Screening of Ionic Liquids and Deep Eutectic Solvents for Physical CO2 Absorption by Soft-SAFT Using Key Performance Indicators

The efficient screening of solvents for CO2 capture requires a reliable and robust equation of state to characterize and compare their thermophysical behavior for the desired application. In this work, the potentiality of 14 ionic liquids (ILs) and 7 deep eutectic solvents (DESs) for CO2 capture was examined using softSAFT as a modeling tool for the screening of these solvents based on key process indicators, namely, cyclic working capacity, enthalpy of desorption, and CO2 diffusion coefficient. Once the models were assessed versus experimental data, soft-SAFT was used as a predictive tool to calculate the thermophysical properties needed for evaluating their performance. Results demonstrate that under the same operating conditions, ILs have a far superior performance than DESs primarily in terms of amount of CO2 captured, being at least two-folds more than that captured using DESs. The screening tool revealed that among all the examined solvents and conditions, [C4 py][NTf2] is the most promising solvent for physical CO2 capture. The collection of the acquired results confirms the reliability of the soft-SAFT EoS as an attractive and valuable screening tool for CO2 capture and process modeling.


■ INTRODUCTION
Currently, the global energy market is substantially dependent on fossil fuels (i.e., coal, oil, and natural gas) as primary energy resources, with a dominating share of 84.3% through 2019. 1 The price paid for this behemoth dependency is the emergence of dire environmental issues manifested in global warming phenomenon and climate change, posing a threat to the sustenance of humanity on the planet in the long term. The primary precursor for these issues is the increasing global levels of carbon dioxide (CO 2 ) emissions, reaching 33.3 gigatons (Gt) by 2019, with almost 65% of these emissions generated solely from fossil fuel consumption either for power generation or industrial applications. 2 Nowadays, carbon capture, utilization, and storage (CCUS) presents itself as the most feasible solution to this global issue in the near-and medium-terms. 3 The availability of an efficient and economically viable CO 2 capture technology is one of the cornerstones in the CCUS supply chain. 4 The absorption of CO 2 using aqueous solutions of alkanolamines, specifically monoethanolamine (MEA), is considered the industry golden standard, with the intrinsic advantage of being an "end-of-pipe" technology already implemented for purification and separation in other industrial applications. 5 Even though alkanolamines, with MEA as a benchmark, possess a high CO 2 absorption capacity, reactivity, and selectivity toward CO 2 , their large-scale deployment for CO 2 capture is still marred mainly by (1) the cost intensive solvent regeneration energy requirements, (2) degradability, either at the high temperature regeneration conditions or in the presence of oxygenated compounds commonly found in flue gases, (3) solvent losses due to high solvent volatility, and (4) high corrosivity jeopardizing the structural integrity of processing facilities. 4−6 These drawbacks are directly linked to their thermophysical properties such as high CO 2 absorption enthalpy, low thermal and chemical stability, high vapor pressure, and corrosive nature. 7 Over the past decades, task-specific designed solvents such as ionic liquids (ILs), and deep eutectic solvents (DESs) have emerged as potential replacements to conventional aqueous amines for CO 2 capture. Both types of solvents belong to the low-transition-temperature mixtures family (LTTM), wherein a liquid substance is formed at low temperatures from a combination of individual components/species that otherwise would be solid at these conditions. 8,9 The prime advantage of these alternative solvents is their tunability feature, as their synthesis is endowed with an additional degree of freedom due to the numerous combinations of possible starting materials. 9−11 The proper selection of the nature of individual components, and molar ratio in the case of DESs, allows a targeted synthesis of the solvent to specifically tailor unique thermophysical properties suited for a particular application. Among their many unique properties, the most attractive for CO 2 capture are their high chemical and thermal stability, and extremely low volatility, effectively limiting solvent losses and makeup.
A plethora of reviews exist in the literature compiling the extensive lab-scale experimental efforts toward the synthesis and application of ILs 6,7,9,12−16 and DESs 9,17−22 for CO 2 capture. Clearly, any technology newly developed for carbon capture should target improved performance to existing ones in technical, economic, and environmental aspects. 23,24 Nonetheless, the reported effectiveness of these solvents is often asserted with limited overall characteristics such as CO 2 solubility. 25 Additionally, some thermophysical properties of relevance for the design of CO 2 capture processes such as selectivity, enthalpy of absorption, heat capacity, diffusivity, viscosity, and surface tension are usually excluded from solvent performance evaluation. Thus, the conclusions drawn from these evaluations are misleading, as trade-offs between other important processing aspects affecting process design and operation such as heat of regeneration and mass transfer are overlooked. 26 The proper assessment of the potentiality of these solvents in a holistic manner would entail obtaining all relevant thermophysical properties at all industrially representative operations conditions.
Considering the wide array of synthesized ILs and DESs for CO 2 capture, experimentally acquiring all relevant information would be an unfeasible feat. Thus, it is imperative to develop a rapid and reliable screening procedure directly linking molecular characteristics of ILs and DESs to their overall carbon capture process performance. This can be accomplished through a reliable and appropriate thermodynamic model capable of reproducing the thermodynamic behavior of pure solvents and their mixtures with CO 2 , integrated with rudimentary engineering calculations to assess the performance of these solvents in terms of key performance indicators (KPIs) such as solvent cyclic working capacity and energy of regeneration, among others. Robust screening tools of the sort have been successfully developed for the screening of various materials for CO 2 capture, such as conventional aqueous alkanolamines and their blends, 27,28 novel water-free hybrid solvents, 29,30 and adsorbents, 30−32 relying on sophisticated molecular modeling tools such as molecular-based equations of state (EoSs) and molecular simulation techniques to fully characterize the thermodynamic behavior of the CO 2 capture media.
In the case of screening tools developed for ILs and DESs for CO 2 capture, several contributions attempted such an endeavor, 25,33−52 with the limiting factor of their framework being the thermodynamic model of choice. These works relied on a variety of available models to obtain necessary thermodynamic information on these solvents, including conventional semiempirical correlations, 25,33,44,46−50 artificial intelligence (AI) techniques, 34,52,53 quantum chemistry calculations using the conductor like screening model for real solvents (COSMO-RS/SAC), [35][36][37][38][39][40]42,54 or a combination of these techniques. 43,45 Regression modeling approaches suffer from poor predictive accuracy outside the range of data used for fitting, which can be enhanced with the utilization of an extensive amount of experimental data that might not be available. Moreover, they lack robustness as the full characterization of a single solvent would entail several models developed for each property of interest. Conversely, the COSMO-RS approach can be deemed predictive without need to adjust parameters to experimental data; however, in some cases the accuracy of the results is qualitatively compared to experimental data. The primary application for these models for solvent screening and evaluation has been limited to the confines of CO 2 solubility and enthalpy of absorption. However, these models have the potential to predict other thermophysical properties of importance for solvent screening (i.e., diffusivity, viscosity, surface tension, etc.) once integrated with other modeling approaches. For example, COSMO-RS have been used to obtain molecular descriptors for a wide variety of ILs, which integrated with machine learning algorithms, providing accurate predictions for viscosities of pure ILs. 55 The reader is referred to the comprehensive reviews of Vega et al. 10 and Alkhatib et al. 11 on the thermodynamic models applied in recent years on the characterization of ILs and DESs and criteria for model selection.
Alternatively, equations of state, particularly those based on the Statistical Associating Fluid Theory (SAFT), 56−59 provide a viable alternative for the development of solvent screening tools owing to their outstanding ability in capturing the physical nature of complex fluids in a representative manner. Additionally, these models are capable of predicting a full range of information needed for accurate process design and modeling, along with offering additional fundamental insight and understanding on the thermodynamic behavior of these solvents, encompassed in a single model. The strength of SAFT-based models stems from their strong theoretical framework deeply rooted in statistical mechanics based on the work of Wertheim's first-order thermodynamic perturbation theory (TPT1) of association. 60−63 This framework allows the explicit consideration of governing intermolecular interactions, such as dispersive, associating, polar, and others, expanding the range of fluids that can be modeled, being suitable to accurately describe the complex nature of ILs and DESs with the correct balance of accuracy, predictive power, and robustness. 10,11,64 Among the wide variety of SAFT-based EoSs, soft-SAFT by Blas and Vega 65,66 stands out for its robustness and accuracy in modeling the complex thermodynamic behavior of ILs 10,64,67−75 and DESs 11,76−78 and their wide range of applications. Moreover, soft-SAFT EoS has been successfully employed for the screening of a variety of solvents for CO 2 capture. [27][28][29]31 In this contribution, we demonstrate the feasibility of employing a molecular-based EoS, the soft-SAFT EoS, 65,66 for the screening of physical solvents (i.e., ionic liquids and deep eutectic solvents) for CO 2 capture at industrially relevant conditions. This presents a key step toward the development of a suitable screening tool to identify the most promising solvents for CO 2 capture as well as for a more reliable process design simulation. The modeling tool is used to fully characterize the thermodynamic behavior of the pure solvents and their mixtures with CO 2 with a limited number of adjustable parameters. The soft-SAFT molecular models characterizing the thermodynamic behavior of pure ILs and DESs, and their mixtures with CO 2 , are transferred from previous contributions, while those describing the viscosity of pure solvents are developed in this work. These models are subsequently used in a predictive manner to describe key physical properties for CO 2 capture, integrated with simple engineering calculations to assess the potentiality of these solvents using KPIs such as cyclic working capacity, heat of regeneration, viscosity, diffusivity, and solvent consumption.
The reference term (a ref ) accounts for the contribution due to segment−segment repulsive and dispersive interactions for a Lennard-Jones (LJ) fluid in a single expression, computed from the LJ EoS of Johnson and co-workers. 79 The chain term (a chain ) represents the contribution arising from the formation of chains by connecting individual segments. The association term (a assoc ) exemplifies the contribution attributed to strong and highly directional short-range interactions such as those arising from the formation of hydrogen bonds. The expression for these two terms are obtained directly from Wertheim's TPT1, 60−63 being the same across all SAFT-based models.
Lastly, in the case of polar interactions, an additional term is added (a polar ) to explicitly account for the contributions arising from polar forces such as dipolar and quadrupolar interactions. The expression for the polar term is based on the theory of Gubbins and Twu, 80,81 and its subsequent extension to chainlike fluids by Jog and Chapman. 82, 83 The thermodynamic modeling of fluids within the soft-SAFT framework entails acquiring a set of molecular parameters descriptive of major structural and energetic features of the fluid. The characterization of nonassociating fluids requires three parameters, namely, segment diameter (σ i ), chain length (m i ), and dispersive energy between segments (ε i ). In the case of associating fluids, two additional parameters are included, related to the volume (k HB ) and energy (ε HB ) of association. Lastly, explicitly modeling polar interactions requires the fraction of polar segments in the molecule (x p ) and the effective dipolar (D) or quadrupolar moment (Q). These molecular parameters are typically treated as adjustable parameters fitted using experimental saturated liquid density and vapor pressure of the pure fluid, while the dipolar and/or quadrupolar moment is taken from experimental information, and x p is set a priori based on physical arguments. 84,85 The extension of the model to multicomponent mixtures is straightforward for the chain, association, and polar terms, as they are explicitly written for mixtures. Alternatively, extending the reference term to multicomponent mixtures is done using the generalized Lorentz−Berthelot (LB) mixing rules, used to calculate the crossed diameter (η ij ) and energy (ε ij ): where η ij and ξ ij (η ij = 1 − l ij and ξ ij = 1 − k ij in classical EoSs) are the adjustable size and energy binary interaction parameters, respectively. For mixtures of associating molecules, cross-association interactions between different molecules or different functional groups within the same molecules are computed using combining rules in a manner analogous to those in eqs 2a and 2b. The volume of cross-association is calculated by the mean arithmetic radius of the association site, while the energy of cross-association is calculated through the modified geometric average as follows: jj HB HB HB 3 3 3 i k j j j j j j j j j y where α ij HB is a binary parameter that generalizes the energy of cross-association, fitted to binary mixture VLE data, if needed. The model is used in a fully predictive manner whenever the binary parameters (η ij , ξ ij , and α ij HB ) are set to unity, without fitting to binary mixture vapor−liquid equilibrium (VLE) data.
Free-Volume Theory (FVT). Within the soft-SAFT framework, the viscosity of pure fluids is obtained using the free-volume theory (FVT), 86,87 wherein, the dynamic viscosity of a pure fluid (μ) is computed as the sum of two terms: where μ 0 is the viscosity of the dilute gas, which can be omitted for solvents with low vapor pressure due to its small contribution to the total viscosity, 88 and Δμ is the correction term used to account for the viscosity of the dense-state. By omitting the dilute gas term, the expression for the total viscosity becomes solely dependent on the dense-correction term which is computed based on the expression proposed by Allal and co-workers, 86,87 relating the microstructure and the free space among the molecules with the viscosity of the fluid using thermodynamic variables such as pressure, temperature, and density. The FVT dense term contains three adjustable parameters, α, B, and L v , which are typically fitted to pure fluid or mixture viscosity data. Additional details on each FVT expression and their implementation within the framework of the soft-SAFT equation can be found in previous contributions. 89,90 Extending viscosity calculations to multicomponent mixtures is done using linear compositional rules for each FVT adjustable parameter, 90 as such: It should be noted that with these mixing rules, the model is extended to mixtures in a fully predictive manner, as no additional binary parameters are fitted to experimental viscosity data of the binary mixture. A detailed discussion will be provided in the results section.

Journal of Chemical & Engineering Data
pubs.acs.org/jced Article Key Performance Indicators for solvent screening. To properly evaluate the performance of ILs and DESs for CO 2 capture, the effectiveness of these solvents on three important elements with direct impact on the techno-economic feasibility of the absorption process is assessed. These elements are CO 2 solubility, mass transfer, and regeneration energy. For each element, a selection criterion is proposed based on fundamental engineering calculations, with the required input thermophysical properties directly predicted from soft-SAFT models developed for these solvents and their binary mixtures with CO 2 . These selection criteria are subsequently used to capture the trade-offs in assessing the potentiality of these solvents in a comparative manner. As all the examined solvents present a physical absorption behavior for CO 2 , their performance is assessed on the basis of a physical CO 2 absorption for precombustion CO 2 capture. In this process, the absorption of CO 2 from the feed flue gas takes place in an absorption column operated at low temperatures (298−313 K) and high pressures (1000−3000 kPa). The regeneration of the solvent occurs either by decreasing the operating pressure of the stripping column using a pressure swing absorption (PSA), or by increasing the temperature using a temperature swing absorption (TSA). For this work, it was opted to assess the performance of these solvents using a PSA process. It should be noted that the selectivity and effect of impurities on the performance of these solvents, though important for a final decision on the solvents and the design of CO 2 capture systems, are not considered herein, as it is out of the scope of the present work.
In the case of CO 2 solubility, Henry's constant is typically used as a quick calculation to screen physical solvents. However, a more apt criterion to evaluate solvents based on CO 2 solubility is the solvent cyclic working capacity (Δα CO 2 . The cyclic working capacity, on the basis of moles of CO 2 captured per mole of solvent consumed, is defined as the difference between the amount of CO 2 absorbed in the solvent at absorption conditions and the amount of CO 2 remaining in the solvent at desorption conditions, and it is directly related to the amount of CO 2 effectively captured from the feed flue gas. In the case of effectiveness of mass transfer, the simplest criterion to capture the effect of solvent thermophysical properties on mass transfer would be in terms of diffusion coefficient of CO 2 (D CO 2 ), as it is inherently and inversely related to the viscosity of the solvent. A variety of methods are available in the literature for the prediction of diffusion coefficient of CO 2 from thermophysical properties of solvents, 91,92 yet in this work, the diffusion coefficient of CO 2 is determined using the Wilke-Chang expression, 93 expressed as where D CO 2 is the diffusion coefficient of CO 2 in the examined solvent, MW is the molar mass of the solvent in g mol −1 , μ is the viscosity of the solvent in cP, V m is the molar volume of CO 2 in cm 3 ·g −1 , and φ is the interaction parameter with a value of 7.5 for ILs. 91,92 We have assumed that the same interaction parameter applies to DESs, as they are analogues to ILs. Evaluating the energy consumption required for solvent regeneration is the last key factor in evaluating the performance of solvents for CO 2 capture. As previously mentioned, the performance of the examined ILs and DESs for CO 2 capture is evaluated using a PSA process at T = 298 K. The energy consumption for solvent regeneration (Q) using a PSA process, for which the temperature of absorption and desorption is constant, is expressed as 33,44,48,49,94,95 where, Q des is the energy consumed to vaporize CO 2 from liquid to vapor phase at the desorption conditions, which is the same as the enthalpy of dissolution (H dis ). Q ex is the energy consumption due to the difference of excess enthalpy (ΔH ex ) between the streams at absorption and desorption conditions. It was previously demonstrated, 44 that the contribution of the energy consumption due to excess enthalpy was generally less than 5%, effectively simplifying the energy needed for solvent regeneration to be fully dependent on the enthalpy of dissolution, which can be derived from the Clausius− Clapeyron equation using information obtained from the soft-SAFT EoS, as It should be noted, that the negative value of the enthalpy of absorption is equal to the enthalpy of desorption.
Coarse-Grain soft-SAFT Molecular Models of Ionic Liquids and Deep Eutectic Solvents. The key step in applying SAFT-based EoSs to real fluids is the proposition of a simplified coarse-grain molecular model of the examined fluid, capturing its basic structural and energetic features. Following our previous work, carbon dioxide was modeled as a homonuclear chainlike LJ fluid, explicitly considering its quadrupole moment. The fraction of quadrupolar segments was fixed a priori, ( = x p 1 3 ), based on the geometry of the molecule, while its quadrupole moment was included in the fitting procedure, yielding a value relatively similar to the experimental one. 96,97 The development of coarse-grain molecular models for ILs and DESs is relatively similar due to their parallel molecular nature, being mixtures of individual components/species bounded by some form of physical strong and highly directional interactions, well captured by the association term in SAFT-type equations. For these substances, the mixture of their individual components can be modeled either as a single substance (pseudopure component approach), or as a binary mixture of the individual components (individual-component approach). 10,11,98 A graphical representation of the soft-SAFT molecular models for selected ILs and DESs examined in this work is depicted in Figure 1. Additionally, a summary of the ILs and DESs modeled is included in Table 1.
The ILs were modeled using the pseudopure component approach as associating single chainlike fluids with the attractive interactions between the counterions mimicked through the addition of specific association sites (defined a priori) dependent on the structure and charge distribution of the IL. For ILs with [BF 4 anions, a single association site of type A was included in their models as the neutral counterions pair is quasi-spherical with a very localized charge, with A−A interactions allowed in the model due to the dual nature of the association site. 68 were modeled with two association sites, one site of type A representing the interaction between the anion and the cation, and one site of type B, mimicking the delocalized charge due to the oxygen molecule on the anion, with only A−B interactions allowed in the model. 99,100 In a similar fashion, ILs with [NTf 2 ] − , [C 4 F 9 SO 3 ] − , and [N-(C 2 F 5 SO 2 ) 2 ] − anions were modeled with three association sites, one site of type A and two sites of type B, as a consequence of a higher degree of charge delocalization in the anion, again only allowing A−B interactions in the model. 69,72,73 The soft-SAFT molecular parameters for the examined ILs were transferred from previous contributions, 68,69,72,73,99,101 and are included in Table S1 in the Supporting Information for completeness.
The DESs were modeled using the individual-component approach. 11 In this manner, each individual component is characterized by a separate set of parameters and an additional binary interaction parameter accounting for deviations in size or energy between both components. The hydrogen-bond acceptors (HBAs) such as tetraalkylammonium chloride salts  The models are subsequently used in a predictive manner to assess the potentiality of these solvents in terms of KPIs such as cyclic working capacity, enthalpy of absorption, and diffusivity, capturing the trade-offs between these properties on the effectiveness of ILs and DESs for CO 2 capture. Viscosity Calculations of pure ILs and DESs. The description of the liquid density at atmospheric pressure for the pure solvents as obtained from soft-SAFT with the transferred molecular parameters, compared to experimental data for ILs 72,99,103−107 and DESs 108−111 is shown in Figure S1 in the Supporting Information, highlighting the quantitative accuracy of the models used in this work for ILs and DESs.
The viscosities of the examined solvents were obtained in this work using the FVT treatment combined with soft-SAFT, requiring three additional molecular parameters regressed using pure viscosity experimental data. The FVT treatment parametrization procedure for ILs is straightforward, modeled using the pseudopure component approach, fitting the FVT parameters to available viscosity experimental data of pure ILs, using the previously regressed soft-SAFT molecular parameters provided in Table S1 in the Supporting Information. In the case of DESs, modeled using the individual-component approach as binary mixtures, the fitting procedure was modified, as some of the individual components are solid at a wide range of conditions, with no experimental viscosity data. Hence, the spider-web methodology 112 was utilized to fit the FVT parameters for the solid components of the examined DESs. With this approach, the FVT parameters were fitted simultaneously to the viscosity of pure compounds and mixtures, assuming the FVT parameters of one of the compounds (i.e., EG which is a liquid) in the mixture are available. The sequence for parametrizing the FVT parameters for the individual components in the investigated DESs is shown in Figure 2. It should be noted that the sequence provided in this figure relies on using the previously regressed soft-SAFT molecular parameters in a transferable manner in order to obtain thermodynamic information needed for FVT treatment such as temperature, pressure, and density. Additionally, the optimization for FVT parameters is carried out not only attending to the minimal deviation respect to viscosity data, but also looking for meaningful trends for the compounds of the same family. In this regard, the tetraalkylammonium salts present a linear increase of the barrier energy α and a second-order polynomial and linear decrease of the freevolume overlap B and correlation length L v , respectively, with the molecular weight. This type of trend is similar to that found in previous contributions for other families of compounds. 90 Figure 2. Schematic procedure for the sequence of parametrizing free-volume theory (FVT) parameters to describe the viscosity of the individual components of pure DESs examined in this work. The green boxes refer to either molecular parameters (MP), or FVT parameters of the examined substances, included in Tables S1 and S2 in the Supporting Information, while the blue boxes refer to experimental viscosity data (μ) for the examined pure fluid (EG) or DES. See notation in Table 1.

Journal of Chemical & Engineering Data pubs.acs.org/jced Article
With the FVT parameters obtained in this work for ILs and DESs, included in Table S2 in the Supporting Information, the calculations for the pure solvents viscosities from FVT + soft-SAFT are in excellent agreement with experimental data for DESs, 113−119 and ILs 106,120−130 studied in this work, as depicted in Figure 3 and Figure 4, respectively.
Solubility of CO 2 in the Selected ILs and DES. Once the thermodynamic behavior of the pure solvents was assessed, the ability of soft-SAFT in capturing the performance for the solubility of CO 2 in these solvents was studied next. A summary of the binary interaction parameters between CO 2 and ILs, and CO 2 and the individual components of each DES modeled in this work, is included in Table S3 in the Supporting Information, either transferred from previous works 68,71,73,101,102 or developed herein. These binary inter-action parameters were obtained, when needed, by fitting to an intermediate isotherm and transferred to other conditions in a predictive manner. It should be noted, that in most cases, the binary parameters were fixed to unity, with the CO 2 solubility in these solvents directly predicted from the molecular parameters of the pure fluids (see Table S3), reinforcing the robustness of the soft-SAFT model and the molecular parameters. The soft-SAFT computations for the solubility of CO 2 in ILs and DESs investigated in this work are depicted in Figure 5, validating the robustness of soft-SAFT, accurately capturing the trends for the experimental CO 2 solubility in ILs 107,131−140 and DESs, 111,141−143 respectively.
Evaluation of ILs and DESs for CO 2 Capture. The performance of the selected ILs and DESs for CO 2 capture is evaluated next based on three important factors affecting the   Table S2 in the Supporting Information. Soft-SAFT + FVT calculations (lines) compared to experimental data 106,120−130 (symbols).

Journal of Chemical & Engineering Data
pubs.acs.org/jced Article techno-economic feasibility of CO 2 capture systems, namely, CO 2 absorption capacity, mass transfer, and energy of regeneration. All calculations were performed using soft-SAFT models in a fully predictive manner. CO 2 Absorption Capacity. One of the key factors in deciding the choice of solvent for CO 2 capture is its affinity to CO 2 as this dictates the efficiency of the solvent in removing CO 2 from the incoming feed gas and has a direct impact on the capital and operating expenditure for the process. Typically, the physical CO 2 absorption process occurs at low temperatures and high pressures; as such, the solubility of CO 2 in the examined solvents was evaluated at T = 298 K, and in the pressure range of P = 0.1−2500 kPa. The absorption loading, represented as mole of CO 2 per mole of solvent of the examined ILs and DESs, was directly obtained from soft-SAFT in a fully predictive manner, and the results are depicted in Figure 6. It can be observed that, under the same absorption conditions, the majority of the examined ILs, except [ Figure 6a−c. For the examined ILs, the solubility of CO 2 is affected by both the chain length and type of the cation, and the type of anion. 3,44 For ILs with the same cation, for example, [C 2 mim] + or [C 4 mim] + , depicted in Figure 6a−b, the increase in CO 2 solubility is associated with an increase in the molecular weight of the anion. For [C 2 mim] + -based ILs, the increase in CO 2 solubility with the different anions is in the order of [CF 3 Figure 6d. The CO 2 solubility in DESs is also affected by the type of HBD and HBA, analogues to the cation and anion in ILs. For the same HBD, for example, LA and LevA, increasing the chain length of the HBA from tetraethylammonium halide salt to tetrabutylammonium halide almost doubled the CO 2 solubility in the examined DESs. In contrast, for the same HBA, for example, choline chloride [Ch][Cl], the choice of HBD had an effect on the CO 2 solubility with [Ch][Cl]/GLY having twice the absorption capacity of [Ch][Cl]/EG. This might be related to the CO 2 solubility in the pure HBD, as glycerol has a higher CO 2 solubility than ethylene glycol. It should be noted that the conclusions drawn here on the effect of the individual components of DESs on CO 2 solubility are only based on the solvents examined in this work. A larger variety of DESs at different molar ratios should be studied to provide more insights.
The cyclic working capacity of each solvent examined in this work was determined using eq 6. The chosen absorption conditions are T = 298 K and two operating pressures, at 1000 and 2500 kPa, to examine the effect of absorption conditions on cyclic working capacity, while the chosen desorption conditions are T = 298 K and P = 10 kPa. Moreover, the amount of solvent required for CO 2 capture at these conditions has been evaluated on the basis of tons of solvent per ton of CO 2 . These results were obtained from the predictions shown in Figure 6 using soft-SAFT EoS. Depicted in Figure 7a,b, are the cyclic working capacity and solvent consumption for the examined ILs and DESs. It is observed that increasing the absorption operating pressure from 1000 to     2), at T = 298 K, P abs = 1000 kPa, and P des = 10 kPa, is compared to that for traditional aqueous alkanolamines. On the bais of results from a previous contribution, 27 the cyclic working capacity of 30 wt % aqueous MDEA, and a 20/10 wt % aqueous blend of MDEA + PZ at those conditions are ca. 0.55 and 0.75 mol CO 2 ·mol −1 amine, respectively, while the amount of alkanolamine consumed to capture 1 ton of CO 2 is 4.97 and 0.98 tons. These values are far more superior to those obtained for the promising IL and DES. This demonstrates that the feasibility of these alternative solvents is not in terms of amount of CO 2 captured, so reaching a level comparable to that obtained with aqueous alkanolamines would require a higher operating pressure to increase the amount of CO 2 captured. Yet, this is not necessarily a negative point, as it is observed that the physical absorption process with these solvents tends to become more efficient at higher absorption operating pressures, by increasing the amount of CO 2 captured and decreasing the amount of solvent consumed. This will have direct implications on the capital costs of the processing units and operating costs for heating and cooling requirements being reduced due to the lower solvent flow rate. 25,46 In contrast, the effect of increasing the absorption operating pressure will manifest in increased sizing and utility consumption of the compression unit, which has been demonstrated to be dependent on the operating conditions rather than the type of solvent. 46 Effect on Mass Transfer. Mass transfer is another vital processing aspect of key importance for the design of CO 2 capture units, due to its effect on the capital cost associated with the height of the absorption column. The inclusion of this factor is typically neglected in assessments based on equilibrium and energy balances of the process. The work of Mota-Martinez and co-workers 25,46 established that diffusivity and viscosity are among the most important thermodynamic properties directly impacting the design of the column based on the effectiveness of mass transfer. The effect of viscosity on the economics of the process has a higher contribution to the capital cost in terms of height of the absorber, than the operating cost in terms of energy required for solvent pumping.
All the required thermophysical properties to predict the diffusion coefficient of CO 2 in ILs and DESs, using eq 7, have been obtained from soft-SAFT at T = 298 K, and in the absorption pressure range of 1000 to 2500 kPa, to examine the effect of absorption pressure on pure solvent viscosity and CO 2 diffusivity in the solvent, as depicted in Figure 8 and Figure 9. The effect of the absorption operating pressure on the viscosity Journal of Chemical & Engineering Data pubs.acs.org/jced Article of pure ILs and DESs is negligible as they are incompressible fluids, while the effect on the CO 2 diffusion coefficient in the examined solvents is more pronounced. The increase in the diffusion coefficient at higher absorption operating pressures implies an enhancement in the mass transfer of CO 2 and its absorption in these solvents. This will reflect on the design of the absorption column being smaller in size and subsequently reducing the capital expenditure, which has also been previously demonstrated in the case of increased cyclic working capacity. Moreover, it can be seen that the higher is the viscosity of the solvent, the lower is its CO 2 diffusion coefficient. Energy of Regeneration. The energy of desorption required for regenerating ILs and DESs examined in this work has been evaluated, using eq 9, at T = 298 K, and in the pressure range of 10−1000 kPa, to investigate the effect of desorption pressure on the desorption enthalpy, as shown in Figure 10. It is observed that for all the examined solvents, increasing the desorption pressure resulted in a reduction in the magnitude of the enthalpy of absorption. Thus, at higher regenerator operating pressures, the energy required for solvent regeneration would be lower; however, the amount of CO 2 recovered would also decrease.
The lowest enthalpy of desorption is observed for the ILs In this aspect, the advantage of ILs and DESs compared to conventional aqueous alkanolamines manifests itself. The enthalpy of desorption for a 30 wt % aqueous MDEA solution is 52.5 kJ·mol −1 CO 2 , 144 being four times higher than that observed for ILs and DESs with the lowest desorption enthalpy examined in this work. Additionally, the presence of large quantities of water as a cosolvent with alkanolamines adds a parasitic load on the amount of energy required for regeneration due to the high heat capacity and enthalpy of vaporization of water. 29 Comparative Assessment of ILs and DESs for CO 2 Capture. The effective assessment for the potentiality of ILs and DESs examined in this work for CO 2 capture requires taking into account trade-offs between the selected KPIs, namely, cyclic working capacity, CO 2 diffusion coefficient, and enthalpy of desorption, due to their influence on capital and operating costs, as previously highlighted. The solvent performance is assessed at T = 298 K and P = 2500 kPa for absorption, while the desorption occurs at T = 298 K, and in the pressure range of 10−1000 kPa. Provided in Figure 11 is the comparative assessment of ILs and DESs for CO 2 capture based on the chosen selection criteria; the optimal solvents would be those with a high cyclic working capacity, low enthalpy of desorption, and high CO 2 diffusion coefficient. The thick lines reflect the evolution of cyclic working capacity and enthalpy of desorption with a change in the desorption pressure from 10−1000 kPa, with the large spheres at the end of the line denoting desorption at 10 kPa. Moreover, the size of the sphere denotes the magnitude of the CO 2 diffusion coefficient in these solvents: the larger the better.
For the examined ILs as shown in Figure 11a, it is observed that the solvents with high working capacity and low enthalpy of desorption are [NTf 2 ] − -based ILs (i.e., [ SO 3 ]. With the inclusion on the mass transfer effect in the form of CO 2 diffusion Figure 11. Comparative assessment of (a) ILs, and (b) DESs examined in this work for CO 2 absorption. All calculations were obtained from soft-SAFT at T = 298 K, P abs = 2500 kPa, and P des = 10−1000 kPa (lines from right to left). See text for details. times the amount of CO 2 captured, and 7 times higher CO 2 diffusion coefficient, indicating that this solvent is far more efficient than the best performing DES. In fact, the majority of the examined ILs, even the poorest ones, have a better performance than the best DES in terms of working capacity.

■ CONCLUSIONS
In this work, the soft-SAFT EoS has been employed as a reliable platform for the screening of a variety of ionic liquids and deep eutectic solvents as physical solvents for CO 2 capture. The models for the pure solvents and their binary mixtures with CO 2 have been validated against available experimental data, using previously developed molecular models for the examined compounds. Subsequently, the models were used in a fully predictive manner to assess the performance of these solvents in terms of key performance indicators such as cyclic working capacity, enthalpy of desorption, and CO 2 diffusion coefficients. These selection criteria were used due to their influence on the technoeconomic feasibility of the CO 2 capture process. The inclusion of all selection criteria assisted in capturing the trade-offs between CO 2 solubility, energy of regeneration, and mass transfer. It was determined that the performance of the majority of ILs was far more superior to DESs at the examined conditions, mainly in terms of working capacity. The most promising solvent identified with the framework developed in this work was [C 4 py][NTf 2 ] IL due to its high working capacity, low enthalpy of desorption, and high CO 2 diffusion coefficient. A comparison with selected physical solvent aqueous amines at the same conditions shows that even the best performer solvents investigated in this work present lower CO 2 loading capacity and higher solvent consumption, while they show a clear advantage in terms of the enthalpy of desorption. Still for a fair detailed assessment, the best operating conditions for each solvent should be investigated separately.
Overall, this work demonstrates the promise held by SAFTbased models, specifically the soft-SAFT EoS, for the design of screening tools for solvents for CO 2 capture that assist in guiding the development and selection of the most promising systems, based on key process parameters. The procedure developed in this work has a high degree of reliability and robustness, being flexible to investigate the effects of a wide range of operating conditions on the performance of the examined solvents for CO 2 capture. An additional factor to be included in future work for a more holistic screening of solvents will be the inclusion of selectivity and the effect of impurities on the CO 2 capture performance of ILs and DESs. Also, please notice that the cost of the solvent, key in the case of ILs for large scale applications, was not considered in this work, as it was based on thermophysical properties.
As a final remark, results presented here are a showcase on how SAFT models and their application have evolved since the original publication 30 years ago, from some refined VLE calculations of complex fluids compared to simpler EoS, to the prediction of a whole set of required thermophysical properties for process design and its application to current environmental/engineering challenges.
Soft-SAFT molecular parameters, FVT parameters of pure fluids, binary interaction parameters for CO 2 + ILs/ DESs, and validation results for pure solvent liquid density (PDF)