Gibbs energy minimization thermodynamics for <scp> CO ₂ </scp> capture by aqueous amines: Conservation laws, concentration variables, and parameter consistency

Abstract Gibbs‐energy‐minimization thermodynamic models (GEM–TMs) provide a rigorous framework for computing chemical equilibrium in aqueous‐amine capture systems, but published implementations and fitting practices are often inconsistent. This work shows that many reported discrepancies in fitted equilibrium constants, activity coefficients, and predicted speciation arise from (i) incomplete or rank‐deficient mass/moiety conservation constraints, (ii) inconsistent use of concentration variables and standard states, and (iii) ill‐posed strategies that co‐fit equilibrium constants (or standard‐state properties) and activity‐coefficient parameters to the same solubility data. We give a transparent treatment of the nonstoichiometric (NS) and stoichiometric GEM–TM formulations, clarify their distinction from numerical solution algorithms, and derive exact conversions of equilibrium constants and standard chemical potentials across molality, mole‐fraction, and molarity conventions for chemical potential models based on the infinite‐dilution reference state. Using aqueous methyldiethanolamine (MDEA) and monoethanolamine (MEA) as representative systems, we demonstrate that distinct formulations and parameterizations can reproduce –loading data while implying substantially different equilibrium constants, activity coefficients, and speciation, thereby exposing non‐unique parameter fits and compensation effects. All case studies are implemented in Reaktoro, a free open‐source Gibbs‐energy minimization framework that enables fully transparent model construction, constraint enforcement, and reproducible calculations. Implications for machine‐learning (ML) surrogates are discussed in the context of thermodynamically consistent equilibrium baselines.