Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Biomedical and Chemical Engineering and Sciences

First Advisor

Toufiq Reza

Second Advisor

Pavithra Pathirathna

Third Advisor

Jonathan Whitlow

Fourth Advisor

Manolis Tomadakis


Light naphtha is a crucial chemical feedstock that is used to make value-added products from plastic to fuels. Whether it is derived from fossil fuels or renewable resources, naphtha typically contains small fraction of aromatic contaminants such as benzene, toluene, and xylene (BTX). Polar solvents such as sulfolane, ethylene glycol, and glycerol are industrially used to separate BTX from naphtha. Each of these solvents has some processing difficulties though, whether it be sulfolane degrading and corroding equipment or ethylene glycol and glycerol not being particularly selective for BTX over light naphtha. Some of the traditional solvents used for naphtha purification can be mixed with new class of solvents to overcome these difficulties though. For instance, Glycerol (Gly) and ethylene glycol (EG), acting as hydrogen bond donors (HBD), can be combined with a hydrogen bond accepting (HBA) salt such as methyltriphenylphosphonium bromide (METPB) to form Deep Eutectic Solvents (DES). DES do not have the same shortcoming as traditional BTX solvents in that they are liquids at low temperatures, have negligible vapor pressures, and do not leach into cyclohexane if the polarity of HBA and HBD are favorable. DES properties can even be tuned by changing the ratio of the HBD and HBA components. Selecting the correct DES for a specific application e.g., BTX removal from naphtha though can be difficult as there are thousands of potential candidate DES to choose from. Thus, a methodology for selection and evaluation of solvents is needed. Therefore, screening solvents using computation methods was the first step used to evaluate potential DES. This study utilized density functional theory (DFT) modeling with a Conductor Screening Model (COSMO) to select candidate DES for BTX removal from model light naphtha (cyclohexane). DES such as EG and tetrabutylammonium bromide were modeled and the liquid-liquid equilibrium (LLE) with a model naphtha phase of 2-10 mass% benzene in cyclohexane was predicted. DES were made from a combination of various HBAs and HBDs. HBDs considered were ethylene glycol and glycerol, given that they are traditional solvents used for BTX removal. HBAs considered were tetrabutylammonium bromide (N4444Br), tetrahexylammonium bromide (N6666Br), choline chloride (ChCl), and methyltriphenylphosphonium bromide (METPB). This resulted in 4 EG based DES and 4 Gly based DES, all of which were considered at their respective eutectic ratios. COSMO was used to model these candidate DES to first determine cyclohexane and BTX solubility and then LLE of the ternary system (benzene-cyclohexane-DES). Of the evaluated DES those containing ChCl performed worse than pure HBD, having less solubility for benzene, and were not considered for evaluation by a process model. N6666Br containing DES had high solubility for cyclohexane and were also not considered for evaluation by a process model. These DES would leach a majority of model naphtha into the solvent phase and lead to poor separation. DES pairs of METPB:EG, METPB:Gly, N4444Br:EG, N4444Br:Gly were evaluated experimentally to determine the accuracy of model predictions for LLE equilibrium at initial conditions of a 1:1.5 DES:naphtha mass ratio, 25°C, atmospheric pressure, and with initial benzene mass%’s ranging from 2-10%. Cyclohexane was used as a model naphtha compound. Overall, the model predictions for EG-based DES were very accurate, with root-mean-square deviations (RMSD) below 1% for both N4444Br:EG and METPB:EG. The glycerol systems were less accurately modeled, with RMSD’s of 4% for N4444Br:Glycerol and 6% for METPB:Gly. At the specified conditions N4444Br:EG removed the most benzene at 30±1%. METPB:EG and N4444Br:Glycerol performed equally removing 19±2 and 20±2% respectively. METPB:Glycerol was the worst performing DES only removing 9±1% of the benzene from the organic phase. In addition to certifying model accuracy, mass transfer kinetics were determined experimentally for these 4 DES. Half-life times, or the time it takes to reach halfway to equilibrium concentration of benzene, were determined for each DES. N4444Br:EG had the shortest time at 3.5 minutes, followed by METPB:Gly at 3.6 minutes. N4444Br:Gly and METPB:EG had longer half-life times at 7.5 and 8.1 minutes respectively. These mass transfer kinetics and equilibrium were results were then used in a process model to evaluate DES performance in removing benzene from model naphtha. A basic process model involving serial mixing, phase separation, and vacuum flashing was used to evaluate DES made of N4444Br and MEPTB paired with EG and glycerol to purify a 1 tonne/hr stream of model naphtha containing 10 mass% benzene. A series of 7 mixing and phase separation units were used to remove benzene from the model naphtha. These had a 1:1.5 DES:naphtha feed ratio (overall ratio of 4.67:1) and operated at 25°C and 1 bar. This low temperature and pressure in the absorption unit is one of the advantages of DES, as traditional processes operate at 120°C and elevated pressures. The vacuum flash unit operated at 45°C and 0.01 bar and was used to remove the extracted benzene and leached cyclohexane from the DES before it was recycled. Notably none of the evaluated DES leached into the naphtha product (from the absorption unit) or the benzene product (from the vacuum flash unit) which is common in traditional processes. N4444Br containing DES removed most of the benzene in the initial stream but also leached up to half of the starting cyclohexane away from the naphtha product phase. N4444Br removed the most benzene of the DES at 92% while leaching 47% of the initial cyclohexane. METPB:EG removed 71.6% of the initial benzene but leached far less cyclohexane at only 10 mass% and was the best performing DES. These results are promising and show the methodologies used for selected DES are useful in both eliminating poor candidates and accurately predicting the performance of candidate solvents. Future works will consider a wider array of initial solvents as well a more comprehensive process model to evaluate candidate solvents more completely.


Copyright held by author