A systematic and validated framework for selecting greener entrainers towards eco-efficient extractive distillation

Abstract

Extractive distillation (ED) is a key technology for separating close-boiling and azeotropic mixtures by introducing an entrainer that modifies intermolecular interactions within the mixture. Through selective affinity toward specific components, the entrainer increases the non-ideality of the system, enhances relative volatility, and eliminates close-boiling and azeotropic behavior. However, conventional entrainers such as N-methyl-2-pyrrolidone (NMP) raise significant environmental and health concerns, including reproductive toxicity. Meanwhile, many greener candidates, including biobased and greener organic entrainers, as well as natural deep eutectic solvents (NADESs), remain insufficiently evaluated for ED applications. Nevertheless, systematic and validated entrainer selection strategies, supported by reliable vapor-liquid equilibrium (VLE) data and rigorous process performance comparisons for the close-boiling methylcyclohexane-toluene mixture and the azeotropic n-hexane-ethanol mixture containing these underexplored greener entrainers, are unavailable. Therefore, this thesis aims to develop a framework for screening greener entrainers, along with generating reliable VLE data and performing process performance comparison, to identify effective alternatives to NMP for these two representative separation challenges.
To achieve this objective, a multi-stage entrainer selection methodology was developed, integrating initial screening, predictive modelling, experimental validation, and process performance comparison. In Chapter 2, potential greener entrainers were screened using an initial screening phase, which identified potential greener entrainers based on high boiling temperatures, favorable solvency, and compliance with green chemistry principles. Subsequently, key separation metrics, such as selectivity and capacity at infinite dilution, and performance index, were predicted using quantum chemical calculations (COSMO-RS). In addition, relative volatility was predicted using both COSMO-RS and group contribution methods (UNIFAC and modified UNIFAC Dortmund). Promising candidates were then evaluated experimentally through miscibility tests and relative volatility measurements. Selected greener entrainer candidates, along with benchmark entrainer NMP in both mixtures, were examined for their vapor-liquid equilibrium (VLE) data. Chapters 3 and 4 focus on generating reliable VLE data for both mixtures in the presence of each selected greener entrainer at various entrainer-to-feed ratios (E/F) and operating pressures. The experimental data were examined through thermodynamic consistency tests. Moreover, the data were regressed using activity coefficient models (NRTL and UNIQUAC) to obtain the optimum binary interaction parameters. In Chapter 5, these parameters were implemented in Aspen Plus V.12 for ED process simulations. Greener entrainers were evaluated according their performance compared to NMP using economic indicator (total annual cost) and sustainability metrics, such as energy requirements, water consumption, and CO2 emissions. In addition, the use and limits of predictive tools for ED entrainer screening were evaluated.
The screening results in Chapter 2 identify promising greening entrainers. For the methylcyclohexane–toluene mixture, gamma-valerolactone (GVL), n-butyl-2-pyrrolidone (NBP), and dimethyl isosorbide (DMI) were selected for VLE data experimental evaluation, while NBP, DMI, and guaiacol were selected for the n-hexane–ethanol mixture. Some NADESs and other biobased and greener organic entrainers exhibited favorable predicted selectivity but showed immiscibility issues in the investigated mixtures and were therefore excluded from further evaluation. In Chapters 3 and 4, the measured VLE data for both mixtures containing each entrainer passed the thermodynamic consistency test, indicating the reliability of the data. The VLE data confirmed that the selected greener entrainers significantly increase the relative volatility of the mixtures and effectively remove close-boiling and azeotropic behavior. Process simulations discussed in Chapter 5 indicate that for the separation of methylcyclohexane-toluene, GVL reduces total annual cost (TAC) by 5.6% compared to NMP, while also having a lower energy intensity of 6.1%. In contrast, DMI and NBP exhibit slightly higher TAC values of 6.2% and 15.2%, respectively, along with slightly higher energy intensity of 3.0% and 18.2%. For CO2 emissions and water consumption, GVL remains comparable to NMP, with both exhibiting values of 0.04-0.05 kgCO2/kgproducts and 0.05 m3water/kgproducts. For the n-hexane-ethanol mixture, NBP shows a comparable TAC, with 1.6% slightly higher than that of NMP, with a 3.1% increase in energy intensity. Meanwhile, guaiacol and DMI demonstrate increasing TAC by 10.9% and 14.0%, respectively, along with energy intensity that are 12.5% and 18.8% higher. In addition, CO2 emissions for these options remain comparable at around 0.05-0.06 kgCO2/kgproducts and water consumption is similar at approximately 0.05-0.07 m3water/kgproducts. Importantly, these greener entrainers have significantly lower toxicity profiles than NMP. Furthermore, the TAC deviation threshold of 32% from predictive models compared to the experimental-based NRTL model makes them suitable for early-stage entrainer screening. However, experimental validation remains necessary for detailed process design.
Overall, this thesis provides a systematic and experimentally validated framework for selecting greener entrainers in extractive distillation. The work delivers new and reliable VLE data along with optimum binary interaction parameters and a comparison of process-level performance. The findings demonstrate that replacing conventional entrainers, such as NMP, with greener alternatives is technically feasible, economically viable, and environmentally sustainable. This supports the transition toward more eco-efficient extractive distillation processes.