Carbon capture assessment for phosphonium-based ionic liquids

RODRIGUEZ REARTES, SABRINA BELÉN; OUEDGHIRI BEN OTMANE, FÁTIMA; LLOVELL, FÈLIX

Barcelona, España

Congreso; 15th Mediterranean Congress of Chemical Engineering (MeCCE-15); 2023

IQS

Carbon dioxide emissions reduction has become a global concern in order to mitigate climate change. Renewable energy sources are being implemented whereas carbon capture technologies are also used and studied to reduce greenhouse gases emissions. The development of new and more efficient carbon removal technologies includes the application of ionic liquids (ILs) for CO2 capture from gas streams. ILs have kept attention due to its high absorption performance and physical features. Even though several CO2 capture technologies exist, absorption-based technology is the most widely used, present in 72% of the facilities, being amine-based solvent (i.e., MEA and MDEA) utilized in 69% of them, accounting for 55% of total CO2 capture capacity [1]. MEA and MDEA have handicaps related to its degradation during process operation and its corrosive properties. In this context, new CO2 absorbents are tested. Among them, ILs are being considered due to their low volatility, high thermal stability and non-corrosiveness. ILs can be “design” by combining different cations and anions considering the final operation conditions and the flue stream composition to treat.Particularly, some phosphonium cation/anion combinations have been studied in literature with promising results. A complete characterization of these compounds is required to select the most suitable CO2 absorber, but available experimental data is scarce; then modelling tools can assist. In this work, the potential of phosphonium-based ILs as CO2 absorbers at different operating conditions for industrial application is assessed through the soft-Statistical Association Fluid Theory (soft-SAFT) equation of state (EoS). soft-SAFT EoS accounts for the shape of the molecule, hydrogen bonding formation and polarity effects and was used in the past to study many types of families of ILs and its mixtures with success, and will be used here in combination with quantum-chemical approaches, such as Turbomole-COSMO. This last approach allows to obtain the charge distribution profiles and describe the key interactions in these compounds and help build a physically sound model. Different anions are combined with the trihexyltetradecylphosphonium cation [P66614]+ to generate ILs. The resulting ILs are characterized by a complete description of their pressure-temperature-density diagrams, and derivative properties. Transport properties, such as the viscosity, are also modeled using the Free-Volume theory coupled to soft-SAFT EoS. Then, CO2 absorption isotherms are described and compared to experimental data when available. The CO2 absorption capacity of the different ILs, considering both diluted and concentrated mixtures in CO2, is assessed through the calculation of Henry’s law constants and the solvation enthalpies and entropies at different conditions, proposing a preliminary list of potential compounds for different types of flue gases.Figure 1 illustrates the densities obtained through soft-SAFT EoS for [P66614][Cl] at different pressures with a highly agreement with the experimental data [2]. Figure 2 shows the soft-SAFT predictions for carbon dioxide absorption by [P66614][Cl] at different temperatures, presenting also a very good agreement with available experimental data. Fig. 1 - Calculated densities of [P66614][Cl] at different pressures. Lines: soft-SAFT calculations. Symbols: experimental data from [2].Fig. 2 - Predicted carbon dioxide absorption for [P66614][Cl] at different temperatures. Lines: soft-SAFT calculations. Symbols: experimental data from [3].In order to complement the molecular model-based calculations, a machine learning approach is going to be used. In particular, artificial neural network (ANN) algorithms are to be applied to predict the solubility of CO2 in ILs (primarily phosphonium-based, but could be extended to others ILs). This approach has been implemented by other researchers with success in predicting the solubility of fluorinated refrigerants in ILs. In our study, we will use a feed-forward ANN to map the relation between the properties of pure compounds (i.e, CO2 and ILs) and the solubility of CO2 in ILs. Several statistical indicators will be used to evaluate the developed ANN model including coefficient of determination (R2), absolute average relative deviation (AARD), root mean square error (RMSE), and average standard deviation (SDavg) for each output. We aim that such ANN could be employed as a prescreening tool by predicting the solubility of CO2 in new designed ILs (not considered initially to build the ANN) aimed at the capture of CO2.