https://ifp.hal.science/hal-03499646Tsanas, ChristosChristosTsanasIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesde Hemptinne, Jean-CharlesJean-Charlesde HemptinneIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesMougin, PascalPascalMouginIFPEN - IFP Energies nouvelles - IFPEN - IFP Energies nouvellesCalculation of Phase and Chemical Equilibrium for Multiple Ion-Containing Phases Including Stability AnalysisHAL CCSD2022phase equilibriumchemical equilibriumflash algorithmliquid-liquid equilibriumelectrolytes[PHYS] Physics [physics][CHIM] Chemical SciencesBelli, Catherine2021-12-21 15:18:152021-12-21 18:28:192021-12-21 18:28:19enJournal articleshttps://ifp.hal.science/hal-03499646/document10.1016/j.ces.2021.117174application/pdf1A reactive flash algorithm based on the work of Gautam, Seider, White (Gautam and Seider, 1979a, 1979b, 1979c; White and Seider, 1981) and Barreau et al. (2006) is presented. The original algorithm minimizes the total Gibbs energy of the system with respect to an elemental material balance constraint and the overall system electroneutrality. In our approach, we perform the minimization taking into account the charge/electric potential contribution to the Gibbs energy to allow application to multiple electrolyte phases. Moreover, the composition derivatives of fugacity coefficients are not ignored in the derivation of the working equations. Finally, the need of the overall electroneutrality equation is discussed. The algorithm is applied to a demixing amine system (water/DEEA/MAPA) with carbon dioxide, modeled with eNRTL for both phases (Mouhoubi et al., 2020). In LLE conditions, both phases are electroneutral, chemical reaction equilibrium is satisfied and phase equilibrium of each molecule and ion individually has been established. Stability analysis based on the electrochemical potential allows the determination of an electroneutral ion-containing phase. Although calculations are compared with experimental data, the purpose of the work was not the parameterization of the model, and therefore its accuracy is not addressed. Since our implementation of eNRTL did not satisfy the Gibbs-Duhem equation, its composition derivatives were incompatible with the algorithm and certain terms had to be ignored. Even though the algorithm is generally expected to converge quadratically, convergence in this work resembles the one of a partial Newton method due to the omitted derivative terms.