Co₂ solubility measurements and modelling in amine-NMP solvent blends.

Abstract

In this study, solvent blends of monoethanolamine (MEA) or 2-(2-aminoethoxy)ethanol (DGA) with N-methyl-2-pyrrolidone (NMP) or water (H2O) were selected for investigations of carbon dioxide(CO2) solubility due to the high CO2 absorption capacities of the individual solvents. A static synthetic apparatus, consisting of a stirred equilibrium vessel and a gas reservoir, each submerged in their own temperature-controlled environment, was used to measure the CO2 solubility for the systems and conditions stated above. Isothermal solubility measurements were performed at 40 °C over a pressure range of 0.1 – 1.5 MPa for the systems of CO2 in various solvent blends. These included 20% MEA–80% NMP, 30% MEA–70% NMP, 51% DGA–49% H2O, 40% DGA–60% H2O, 30% DGA–70% H2O, 51% DGA–49% NMP, 40% DGA–60% NMP and 30% DGA–70% NMP (by mass). The recorded temperature-pressure-overall composition (T-P-z) data were converted to T-P-mole fraction (T-P-x) data. Results were displayed on pressure vs. CO2 loading (P-𝛼𝐶𝑂2) graphs andcompared to literature data. Further comparisons were made between the various solvent blends. Thermodynamic modelling of the experimental data for the DGA systems was performed using MATLAB®. Due to a solvent blend of a water-lean amine and a physical solvent, two models were fitted to the experimental data by regression of the model parameters, and the results combined and displayed on P-𝛼𝐶𝑂2 graphs with the respective experimental data. The Posey-Tapperson-Rochellemodel was used for DGA, and the Peng-Robinson equation of state (PR-EOS) with modified van der Waals-Berthelot mixing rules was used for NMP in water-lean cases. Only the Posey-Tapperson-Rochelle model was used for amine-water systems. The results indicated that the water-lean blends, MEA-NMP and DGA-NMP, have a higher CO2 loading at the same pressure when compared to the corresponding MEA-H2O and DGA-H2O blends. An increase from 30% DGA–70% H2O to 40% DGA–60% H2O (by mass) resulted in a viscosity increase of 0.65 Pa.s at 40 °C and an increase in CO2 loading of 0.079 molCO2/molamine at 0.63 MPaand 40 °C. Comparing the 30% MEA–70% NMP and 30% DGA–70% NMP (by mass) blends, it was observed that the DGA blend had an increase in CO2 loading of 0.12 molCO2/molamine at 0.24 MPa and40 °C. Thermodynamic modelling for the CO2-51DGA-49H2O system gave a root mean square error of 3.75%, an absolute average deviation (AAD) of 98.24 and an average absolute relative deviation (AARD) of 22.69%, while modelling for the CO2–51 wt% DGA–49 wt% NMP system gave a root mean square error of 0.61%, an AAD of 13.13 and an AARD of 2.94%. Based on the calculated error and AARD, regression for the Posey-Tapperson-Rochelle and PR-EOS model parameters gave the closest results to the CO2–30 wt% DGA–70 wt% NMP measured data. From this work, it was concluded that in terms of the viscosity and CO2 loading at 40 °C, the DGA-NMP blends show promise compared to the DGA-H2O and MEA-NMP blends. The 40 wt% DGA–60 wt% NMP solvent blend was the best-performing DGA-NMP blend. Further experiments to determine the changes in viscosity and CO2 loading of regenerated solvents for a range of DGA-NMP blends are recommended, and further modelling analyses for data prediction are recommended for continuation of this work.