. At a molecular level, the possible interaction of CO 2 with any chosen solvent depends highly on the H-bond acceptor or donor capacity. The interaction strength of CO 2 can hence be determined with the order as At a lab or plant scale, the same concept is understood by the reaction mechanism of zwitterions, proton exchange reactions, and formation or dissolution of bicarbonates, unstable/stable bicarbamates, and dicarbamates. 44,45 3.2. Vapor Pressure Analysis. The potential applications of any solvent in diverse fields of chemical engineering depend on many important characteristics such as thermal and mechanical stability, low degradation and toxicity levels, the extent of biodegradation offered, recyclability, vapor pressures, etc. Among these many essential features, vapor pressure is considered to be very important for the CO 2 capture process. This is due to the fact that any solvent exhibiting high vapor pressure will lead to huge losses during regeneration. Also, if the vapor pressure is too high, the solubility of CO 2 or other acid solute gases decreases at high temperatures. The latter is because, at high temperatures, the diffusivity is expected to increase considerably. On the other hand, if the vapor pressure is too low, e.g., in the case of pure ionic liquids, the diffusion is also too less at low temperatures. Hence, an optimum vapor pressure is desired        (1) where N, Y i exp , and Y i mod indicate the number of data points, experimental value, and modeled or COSMOtherm estimated value of any variable, respectively.
The vapor pressures of aqueous (aq) (MDEA + 2-MPZ), aq (TMSO 2 + 2-MPZ), and aq ([bmim][Ac] + 2-MPZ) have been measured using the validated experimental methodology of our previous work. 51  3.3. Estimation of Pure Component Density and Viscosity. The experimental analysis of pure component density and viscosity has been reported by many researchers for different purposes in carbon capture systems, mainly for estimation of kinetic parameters and pumping costs, 54,55 concise designing of absorption−stripper columns, 56 understanding the nonideal behavior through analysis of viscosity deviation or excess molar properties, 57 etc. In the present work, pure component density and viscosity are estimated through the quantitative structure− property relationship (QSPR) approach, which consists of many inherent properties of the involved molecules ( Table 2). The pure component density and viscosity measurements have been carried out using a density and sound velocity meter (DSA 5000 M, Anton Paar, Austria) and an Anton Paar AMVn rolling ball viscometer with the standard uncertainties of u(T) = 0.01 K, u(P) = 0.2 kPa, u(ρ) = 0.5 kg·m −3 , and u(η) = 0.07 mPa·s. The adopted detailed methodology can be referred to from our previous work. 51 A comparison of the experimental and COSMO estimated density and viscosity of the involved chemical species is presented in Tables 5 and 6. For the case of 2-MPZ, since it is in crystalline form, the experimental measurement was difficult and hence viscosity was estimated using Aspen plus 58 and  , i.e., 0.707 × 10 2 and 1.067%, respectively. A major reason for this huge deviation, especially with respect to viscosity, is owing to the fact that the structural relationship utilized for the prediction of pure component properties, i.e., QSPR, is assumed to be independent of temperature, which is usually not due to non-Newtonian and nonideal liquid systems. However, similar observations are also reported elsewhere. 59 As a matter of fact, it can be concluded from the present comparison of experimental and COSMO-computed values of density and viscosity that the estimation is not accurate for both the properties. Further, the prediction of pure component density and viscosity does not get reflected in other evaluations for the reason that, as indicated in Table 2, the other properties are calculated as a function of chemical potential. Hence, effective prediction of chemical potential results in accurate predictions of the properties under consideration such as vapor pressure, except pure component density and viscosity. 3.4. Infinite Dilution Activity Coefficients. The magnitude of nonideality in liquid solutions is often in terms of infinite dilution activity coefficients (IDACs). Since the solvents associated with absorption of CO 2 and consequently the same solvent for desorption while releasing CO 2 from it are expected to show nonideality behavior, such nonidealities are also variously expressed in terms of property deviations such as density and viscosity deviations from ideality. A few important thermodynamic properties, for instance, partition coefficient, separation factors, and Henry's law coefficient, are also dictated by activity coefficients. However, the experimental calculation of IDACs is difficult and expensive. Hence, a preliminary analysis through the apposite prediction method is of huge interest. 60−62 The estimation of IDACs of aqueous, nonaqueous, organic, or ionic liquid systems, through COSMO-RS, is reported in the literature to be very efficient owing to the reason the same being evaluated in the absenteeism of meanfield approximation. 63 The nonpolarity or active polarity of a compound is decided by the higher or lower value of IDACs in the system (Table 2). Table 7 presents the COSMO predicted infinite dilution activity coefficients of MDEA, 2-MPZ, and TMSO 2 in H 2 O as a function of temperature in the range of 298.15−333.15 K. With an increase in the temperature, IDACs of MDEA, TMSO 2 , and 2-MPZ in water were found to increase. Although the IDACs values for MDEA and 2-MPZ are very low in comparison to the TMSO 2 values, for the currently studied compounds, all of the IDACs are greater or less than unity, indicating the highly nonideal behavior of such systems. 43 3.5. Vapor−Liquid Equilibrium Relationship of Aq (MDEA + 2-MPZ), Aq (TMSO 2 + 2-MPZ), and Aq ([bmim][Ac] + 2-MPZ). At a specific temperature and pressure, the extent of reaction resulting in formation of products or scattering of various constituents is dependent on the chemical potential and subsequently on the Gibbs' free energy. Further, the maximum gas solubility at equilibrium is also dependent on the values of the activity coefficients of the solvent. The chemical potential is also considered a sum total of various energies as well as factors affecting the energies such as internal, density, temperature, enthalpy, etc., of any molecule. This indicates the dependency of acid−gas separations at a molecular level on the thermodynamic properties such as chemical potential, enthalpy, Gibb's free energy, etc.
The activity coefficients and binary interaction parameters for NRTL, WILSON, and UNIQUAC 4 models of the systems under study were estimated and are presented in Tables 8, 9, and 10, respectively. The majority of the binary interaction parameters of the studied systems are found to be different because the systems are asymmetric, i.e., τ ji ≠ τ ij.

64,65
The obtained results are presented graphically for aq (MDEA + 2-MPZ), aq (TMSO 2 + 2-MPZ), and aq ([bmim][Ac] + 2-MPZ) in Figures 4, 5, and 6, respectively. The mole fraction of water is 0.7 for the presented data. The activity coefficients of 2-MPZ are found to be very less when compared to MDEA and H 2 O (Figure 4a). The temperature dependency of the parameter is observed not to be high. Also, as a function of the MDEA concentration, the activity coefficients of 2-MPZ were found to show an inverse relationship. However, for both MDEA and H 2 O, it does not change much. Insignificant deviations of H E (excess enthalpy) and G E (excess Gibbs free energy) with respect to the temperature change from 303.15 to 323.15 K are observed (Figure 4b). Nevertheless, both the properties were found to have negative values that increase as a function of the MDEA concentration. The   (Figure 4c). Similarly, with respect to the TMSO 2 concentration, the activity coefficients of TMSO 2 and H 2 O are observed to decrease and increase slightly simultaneously as a function of TMSO 2 (Figure 5a). The behavior of 2-MPZ is similar to that found for the aq (MDEA + 2-MPZ) system. The behavior of all three systems with respect to chemical potential is found to be very similar (Figures 5c and 6c). The activity coefficients for [bmim][Ac] are very less when compared to 2-MPZ and H 2 O at any given concentration (Figure 6a). This is also expected due to the lesser chemisorption but greater physisorption behavior conferred by the ionic liquid to the acid gas. Also, though the H E and G E values obtained in the aq ([bmim][Ac] + 2-MPZ) system are negative but are a function of [bmim][Ac], the values are found to further decrease unlike those for the other two systems (Figure 6b).
3.6. CO 2 Solubility in Aq (MDEA + 2-MPZ), Aq (TMSO 2 + 2-MPZ), and Aq ([bmim][Ac] + 2-MPZ). The experimental measurement at the lab scale or estimation of CO 2 solubility in suitable solvents has been a key research area for application. 29,34 Considering the lower partial pressures of CO 2 in the flue gas stream, the CO 2 solubility has been predicted using the activity coefficients estimated through COSMO-RS over the temperature and pressure range of 303. 15 ). An inverse relationship between the CO 2 solubility and temperature for all of the studied solvent blends was observed (Figure 7). Further, although increasing the CO 2 pressure as well as the activator 2-MPZ concentration in all solvents resulted in an increase in CO 2 solubility (Table 11), however, in the case of the aq (MDEA + 2-MPZ) system, CO 2 solubility is observed to be almost similar over the chosen compositional range. This may be attributed to the fact that MDEA is a tertiary amine when compared with TMSO 2 and [bmim][Ac], which have higher CO 2 solubility. Hence, with the increasing concentration of 2-MPZ in the aqueous blends of (MDEA + 2-MPZ), the concentration of MDEA is also simultaneously decreased. Hence, the CO 2 solubility at high MDEA concentration can be understood to be compensated by a decrease in the 2-MPZ concentration. Decisively, the highest CO 2 solubility is observed for the aq (MDEA + 2-MPZ) concentration at 303.15 K. However, it should be also considered that there may be a deviation when the same solvents are studied experimentally for CO 2 absorption. This discrepancy may be attributed to variables affecting the process, nonideality associated with gas and liquid phases, vapor pressures, temperature, maintenance of the partial pressure in the system, etc.
3.7. Estimation of Henry's Constant in CO 2 and N 2 O and Free Energy of Solvation. Henry's constant signifies the physical solubility conferred by any solvent selectively to a gas that can either be measured experimentally or through the mathematical expressions available in the literature. 68 H CO 2 or H N 2 O is a contributive property through misfit in the intermolecular interactions, hydrogen bonding, and van der Waals forces of attraction. The estimation of Henry's law coefficient involves the estimation of solvation free energies, which are  further related to the chemical potential associated with the gas and solvent at a specified temperature and pressure ( Table 2). The calculated results are reported in Figure 8 and Table 12 for all of the systems under study. A low value of Henry's law coefficient indicates a higher CO 2 solubility. It can thus be inferred from the obtained results that increasing the concentration of 2-MPZ from ≈0.5 to ≈1.5 mol·kg −1 , for all base solvents of TMSO 2 and [bmim][Ac], yields an increase of physical solubility, whereas the maximum physical solubility in the case of aq (MDEA + 2-MPZ) is obtained at a 1.008 mol·kg −1 concentration of 2-MPZ. Increasing the 2-MPZ concentration further in the aq (MDEA + 2-MPZ) system results in a decrease in physical solubility that may be owing to the fact that the number of amino groups in the overall blend has increased, which majorly contributes to chemical solubility. Further, the solubilities are found to be much higher at low temperatures for all systems. Henry's law coefficients for CO 2  3.8. Estimation of the Dissociation Constant (pK a ) of MDEA and 2-MPZ. Proton exchange is considered to be one of the major reactions that occur during the interaction between CO 2 and amines. 2,69 This proton exchange reaction rate constant is often described using the dissociation constant (pK a ) of the reaction. The −ve log of pK a of conjugate acid reveals the extent of basicity offered by the chosen solvent. The above property can be either measured experimentally using the acid−base titration method or can also be predicted using quantum methods. The dissociation constants of 2-MPZ and MDEA have been calculated using COSMO. Initially, single conformers with ground-state energy were selected. The selected geometry was then edited to create a cationic structure of the same. Further, this cationic structure was optimized to calculate the energy using Turbomole software. The generated cationic structures along with the optimized energy are presented in Figure 9. Based on the free-energy change in any molecule and corresponding cationic structure, the pK a values were estimated in water, acetonitrile, and tetrahydrofuran solvents. The predicted pK a values of MDEA and 2-MPZ are given in Table 13 at 25°C. Simulated results show that the pK a values of 2-MPZ are relatively higher compared with MDEA in all solvents, demonstrating it to have the possibility of enhanced CO 2 solubility.

CONCLUSIONS
A comprehensive thermodynamic analysis through the COSMO-RS theory has been carried out for the proposed enhanced CO 2 solubility by 2-methylpiperazine in aqueous solvents of N-methyldiethanolamine, sulfolane, and 1-butyl-2-methylimidazolium acetate.