Degummed soybean oil was obtained from Northwood Mills (Northwood, ND, USA). A typical fatty acid profile for this oil is provided in supplemental Table (S1). The cracking catalyst was commercial ZSM-5 (CVV5524G, Zeolyst International, Conshohocken, PA, USA) in NH₄ form, having a Si/Al ratio of 50 and a surface area of 425m²/g. The catalyst was calcined as described in section 2.2.1, below. The alkylation catalyst was beta zeolite (CP814E, Zeolyst International, Conshohocken, PA, USA) having an Si/Al ratio of 25 and an Na₂O concentration of 0.05 wt%. The surface area of the catalyst was 680m²/g. Sulfolane with a purity of 97 wt% was used to extract the BTX out of the catalytically cracked soybean oil. Benzene, toluene, o-, m-, and p-xylene were obtained from Sigma–Aldrich (St. Louis, MO, USA). The reaction gases, propylene, hydrogen and nitrogen at 99.9% purity, were obtained from Praxair, Inc. (Danbury, CT, USA).

Table S1
Typical soybean oil fatty acid profile.

Gases used for gas chromatography (GC) analysis were obtained from Praxair (Praxair, Inc., Danbury, CT, USA) at 99.999% or greater purity. For detailed quantitative characterization of samples by gas chromatography, a number of standards were used. For identification, the following standard mixtures were purchased from Supelco (Bellefonte, PA, USA): isoparaffin-, aromatic-, naphthene-, and olefin-alphagaz PIANO [32ASTM D5134-98 Standard Test Method for Detailed Analysis of Petroleum Naphthas through n-Nonane by Capillary Gas Chromatography., 2008, ]; naphtha, reformate, and alkylate qualitative reference standards; petroleum crude qualitative and quantitative standards [33ASTM D5307-97 Standard Test Method for Determination of Boiling Range Distribution of Crude Petroleum by Gas Chromatography, 2007, ].

For quantification of the cracking products, individual chromatographic standards of analytical grade were used representing the complete series of unbranched alkanes (C₅-C₁₈), selected alkenes (C₆, C₉, C₁₄, C₁₈), and aromatic hydrocarbons (benzene, toluene, o-xylene, m-xylene, p-xylene 1,2,4-trimethylbenzene, indane, naphthalene). Various solvents were purchased from Fischer Scientific (Waltham, MA, USA): acetonitrile (HPLC grade), methylene chloride (GC grade). N-methyl-N-trimethylsilyltrifluoracetamide (MSTFA) was used as a derivatization agent for GC analysis of carboxylic acids and alcohols (Supelco, Bellefonte, PA, USA).

Internal standard calibrations were performed with a mixture of benzene-d₆ (102.1 mg·mL^-1), 2-chlorotoluene (100.1 mg·mL^-1) and o-terphenyl (49.8 mg·mL^-1) in methylene chloride. All were purchased from Sigma-Aldrich (St. Louis, MO, USA). These standards were selected because they are representative of the classes of compounds of interest, will not be produced during the reactions, and provide signals that are unlikely to interfere with the product compound outputs.

For the quantification of acids and alcohols in derivatized samples, a calibration mixture (0.10-75.0 mg·mL^-1) was used consisting of several representative carboxylic acids (acetic, propionic, butyric, hexanoic, octanoic, decanoic, and palmitic), n-butanol, n-hexanol, 1,3-propanediol, glycerol, and n-decanol. For identification, a standard mixture consisting of C₁–C₁₆ carboxylic acids and C₁–C₁₀ alcohols was employed. Geraniol (Aldrich, St. Louis, MO, USA) and o-terphenyl were used as recovery (10 mg·mL^-1 in acetonitrile) and internal standards (50.0 mg·mL^-1 in methylene chloride), respectively.

Both the cracking and alkylation catalysts were activated by calcination. The cracking catalyst was heated at 600°C while the alkylation catalyst was heated at 450^oC, both for six hours in an oven. The catalyst was allowed to cool down to room temperature in a nitrogen atmosphere. The calcinated catalyst was then transferred to an air tight container until use.

All cracking experiments were conducted in a 500 mL volume, high temperature, high pressure batch reactor (Parr 4575 series HP/HT reactor, Moline, IL, USA) as shown in Fig. (3). The specified amount of activated catalyst added to the reaction vessel was based on the oil/catalyst weight ratio in the design of experiments for the weight of 200 mL of soybean oil. The reactor was purged with nitrogen to insure an inert environment.

Upon completion, the liquid and residual contents were measured to obtain the overall yield data. No tar was found in the liquid after collecting the product on the 500 mL batch reactor setup as all of the highly viscous material solidified in the reactor upon cooling and was mixed with the catalyst to form a residual matter. All of the residual matter was carefully collected from the agitator blades, cooling coil, thermowell, other internal parts of the reactor and the reactor vessel. The difference in the weight of the catalyst before and after the reaction was measured and assumed to be coke. While it is possible that minor levels of absorbed reagents and other material could have also been present, characterizing all of these as coke was deemed adequate for the proof of concept level study being conducted.

The concentration of BTX in the ALP was increased by selectively separating out the distillate fractions that were most likely to contain the selected aromatic hydrocarbons based on the boiling points of each component. Batch distillations were performed at atmospheric pressure in a quartz round bottom flask equipped with a distillation column and a water cooled condenser. Distillation temperatures ranged from 80 to 144^oC.

For the sulfolane extraction experiments, soybean oil was cracked at the near optimum conditions found in the initial cracking optimization study in the same reactor. The ALP was generated by batch distillation at 135^oC.

Fig. (3) Cracking reactor system.

For the process option shown in Fig. (2), sulfolane was used to extract the BTX from the ALP. Because the ALP is a unique process mixture, simple extraction experiments were required to determine the partitioning coefficients of BTX into sulfolane from the ALP. The sulfolane extraction experimental setup replicated a mixer/settler arrangement. All experiments were conducted in 20mL volume test tubes and at atmospheric pressure. A sonicator (Fisher Scientific model FS60H, Waltham, MA, USA) was used as the mixer and a centrifuge (Centrific Model 228, Fisher Scientific, Waltham, MA, USA) was used as the settler.

Five grams of ALP were added to each of six test tubes. The proper amount of sulfolane was added to each test tube for the three solvent- to-solute ratios and mixed in the sonicator for 10 minutes. The sonicator internal heater was used to maintain the extraction temperature at the desired condition (30, 50, or 70^oC). After mixing, the test tubes were placed in the centrifuge and spun for four minutes. The two phases, ALP rich solvent (aqueous) and ALP lean solvent (organic), were collected, weighed, and stored for analysis.

The next step in this process option Fig. (2) is to separate the BTX from the sulfolane and purify the BTX fractions that will be used in the alkylation reactions. The recovery of BTX from sulfolane and the subsequent separation of the BTX product mixture into pure components are well developed and used in commercial petroleum-based aromatics production processes. Thus, it was deemed sufficient to simulate these steps using the ChemCad process simulation system as described in section 2.6, below.

The octane number was measured on a Zeltex Inc. (Hagerstown, Maryland, USA) octane and fuel analyzer (ZX-101XL) having a detection limit of 99.5 ON. Values above this point were estimated using literature values, as described in section 2.7, below. Prior to each test, the accuracy of the instrument was confirmed by calibrating the instrument to octane numbers of compounds within its detectable range, namely hexane and cyclohexane.

GC quantification was performed following the method developed by Kubatova and co-workers [34Fegade, S.; Tande, B.; Cho, H.; Seames, W.; Sakodynskaya, I.; Muggli, D.; Kozliak, E. Aromatization of propylene over HZSM-5: a design of experiments a DOE Approach. Chem. Eng. Commun., 2013, 200, 1039-1056.
[http://dx.doi.org/10.1080/00986445.2012.737385] -37Sťávová, J.; Stahl, D.C.; Seames, W.S.; Kubátová, A. Method development for the characterization of biofuel intermediate products using gas chromatography with simultaneous mass spectrometric and flame ionization detections. J. Chromatogr. A, 2012, 1224, 79-88.
[http://dx.doi.org/10.1016/j.chroma.2011.12.013] [PMID: 22245174] ]. This method uses a GC-FID/MS (Agilent 7890N GC, 5975C MS, Santa Clara, CA, USA) equipped with an autosampler (7386B series) and a split/splitless injector. Separations were accomplished using a 100 m long DB-1MS capillary column with a 0.25 mm internal diameter and 0.25 µm film thickness and a constant helium flow rate. The MS and FID data were simultaneously acquired by employing a two-way splitter with a helium makeup gas with a split flow ratio of 1:2 (MS:FID). The MS data (total ion chromatogram, TIC) was acquired in the full scan mode using electron ionization. The standards used are listed in section 2.1 and products were identified based on the NIST MS library.

Key fuel physical properties were determined using commercial (ASTM) standard methods [38ASTM D1655-11b Standard Specification for Aviation Turbine Fuels., 2011, ]. Standardized test methods were used to determine the following critical physical properties of the refined fuel: density [39ASTM D1298-99 Standard Test Method for Density, Relative Density (Specific Gravity), or API Gravity of Crude Petroleum and Liquid Petroleum Products by Hydrometer Method., 2003, ], total acid number [40ASTM D974-11 Standard Test Method for Acid and Base Number by Color-Indicator Titration., 2011, ], freeze point [41ASTM D5972-05 Standard Test Method for Freezing Point of Aviation Fuels (Automatic Phase Transition Method). , 2005, ], flash point [42ASTM D3828-05 Standard Test Methods for Flash Point by Small Scale Closed Cup Tester. , 2005, ], and heat of combustion [43ASTM D4809-00 Standard Test Method for Heat of Combustion of Liquid Hydrocarbon Fuels by Bomb Calorimeter (Precision Method)., 2005, ]. Numerous other specification tests are required to fully certify fuels, such as distillation analysis [44ASTM D86-11a Standard Test Method for Distillation of Petroleum Products at Atmospheric Pressure. , 2011, ], and the fraction of olefins in the fuel, and the fraction of aromatic hydrocarbons in the fuel [45ASTM D1319-10 Standard Test Method for Hydrocarbon Types in Liquid Petroleum Products by Fluorescent Indicator Adsorption., 2010, ]. However, these were not performed in the present work as the suite of tests that were performed was considered adequate to assess the potential of the proposed process pathways to generate a viable aviation gasoline fuel product

Catalytic cracking of TG oils using HZSM-5 to produce an aromatics-rich liquid product has been well studied [5Idem, R.; Katikaneni, S.; Bakhshi, N. Catalytic conversion of canola oil to fuels and chemicals: roles of catalyst acidity, basicity and shape selectivity on product distribution. Fuel Process. Technol., 1997, 51, 101-125.
[http://dx.doi.org/10.1016/S0378-3820(96)01085-5] -14Bertero, M.; Sedran, U. Immediate catalytic upgrading of soybean shell bio-oil. Energy, 2016, 94, 171-179.]. Therefore, complete optimization of this process step was not considered to be necessary. Instead, a series of 16 experiments were performed using the DOE scheme specified in (Table 1) in order to identify reaction conditions that would be productive for ALP generation using the available experimental facilities. For example, the cracking reaction temperature range was chosen based on results in previous work [53Kadrmas, C.; Khambete, M.; Kubátová, A.; Kozliak, E.; Seames, W. Optimizing the production of renewable aromatics via crop oil catalytic cracking. Process, 2015, 3, 222-234.
[http://dx.doi.org/10.3390/pr3020222] ]. Five factors were varied in a two-level half fractional factorial design. The order of experiments was randomized in order to mitigate the effect of any potential background variables. Soybean oil was chosen since it is the most common crop oil in the world, is readily available, and has been extensively studied in our previous work on TG cracking [35Seames, W.; Luo, Y.; Ahmed, I.; Aulich, T.; Kubátová, A.; Štávová, J.; Kozliak, E. The thermal cracking of canola and soybean methyl esters: improvement of cold flow properties. Biomass Bioenergy, 2010, 34, 939-946.
[http://dx.doi.org/10.1016/j.biombioe.2010.02.001] , 53Kadrmas, C.; Khambete, M.; Kubátová, A.; Kozliak, E.; Seames, W. Optimizing the production of renewable aromatics via crop oil catalytic cracking. Process, 2015, 3, 222-234.
[http://dx.doi.org/10.3390/pr3020222] -55Kubátová, A.; Št’ávová, J.; Seames, W.; Luo, Y.; Sadrameli, S.; Linnen, M.; Baglayeva, G.; Smoliakova, I.; Kozliak, E. Triacylglyceride thermal cracking: pathways to cyclic hydrocarbons. Energy Fuels, 2012, 26, 672-685.
[http://dx.doi.org/10.1021/ef200953d] ]. However, the overall conclusions reached should be applicable to a wide variety of TG oil feedstocks.

Table 1
Design of experiments for catalytic cracking of soybean oil.

A summary of the composition of the ALP generated at each reaction condition is shown in Table (2). It should be noted that these results exclude quantification of short chain fatty acids which have been found in previous work to be decarboxylated during soybean oil HZSM-5 catalytic cracking to concentrations on the order of <0.05 wt% of the ALP [53Kadrmas, C.; Khambete, M.; Kubátová, A.; Kozliak, E.; Seames, W. Optimizing the production of renewable aromatics via crop oil catalytic cracking. Process, 2015, 3, 222-234.
[http://dx.doi.org/10.3390/pr3020222] ].

The maximum conversion of TG oil to aromatic hydrocarbons was around 35 wt% based on the inlet mass of TG oil Table (2), Run 8). Note that this is a more rigorous measure than is often seen in the literature where the fraction of the GC elutable and quantified compounds is used, which would be a much higher value. We use fraction of inlet oil mass as it is a more accurate measure of conversion efficiency. This value was obtained at the higher cracking temperature (430^oC), atmospheric pressure, longer residence time (60 min), higher catalyst-to-reactant ratio, and in the absence of hydrogen. Runs 10, 11, 15, and 16 yielded similar results.

Within the bounds of the DOE, there was a clear positive effect leading to higher aromatics concentrations when the quantity of catalyst used (representing the catalyst-to-feedstock ratio) was increased. Reaction temperature, within the narrow range studied, had little effect. While we had postulated that hydrogen addition might inhibit aromatics formation, the results suggest that this is not the case.

Table 2
Composition results for the ALP excluding fatty acids (PAHs = polycyclic aromatic hydrocarbons, w/w%= mass of compounds in the category per total GC elutable mass in %).

Several experiments were carried out to understand the progress of the reaction and also to determine the selectivity of the catalyst being studied. Table (3) provides a summary of the alkylation reactions carried out along with the products that were identified. Please note that these proof-of-feasibility experiments provide insight into the reasonableness of the two potential process pathways and do not represent fully optimized experimental reaction data.

For toluene the most common product detected was 1-methyl-x-(1-isopropyl) benzene. Here, the “x” indicates the changing location on the benzene ring. This number can be 2, 3 or 4. 1, 2-dimethyl-x-isopropyl-benzene was the product of alkylation of o-xylene with propylene. The “x” here takes the position 3 or 4. Similarly for m- and p-xylene, the products were 2,4-dimethyl-x-(1-isopropyl) benzene and 1,4-dimethyl-x-(1-isopropyl) benzene with the positions for “x” as 2, 4 or 5 and 2 and 3, respectively. It should be noted that in all of these experiments, the aromatic stream was in stoichiometric excess which allowed us to study the selectivity of propylene for alkylation.

Table 3
The alkylation of aromatics with propylene.

The first three experiments support process pathway 1 Fig. (1). The experiments were highly selective for the target alkylated aromatic compounds and yielded reasonable propylene conversion. A 1:2:2 ratio of cumene, cymene, and isopropyl xylenes were obtained from the model ALP mixture corresponding to the BTX ratio measured from the Table (1), Run 8 experiments. These compounds were used in the fuel formulation blends and calculations described in section 3.5, below. The remaining experiments support process pathway 2 (Fig. 2). Again, the results show high selectivity to the target alkylated aromatic hydrocarbons and high propylene conversion. These results suggest that it should be possible to completely alkylate BTX, as conversion and selectivity of propylene for alkylation of each aromatic, both individually and in a mixture, was high.

Studies were performed of the conditions required to extract aromatic hydrocarbons out of the ALP generated during the cracking reactions to support process pathway 2 Fig. (2). The initial DOE set was based on a two-factor, three-level full factorial experimental design of: 1) solvent-to-solute ratio (3:1, 5:1 and 7:1) and 2) mixer temperature (30, 50, and 70°C). The response measured was extraction yield. Four replicates were performed for a total of 36 runs.

The results are shown in Fig. (4). Increasing the temperature from 30 to 50^oC resulted in a significant increase in extraction yield. However, increasing the temperature from 50 to 70^oC did not significantly improve the yield. Therefore, 50^oC was chosen as the approximate optimum temperature.

Fig. (4) The extraction yield of BTX into sulfolane as a function of temperature and solvent-to-BTX ratio.

Because the extraction yield was greatest at the higher solvent-to-BTX ratio utilized in the DOE, a parametric study was then performed with solvent-to-BTX ratios of 8:1, 9:1, 9.5:1, 10:1, 10.5:1, and 11:1 at an extraction temperature of 50^oC to predict the saturation point. Fig. (5) shows the extraction yield as a function of solvent-to-BTX ratio. As shown, saturation is reached when the solvent-to-BTX ratio approaches 9:1 with no further significant improvement at higher ratios. Therefore a value of 9:1 was used in the simulations described below to generate the process conditions for this portion of the process.

These results provide data for a single stage extraction with an optimum efficiency of around 70%. For a commercially viable process, a recovery of greater than 99% would be desirable. This can be accomplished using a multi-stage extraction system. The single stage data were used to estimate the efficiency of a multiple stage extraction following the methods and assumptions described by Shuler and Kargi [56Shuler, M.; Kargi, F. Bioprocess Engineering: Basic Concepts., (2nd ed. ), 2011, ]. This method assumes that the distribution coefficient is constant, that there is no sulfolane in the raffinate phase, and there is nothing but BTX and sulfolane in the extract phase. With these assumptions, the number of stages required to achieve a 99.5% recovery of BTX was estimated to be three.

Fig. (5) The extraction yield of BTX into sulfolane as a function of solvent-to-BTX ratio.

Recovery of the BTX out of the sulfolane solvent and subsequent purification were studied using process simulation. Cumene was found to represent almost all of the roughly 0.12 wt% non-BTX aromatic hydrocarbons present in the sulfolane and thus it is assumed that all alkylated aromatic hydrocarbons recovered in the solvent were cumene. Assuming a sulfolane-to-BTX ratio of 9:1, the composition of the extract stream from the sulfolane extractor was estimated as shown in Table (4). This composition was used as the feed for the first distillation column.

Table 4
Estimated Aromatics-Rich Solvent Extractant Composition.

Three different configurations were modeled with the most efficient configuration shown in Fig. (2). In this scheme, the first column separates the aromatic hydrocarbons from the sulfolane, allowing the sulfolane to be recycled back to the extractor. In the second column benzene is recovered as the light key to a purity of 99.2%. The third column separates toluene as the light key to a purity of 99.0%, and the last column separates the o- and p-xylenes from the m-xylene and ethylbenzene.

The key design parameters for the optimum simulated system are shown in Table (5). It should be noted that the separation of ethylbenzene from xylene is challenging in traditional distillation, requiring a very large number of separation stages and a high reflux ratio. Even with these rigorous conditions, it is not possible to recover all of the m-xylene. Most of the contaminants in the xylene stream are heavier aromatic hydrocarbons (mostly cumene) which cannot be practically separated from the xylene at this point in the process.

Table 5
Key Design parameters of the figure 1b 4-column configuration to recover and Purify BTX from sulfolane extract.

The key fuel properties from AvGas blend formulations based on the alkylated aromatic hydrocarbons generated during the study are shown in Table (6). Cumene, cymene, and dimethylisopropyl benzene were used in the theoretical blend (CCD) assuming that the same ratio of benzene, toluene, and xylenes were alkylated as were obtained experimentally from catalytically cracking the soybean oil and concentrating the reaction products, as described in section 2.3, to generate the ALP shown in Table 2 as Run 8. Under these conditions, a fuel with an octane number of 96.2 and with acceptable flash point, freeze point, and heat of combustion is projected.

To verify the accuracy of this theoretical result, the ALP from Table (6), Run 8 was alkylated. After alkylation, the product was distilled to separate out the unconverted reactants. The resulting alkylated ALP had an octane number that exceeded the limit of the analyzer (99.5 ON), a flash point of 59.5^oC, and a freeze point below the lower limit of the analyzer (< -75^oC). The difference from the theoretical values may be due to the assumption of linear blending used in the theoretical calculations and/or due to trace quantities of BTX, ethylbenzene, or other hydrocarbons in the experimental mixture. The similarity of the experimental and theoretical results suggests that it is likely feasible to produce a high octane AvGas via a two stage cracking-alkylation pathway from fatty acid-based oils, such as soybean oil.

Table 6
Key Fuel properties for alkylated aromatic blends for avGas production.

To improve the octane number further, a second scenario was considered as shown in the final row of Table (6). In this case cumene, cymene, and 1,2,4-trimethylbenzene (TMB) were used to form a simulated blend (CCT), assuming the same ratios of benzene, toluene, and xylenes as used above. Previous studies [57Nishi, H.; Nowinska, K.; Moffat, J. The alkylation of toluene with methanol on microporous heteropoly oxometalates. J. Catal., 1989, 116, 480-487.
[http://dx.doi.org/10.1016/0021-9517(89)90114-0] ] suggest that the alkylation of xylenes with methanol yielded TMB which has a reported octane number of 110. Substituting the higher octane number TMB for DIMPD, which has a reported octane number of 95, increased the octane number of the fuel blend to 103.2 while maintaining the flash point and lower freeze point within the range of acceptable values for these types of fuels. This higher octane blend could be attractive for those cases where the fuel producer wants to blend in lower octane fuel compounds, such as linear alkanes, while still meeting the fuel specifications of 100LL AvGas.