Experimental Investigation and Modelling of Reaction and Phase Equilibrium using the Example of the Substance System n-Butanol — Acetic Acid — n-Butyl Acetate — Water

This work examines how phase-equilibrium knowledge can be leveraged to model both equilibrium and kinetics of chemical reactions in strongly non-ideal liquid mixtures, using the esterification of n-butanol with acetic acid to n-butyl acetate and water as a case study relevant to multiphase reactors and reactive separation (e.g., reactive distillation). Because literature data on ternary/quaternary phase equilibria and on reaction equilibrium/kinetics are incomplete and often based on limited starting compositions, an extensive new experimental database is generated. Reaction equilibrium is measured across a wide composition space at 80, 100, and 120 °C, including 91 homogeneous liquid equilibrium points and 33 reactive liquid–liquid equilibrium measurements, using a newly built multiphase batch reactor with online gas chromatography and complementary batch experiments coupled to proton NMR. The results show that the pseudo reaction equilibrium constant depends strongly on mixture composition, highlighting non-ideality effects. Additional phase-equilibrium work includes 49 liquid–liquid equilibrium measurements in ternary and quaternary mixtures away from reaction equilibrium (plus selected quaternary data from a circulating still). Reaction kinetics are investigated in 42 experiments using an NMR flow cell operated as a batch reactor, enabling fast, time-resolved composition measurements while varying initial composition, catalyst concentration, and temperature. Phase and reaction equilibrium are modeled thermodynamically consistently using NRTL and UNIQUAC (Gᴱ models), PC-SAFT, and COSMO-RS. For NRTL/UNIQUAC, parameter fitting shows that a single parametrization cannot simultaneously match VLE and LLE with high accuracy, and predicting ternary/quaternary equilibria from binary data is especially challenging for LLE. For reaction equilibrium, Gᴱ models fitted to VLE data best capture the composition dependence of the pseudo-equilibrium constant, while PC-SAFT—despite typically weaker phase-equilibrium predictions due to limited binary parameters—predicts reaction equilibrium well over much of composition space. COSMO-RS, fully predictive with no fitted binary parameters, performs worse for phase equilibria but yields unexpectedly high-quality reaction-equilibrium predictions, notably including prediction (rather than fitting) of the equilibrium constant. Kinetics are described over broad compositions and temperatures with a second-order Arrhenius-type approach formulated in terms of activities or fugacities from the thermodynamic models, although rate constants still show composition dependence due to changes in catalytic proton activity (modeled empirically). Finally, COSMO-RS is shown to support an alternative, thermodynamically consistent kinetic modeling route based on contact probabilities of reacting surface segments, offering a more physically intuitive framework than purely concentration-based kinetics.